SAND2022 16425 HyRAMplus 5.0 Technical Reference Manual

SANDIA REPORT
SAND2022-16425
Printed November 2022
Hydrogen Plus Other Alternative Fuels Risk

Assessment Models (HyRAM+) Version 5.0
Technical Reference Manual
Brian D. Ehrhart and Ethan S. Hecht
Prepared by
Sandia National Laboratories
Albuquerque, New Mexico 87185
Livermore, California 94550
Issued by Sandia National Laboratories, operated for the United States Department of Energy by National
Technology & Engineering Solutions of Sandia, LLC.
NOTICE: This report was prepared as an account of work sponsored by an agency of the United States Government.
Neither the United States Government, nor any agency thereof, nor any of their employees, nor any of their
contractors, subcontractors, or their employees, make any warranty, express or implied, or assume any legal liability
or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process
disclosed, or represent that its use would not infringe privately owned rights. Reference herein to any specific
commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not necessarily
constitute or imply its endorsement, recommendation, or favoring by the United States Government, any agency
thereof, or any of their contractors or subcontractors. The views and opinions expressed herein do not necessarily
state or reflect those of the United States Government, any agency thereof, or any of their contractors.
Printed in the United States of America. This report has been reproduced directly from the best available copy.
Available to DOE and DOE contractors from
U.S. Department of Energy
Office of Scientific and Technical Information
P.O. Box 62
Oak Ridge, TN 37831
Telephone: (865) 576-8401

Facsimile: (865) 576-5728
E-Mail: [email protected]
Online ordering: http://www.osti.gov/scitech
Available to the public from
U.S. Department of Commerce
National Technical Information Service
5301 Shawnee Road
Alexandria, VA 22312
Telephone: (800) 553-6847

Facsimile: (703) 605-6900
E-Mail: [email protected]
Online order: https://classic.ntis.gov/help/order-methods
NT OF E
ME N
RT
ER
A
DEP
GY
• •
IC A
U NIT
ER
ED
ST M
A TES OF A
2
ABSTRACT
The HyRAM+ software toolkit provides a basis for conducting quantitative risk assessment and
consequence modeling for hydrogen, natural gas, and autogas systems. HyRAM+ is designed to
facilitate the use of state-of-the-art models to conduct robust, repeatable assessments of safety,
hazards, and risk. HyRAM+ integrates deterministic and probabilistic models for quantifying leak
sizes and rates, predicting physical effects, characterizing hazards (thermal effects from jet fires,
overpressure effects from delayed ignition), and assessing impacts on people. HyRAM+ is
developed at Sandia National Laboratories to support the development and revision of national
and international codes and standards, and to provide developed models in a publicly-accessible
toolkit usable by all stakeholders.
This document provides a description of the methodology and models contained in HyRAM+
version 5.0. The most significant change for HyRAM+ version 5.0 from HyRAM+ version 4.1 is
the ability to model blends of different fuels. HyRAM+ was previously only suitable for use with
hydrogen, methane, or propane, with users having the ability to use methane as a proxy for
natural gas and propane as a proxy for autogas/liquefied petroleum gas. In version 5.0, real
natural gas or autogas compositions can be modeled as the fuel, or even blends of natural gas with
hydrogen. These blends can be used in the standalone physics models, but not yet in the
quantitative risk assessment mode of HyRAM+.
3
ACKNOWLEDGMENT
This report describes HyRAM+, a revised version of the HyRAM code, which itself was built
upon the previous work of many others. The authors therefore especially wish to thank Katrina
Groth (now at the University of Maryland) for her leadership and contributions to the HyRAM
project, as well as John Reynolds, Greg Walkup, and Erin Carrier for their contributions to
HyRAM. The authors also wish to thank Ben Schroeder and Ben Liu of Sandia, Cianan Sims of
Sims Industries, and Jeremey Rifkin for their contributions to the HyRAM+ code. This work was
supported by the U.S. Department of Energy (DOE) Office of Energy Efficiency (EERE)
Hydrogen and Fuel Cell Technologies Office (HFTO), the DOE EERE Vehicle Technologies
Office (VTO), and the U.S. Department of Transportation Pipeline and Hazardous Materials
Safety Administration (PHMSA). The authors gratefully acknowledge Myra Blaylock, Chris
LaFleur, Dusty Brooks, and Alice Muna at Sandia for helpful discussions and useful insights. The
authors wish to thank Ben Schroeder, Laura Hill, and Christine Watson for their review of this
document. The authors also wish to specifically thank the members of NFPA 2, the Hydrogen
Safety Panel, and HySafe for engaging technical discussions and thoughtful feedback. Finally, the
authors gratefully acknowledge the many productive discussions with users, other researchers,
and the various stakeholders who have provided insight and feedback for this work.
4
CONTENTS
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.1. About HyRAM+ and This Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2. Design Goals and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3. Summary of HyRAM+ Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4. Summary of Changes Made . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2. Quantitative Risk Assessment Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.1. Quantitative Risk Assessment Methodology Overview . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2. Risk Metrics Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3. Scenario Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1. Default Detection and Isolation Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.2. Default Ignition Probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4. Frequency of a Release . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.1. Frequency of Random Leaks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.2. Default Component Leak Frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.3. Frequency of Dispenser Releases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.4. Default Dispenser Failure Probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5. Consequence Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5.1. Facility Occupants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5.2. Detection and Isolation Scenario Consequences . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.3. No Ignition Scenario Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.4. Jet Fire Scenario Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.5. Overpressure Scenario Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6. Harm and Loss Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6.1. Thermal Harm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6.2. Overpressure Harm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3. Physics Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1. Properties of the Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.1. Equation of State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.2. Combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2. Developing Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.1. Orifice Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.2. Notional Nozzles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.3. Initial Entrainment and Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.4. Establishment of a Gaussian Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3. Unignited Releases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.1. Gas Jet/Plume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.2. Tank Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3.3. Tank Emptying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3.4. Accumulation in Confined Areas/Enclosures . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4. Ignited Releases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4.1. Flame Correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5
3.4.2. Jet Flame with Buoyancy Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.4.3. Radiation From a Curved Flame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.4.4. Overpressure in Enclosures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4.5. Unconfined Overpressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4. Summary of Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.1. Python Calculation Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2. Leak Frequency Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3. Unit Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
LIST OF FIGURES
Figure 2-1. Event sequence diagram used by HyRAM+ for flammable gas releases [4]. . . . . 15
Figure 2-2. Fault tree for random leaks of size 0.01% from components. . . . . . . . . . . . . . . . . . 17
Figure 2-3. Box and whisker plots of gaseous hydrogen (top) and liquid hydrogen (bottom)
leak frequencies for the different components and each leak size. The thick
central line is the median leak frequency, the boxes show the 25th and 75th
percentiles of the distribution and the whiskers show the 5th and 95th percentiles. 19
Figure 2-4. Box and whisker plots of gaseous methane (top) and liquid methane (bottom)
leak frequencies for the different components and each leak size. The thick
central line is the median leak frequency, the boxes show the 25th and 75th
Figure 2-5. Box and whisker plot of propane (same default values for both gaseous and
liquid) leak frequencies for the different components and each leak size. The
thick central line is the median leak frequency, the boxes show the 25th and 75th
Figure 2-6. Fault tree for Other Releases from a dispenser [4]. . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Figure 3-1. Graphical representations of state points, calculated using CoolProp [30] which
uses the Leachman et al. [31] equation of state for hydrogen. Top plots show
shading and iso-contours of density as a function of temperature and pressure.
Bottom plot shows shading of density as a function of entropy and tempera-
ture, with iso-contours of pressure and enthalpy. The thick black line shows the
liquid/two-phase/vapor boundary, and the black dots mark the triple point and
critical points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Figure 3-2. Temperature and density of products for the combustion of 298 K, 101,325 Pa
fuels as a function of mixture fraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Figure 3-3. Sketch of plume model coordinates. Gravity acts in the negative y-direction. . . . 42
Figure 3-4. Mapping of scaled distance to scaled overpressure (left) and scaled impulse
(right) for the BST unconfined overpressure model [2]. . . . . . . . . . . . . . . . . . . . . . 51
Figure 3-5. Mapping of scaled distance to scaled overpressure (left) and scaled impulse
(right) for the TNT equivalence unconfined overpressure model [2]. . . . . . . . . . . 52
6
Figure 3-6. Detonation cell size data (points) from the detonation database [69] and fits to
the data (lines) on a linear (left) and logarithmic (right) scale. . . . . . . . . . . . . . . . 53
LIST OF TABLES
Table 2-1. Default ignition probabilities for different fuels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Table 2-2. Default parameters for frequency of random leaks for individual components. . . . 22
Table 2-3. Default probability distributions for component failure. . . . . . . . . . . . . . . . . . . . . . . 26
Table 2-4. Default probability distributions for accident occurrence. . . . . . . . . . . . . . . . . . . . . . 26
Table 2-5. Probit models used to calculate fatality probability as a function of thermal dose
(V ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Table 2-6. Probit models to calculate fatality probability from exposure to overpressures,
where Ps is peak overpressure (Pa) and i is the impulse of the shock wave (Pa·s). . 31
Table 3-1. Detonation cell size fitted parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Table 4-1. HyRAM+ convertible units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7
ACRONYMS AND DEFINITIONS
AIR average individual risk
CFD computational fluid dynamics
DOE U.S. Department of Energy
EERE Office of Energy Efficiency and Renewable Energy
ESD event sequence diagram
FAR fatal accident rate
FTC failure to close
FTO failure to open
HFTO Hydrogen and Fuel Cell Technologies Office
HSE U.K. Health and Safety Executive
HyRAM Hydrogen Risk Assessment Models
HyRAM+ Hydrogen Plus Other Alternative Fuels Risk Assessment Models
MW molecular weight
NFPA National Fire Protection Association
PHMSA Pipeline and Hazardous Materials Safety Administration
PLL potential loss of life
PRD pressure relief device
QRA quantitative risk assessment
SI International System of Units
TNO Netherlands Organisation for Applied Scientific Research
TNT trinitrotoluene
VTO Vehicle Technologies Office
8
1. INTRODUCTION
1.1. About HyRAM+ and This Report
Hydrogen Plus Other Alternative Fuels Risk Assessment Models (HyRAM+) is a software toolkit
that integrates data and methods relevant to assessing the safety of the delivery, storage, and use
infrastructure of hydrogen and other alternative fuels (i.e., natural gas and propane). The
HyRAM+ risk assessment calculation incorporates generic probabilities for component failures
for both compressed gaseous and liquefied fuels, as well as probabilistic models for the effect of
heat flux and overpressure on humans. HyRAM+ also incorporates experimentally validated
models of various aspects of release and flame behavior. The HyRAM+ toolkit can be used to
support multiple types of analysis, including code and standards development, safety basis
development, and facility safety planning.
This report provides technical documentation of the algorithms, models, and data incorporated in
HyRAM+ version 5.0. HyRAM+ version 5.0 is a revised version of the software, and this report
has a lot of similar content due to similarities between the previous versions of this software, as
well as several versions of the previously named Hydrogen Risk Assessment Models (HyRAM)
software. This report and the HyRAM+ software builds off of the models and implementations
from those earlier versions [1–6].
HyRAM+ is free and open source software: you can redistribute it and/or modify it under the
terms of the GNU General Public License version 3, as published by the Free Software
Foundation. This program is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE. See the GNU General Public License for more details
(https://www.gnu.org/licenses/gpl-3.0.html).
1.2. Design Goals and Limitations
HyRAM+ is designed to calculate multiple risk and harm/damage metrics from user-defined
system configurations to provide insights for stakeholders in the safety, codes, and standards
community [6–9]. HyRAM+ contains default inputs and fast-running, reduced-order models
designed to facilitate comparison of different system designs and requirements. Reduced-order
models use simplifying assumptions to approximate behavior much more quickly than more
complex high-fidelity models. As such, the focus of HyRAM+ is on enabling systematic,
defensible risk and consequence assessments for use in risk comparisons and sensitivity analyses,
rather than a focus on absolute model accuracy.
Risk and safety assessment results should be used as part of a decision-making process, not as the
sole basis for a decision. Safety and design decisions involve consideration of many factors and
judgments; these factors include the safety assessments, the assumptions and limitations of safety
assessments, the benefits of a technology, and public preferences. As such, HyRAM+ does not
allow the user to specify an acceptability or tolerability criteria for risk or harm.
9
HyRAM+ is designed to enable defensible, repeatable calculations using documented algorithms.
The algorithms, models, and data in HyRAM+ have been assembled from published, publicly
available sources. The physics models contained in HyRAM+ have been validated against
available experimental data [10–13]. Where model validation is lacking (e.g., for harm models),
HyRAM+ often allows users to choose among different models. HyRAM+ includes default data
for component leak frequencies, which are in most cases specific to the fuel. HyRAM+ also
includes documented, expert-assigned probabilities for ignition. HyRAM+ is designed to allow
users to replace the default parameters (e.g., leak frequency distribution parameters for every
component) and assumptions with system-specific information when such information is
available to the user.
1.3. Summary of HyRAM+ Outputs
The quantitative risk assessment (QRA) mode in HyRAM+ can be used to calculate risk metrics
which are commonly used to evaluate fatality risk in multiple industries [6–9], as well as other
incident frequency information:
• Fatal Accident Rate (FAR), the expected number of fatalities in 100 million exposed hours
• Average Individual Risk (AIR), the expected number of fatalities per exposed individual
• Potential Loss of Life (PLL), the expected number of fatalities per system-year
• Expected number of releases per system-year (unignited and ignited cases)
• Expected number of jet fires per system-year (immediate ignition cases)
• Expected number of overpressures per system-year (delayed ignition cases)
The physics mode of HyRAM+ can be used to calculate multiple physical effects associated with
hydrogen, methane, propane, and blends, including:
• Concentration for an unignited plume
• Jet flame temperature and trajectory
• Jet flame radiative heat flux
• Time histories of concentration, flammable mass, and overpressure due to accumulation and
delayed ignition in an enclosure
• Overpressure from the delayed ignition of a jet/plume
1.4. Summary of Changes Made
HyRAM+ is an open source software, meaning that changes to the code can be observed
directly1 . However, a summary of major changes to the models will be included here for ease of
reference.
The most significant change for HyRAM+ version 5.0 from HyRAM+ version 4.1 is the ability to
model blends of different fuels in the Physics mode. HyRAM+ Physics mode was previously only
suitable for use with hydrogen, methane, or propane, with users having the ability to use methane
1 More detailed changes are given in the source code changelog: https://github.com/sandialabs/hyram/blob/
master/CHANGELOG.md
10
as a proxy for natural gas and propane as a proxy for autogas/liquefied petroleum gas. In version
5.0, real natural gas or autogas compositions can be modeled as the fuel in the Physics mode, or
even blends of natural gas with hydrogen.
Other changes include:
• Changed the output of the dispersion model so that the contour is labeled directly with its
value
• Fixed a bug that was causing notional nozzle calculations to be made for unchoked flows
• Changed units from Pa to kPa for peak overpressure and impulse plots and tables
• Moved Engineering Toolkit calculations into Physics mode, rather than a separate section
• Added ability to calculate steady-state releases for accumulation in graphical frontend
• Changed Tank Mass calculation to calculate temperature, pressure, volume, or mass if
given the other three inputs
• Random seed for occupant location distributions now generates a new value each time GUI
is launched, but it is still editable by user
• Changed GUI save/load functionality to utilize JSON file format instead of binary, so that
save files can now be modified directly via text editor
• Updated default component leak frequency values for gaseous methane based on recent
statistical effort using an updated data set
• Changed default BST Mach Flame Speed to 0.35
• Changed default Thermal Exposure Time in QRA to 30 seconds
• Added HyRAM+ Python module to PyPI and Conda-Forge repositories as "hyram" for ease
in installation of Python only package (first supported in HyRAM+ version 4.1.1)
11
This page intentionally left blank.
12
2. QUANTITATIVE RISK ASSESSMENT METHODOLOGY
In a QRA, multiple models are used together to provide a framework for reasoning about decision
options, based on the input information used in those models. HyRAM+ includes a QRA mode,
which provides calculations and models relevant to estimating the risk of systems for hydrogen
and other alternative fuels. The consequences and fatality risk of jet flames and overpressures are
estimated for different release sizes, each of which can be predicted to occur with different
frequencies [6–9, 14]. These calculations use a subset of the models available in physics mode to
estimate the release behavior (see Section 3).
2.1. Quantitative Risk Assessment Methodology Overview
Risk is characterized by a set of hazard exposure scenarios (i), the consequences (ci ) associated
with each scenario, and the probability of occurrence (pi ) of these consequences. One commonly
used general expression for calculating risk is shown in Equation 1 [5, 14].
Risk = ∑(pi × ci ) (1)

i
In a QRA, the consequences are expressed in terms of an observable quantity, such as number of
fatalities or repair cost in a specific period of time. In HyRAM+, the number of fatalities is used
as the safety metric of interest. The general expression shown in Equation 1 shows risk as a
product of probability; depending on the application, risk may be characterized using either
probability or frequency. The rest of this section will describe the specific implementation for risk
calculations in HyRAM+.
2.2. Risk Metrics Calculations
There are multiple metrics used to express the fatality risk for a system under consideration. One
such metric is the Potential Loss of Life (PLL), which expresses the expected number of fatalities
per system-year. The PLL calculation is shown in Equation 2, where n is one of the possible
safety-significant scenarios (described in Section 2.3), fn is the frequency of that accident
scenario n (described in Sections 2.3 and 2.4), and cn is the expected number of fatalities for
accident scenario n (described in Section 2.5) [6, 9].
PLL = ∑( fn × cn ) (2)
n
Another risk metric related to the PLL is the Fatal Accident Rate (FAR), which is the expected
number of fatalities in a group, per 100 million exposed hours. The FAR for a particular facility
can be calculated using the PLL and the population of the facility [9]. The FAR is calculated
using Equation 3, where N pop is the average number of personnel in the facility, and dividing by
8760 converts from years to hours (24 hours per day and 365 days per year) [6, 14].
13
PLL × 108 PLL × 108
FAR = = (3)
Exposed hours N pop × 8760
The third metric used in HyRAM+ is the Average Individual Risk (AIR), which expresses the
average number of fatalities per exposed individual. It is based on the number of hours the
average occupant spends at the facility [9]. The AIR is calculated using Equation 4, where H is
the annual number of hours the individual spends in the facility (e.g., 2000 hours for full-time
worker) [6, 14].
AIR = H × FAR × 10−8 (4)
2.3. Scenario Models
A release of a flammable fuel could lead to several different physical consequences and
associated hazards. For continuous releases of a fuel, the physical consequences are unignited
releases, jet fires (thermal effects), flash fires or fireballs (deflagration of accumulated gas
dominated by thermal effects), and explosions (deflagration or detonation of accumulated gas
dominated by overpressure effects) [6, 8, 9]. Currently, HyRAM+ calculates harm from thermal
effects of jet fires (for immediate ignition) and overpressure (for delayed ignition). A release of a
liquid fuel (e.g., liquid hydrogen or liquid natural gas) may also form a pool on the ground, but
this is currently not considered in HyRAM+, nor are the thermal effects from the cold
temperatures of these liquid fuels. These scenarios are modeled using an event sequence diagram
(ESD) for release of a flammable fuel (see Figure 2-1).
14
Leak Detected
Leak Shutdown
and Isolated
No Ignition No Ignition
Immediate
Jet Fire
Ignition
Explosion
Figure 2-1 Event sequence diagram used by HyRAM+ for flammable gas releases [4].
The ESD is implemented in HyRAM+ using Equations 5–8 where fRelease is the annual frequency
of a release (see Section 2.4), pIsolated is the probability of release (leak) detection and isolation
before ignition (see Section 2.3.1), pImmed. Ignite is the probability of immediate ignition (see
Section 2.3.2), and pDelayed Ignite is the probability of delayed ignition (see Section 2.3.2) [4].
fIsolated = fRelease × pIsolated (5)
fUnignited = fRelease × (1 − pIsolated ) × (1 − pImmed. Ignite − pDelayed Ignite ) (6)
fJetfire = fRelease × (1 − pIsolated ) × pImmed. Ignite (7)
fExplosion = fRelease × (1 − pIsolated ) × pDelayed Ignite (8)
These equations are defined differently than is typically seen for an ESD, in which each top level
event (split) in the ESD is defined with a probability and complementary probability. These
equations are written in this way to utilize immediate and delayed ignition probabilities that are
each defined as conditional to a leak occurring (see Section 2.3.2). Thus, there are three possible
ignition outcomes, and so a top level event was added to the ESD for immediate vs. delayed
ignition for ease of viewing, even if the equations are written so that the ignition probabilities in
Section 2.3.2 can be used directly.
15
2.3.1. Default Detection and Isolation Probability
The default value for successful detection and isolation of a release (pIsolate ) is 0.9 [6]. This value
incorporates many considerations on detection and isolation including: ventilation, sensor
placement, leak location, and the ability of the sensor and isolation valve to operate successfully
on demand.
Note: This value can vary significantly based on a particular system setup, and so the
user/analyst needs to carefully consider the particulars of the system being assessed and decide if
this default value is appropriate.
2.3.2. Default Ignition Probabilities
The default ignition probabilities are a function of release rate and are given in Table 2-1 [15]. It
should be noted that both the immediate and delayed ignition probabilities are both relative to a
release rate; the delayed ignition probability is not conditional upon the immediate ignition
having not occurred. The total probability of ignition is the immediate and delayed ignition
probabilities added together.
Table 2-1 Default ignition probabilities for different fuels.
(a) Hydrogen (b) Methane and Propane
Release Ignition Probability Release Ignition Probability

Rate (kg/s) Immediate Delayed Rate (kg/s) Immediate Delayed
<0.125 0.008 0.004 <1 0.007 0.003
0.125–6.25 0.053 0.027 1–50 0.047 0.023
>6.25 0.230 0.120 >50 0.200 0.100
2.4. Frequency of a Release
HyRAM+ estimates the annual frequency of a release for release sizes of 0.01%, 0.1%, 1%, 10%,
or 100% of pipe flow area [6–9]. These release sizes are relative to the pipe flow area (A) as shown
in Equation 9, where Cd is the discharge coefficient, and d is the inner diameter of the pipe.
π
A = Cd d 2 (9)
4
The annual frequency of a release for each of the four smallest releases sizes ( fRelease,k , k =
0.01%, 0.1%, 1%, and 10%) is only due to the annual frequency of random leaks
( fRandom Releases,k , see Section 2.4.1), as shown in Equation 10 [6, 14].
16
fRelease,k = fRandom Releases,k (10)
The annual frequency of the largest release size (100%) is due to both random leaks (see
Section 2.4.1) and other releases (see Section 2.4.3), as shown in Equation 11 [6].
fRelease, k=100% = fRandom Releases, k=100% + fOther Releases (11)
2.4.1. Frequency of Random Leaks
The annual frequency of random leaks is obtained for each release size using a fault tree [4]. As
an example, the fault tree for random leaks for leak size 0.01% is shown in Figure 2-2. The fault
trees for random leaks for all other leak sizes are analogous per leak size.
0.01% Leak
OR
Heat Loading Component

Compressor Vessel Valve Instrument Joint Hose Pipe (1 m) Filter Flange Vaporizer Component
Com Com Com Com Com Com Com Com Com Exchanger
Com Com Arm
Com Com
1 Com
2
0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak 0.01% Leak
0.01% Leak 0.01% Leak 0.01% Leak
Figure 2-2 Fault tree for random leaks of size 0.01% from components.
These fault trees are implemented in HyRAM+ to combine individual component leak
frequencies into an overall system leak frequency for each leak size. The annual frequency of
random leaks ( fRandom Releases,k ) is calculated using Equation 12 for each release of size k by
combining the individual component leak frequencies for all the components in the system of
interest, where NComponenti is the number of components of each type and fLeaki,k is the leak
frequency of size k for component i (see Section 2.4.2) [6]. This implementation assumes that the
causes in the fault tree (leaks) are mutually exclusive; this is because a hydrogen leak in a system
can result in a shutdown of the system itself, thereby precluding other releases2 .
fRandom Releases,k = ∑ NComponenti × fLeak,i,k (12)

i
The component types are Vessels (Cylinders/Tanks), Compressors, Flanges, Hoses, Joints, Pipes3 ,
Valves, Filters, Instruments, Heat Exchangers, Loading Arms, Vaporizers, Extra Component #1,
2 While simultaneous releases are still possible, depending on how quickly a system shutdown can occur, treating the
events as mutually exclusive will lead to a higher (and therefore more conservative) leak frequency compared to
treating leak events as independent.
3 The "Pipes" component type is per-meter of pipe; so if a system has 15 m worth of piping, then the "number" of
components for that type is 15. By contrast, the "Hoses" component type is specified as per-hose, not per-length.
17
and Extra Component #2. Some of these components are only used for gaseous fuels (e.g.,
Compressors), while some are only used for liquid fuels (e.g., Vaporizers), although all are
available as inputs for either gaseous or liquid fuels [2].
2.4.2. Default Component Leak Frequencies
In HyRAM+, the annual frequency of a random leak ( fLeak,i,k ) for component i and leak size k is
assumed to be distributed as a lognormal distribution with parameters µ and σ, as noted in
Equation 13 [6–9, 14].
fLeak,i,k ∼ Lognormal(µ, σ2 ) (13)
The geometric mean (which is equal to the median for lognormal distributions) values are used in
the frequency calculations in HyRAM+ as a metric of central-tendency for the distribution4 [4].
The geometric mean is calculated using Equation 14.
median = eµ (14)
The default values for compressed hydrogen are based on generic system leak frequencies and
data from compressed hydrogen systems developed by LaChance et al. [16] and updated by both
Groth et al. [14] and Glover et al. [17]. For liquid hydrogen, leak frequencies were determined
using gaseous hydrogen and liquefied natural gas data as outlined by Brooks et al. [18]. For
compressed methane, values from the analysis by Brooks et al. are used [19]. For liquid methane,
the analysis described by Mulcahy et al. [20] was used. For propane, the generic system (without
the update using hydrogen data) leak frequencies described by LaChance et al. [16] are used. The
µ and σ parameters were obtained from fitted distributions based on the reported distribution
results from each of the original sources [2]. The default leak frequency distributions for each
component are shown in Figures 2-3–2-5, and the default parameters and median values for the
distributions are listed in Table 2-2.
There are two extra components: Extra Component #1 and Extra Component #2 which might not
fall into the other component type categories [4]. The intention is that users can specify a custom
leak frequency distribution for these components while still keeping the leak frequency
distributions for the other components. These are meant to be set by the user.
For each component for which there is no data for a specific fuel, the mean and median leak
frequencies are set to infinity, with µ and σ set to 999 to achieve this effect [2]. If a user has data
for these components for a specific fuel, they can be updated. If a user does not have specific
information for these components for the specific fuel of interest, leak frequency distribution
parameters for another fuel or another component could be used as a proxy. By making the
default median frequency infinity, the risk metric will result in a value of infinity if one of these
components is included in the system without leak frequency data, thereby alerting the user.
4 Currently, only the geometric mean (median) leak frequency is used in risk calculations; future versions of HyRAM+
may use additional information from the lognormal distribution in uncertainty propagation.
18
Liquid Hydrogen Gaseous Hydrogen
Ve Leak Frequency (/yr) Ve Leak Frequency (/yr)
sse ss
10
10
10
10
10
10
10
10
l (T C el
(Ta Co
7
5
3
1
7
5
3
1
an om nk mp
k/C pr /Cy re
yliess linsso
nd or d r
FlaFilteer) FlaFilteer)
Hnoger Hnoger
He I Joisne He I Joisne
at nst VPip t at nst VPip t
Ex ru alve Ex ru alve
cm e cm e
Ve TranVahpaongent Ve TranVahpaongent
sse sfe ri er sse sfe ri er
l (T C r Azer l (T C r Azer
0.01% Leak Area

0.01% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
Hnoger Hnoger
cm e cm e
0.1% Leak Area
0.1% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
19
Hnoger Hnoger
cm e cm e
1% Leak Area
1% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
Hnoger Hnoger
cm e cm e
10% Leak Area
10% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
Hnoger Hnoger
Ex ru al e Ex ru al e
show the 25th and 75th percentiles of the distribution and the whiskers show the 5th and 95th percentiles. Tra Vachamnevne Tra Vachamnevne
ns por get ns por get
fer iz r fer iz r
100% Leak Area
100% Leak Area
Aremr Aremr
the different components and each leak size. The thick central line is the median leak frequency, the boxes
Figure 2-3 Box and whisker plots of gaseous hydrogen (top) and liquid hydrogen (bottom) leak frequencies for
Liquid Methane Gaseous Methane
Ve Leak Frequency (/yr) Ve Leak Frequency (/yr)
sse ss
10
10
10
10
l (T C el
(Ta Co
100
103
106
100
6
3
4
2
an om nk mp
k/C pr /Cy re
yliess linsso
nd or d r
Hnoger Hnoger
cm e cm e
0.01% Leak Area

0.01% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
Hnoger Hnoger
cm e cm e
0.1% Leak Area
0.1% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
20
Hnoger Hnoger
cm e cm e
1% Leak Area
1% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
Hnoger Hnoger
cm e cm e
10% Leak Area
10% Leak Area
an om rm an om rm
k/C pr k/C pr
yliess yliess
nd or nd or
Hnoger Hnoger
Ex ru al e Ex ru al e
show the 25th and 75th percentiles of the distribution and the whiskers show the 5th and 95th percentiles. Tra Vachamnevne Tra Vachamnevne
ns por get ns por get
fer iz r fer iz r
100% Leak Area
100% Leak Area
Aremr Aremr
the different components and each leak size. The thick central line is the median leak frequency, the boxes
Figure 2-4 Box and whisker plots of gaseous methane (top) and liquid methane (bottom) leak frequencies for
Propane
Ve Leak Frequency (/yr)
ss
10
10
10
10
el
(Ta Co
101
7
5
3
nk mp
/Cy re 1
linsso
d r
FlaFilteer)
Hnoger
He I Joisne
at nst VPip t
Ex ru alve
cm e
Ve TranVahpaongent
sse sfe ri er
l (T C r Azer
0.01% Leak Area
an om rm
k/C pr
yliess
nd or
FlaFilteer)
Hnoger
He I Joisne
at nst VPip t
Ex ru alve
cm e
Ve TranVahpaongent
sse sfe ri er
0.1% Leak Area
l (T C r Azer
an om rm
k/C pr
yliess
nd or
21
FlaFilteer)
Hnoger
He I Joisne
at nst VPip t
Ex ru alve
cm e
Ve TranVahpaongent
sse sfe ri er
1% Leak Area
l (T C r Azer
an om rm
k/C pr
yliess
nd or
FlaFilteer)
Hnoger
He I Joisne
at nst VPip t
Ex ru alve
cm e
Ve TranVahpaongent
sse sfe ri er
10% Leak Area
l (T C r Azer
an om rm
k/C pr
yliess
nd or
FlaFilteer)
Hnoger
He I Joisne
at nst VPip t
Ex ru al e
show the 25th and 75th percentiles of the distribution and the whiskers show the 5th and 95th percentiles.
Tra Vachamnevne
ns por get
fer iz r
100% Leak Area
Aremr
for the different components and each leak size. The thick central line is the median leak frequency, the boxes
Figure 2-5 Box and whisker plot of propane (same default values for both gaseous and liquid) leak frequencies
Table 2-2 Default parameters for frequency of random leaks for individual components.
Gaseous Hydrogen Liquid Hydrogen Gaseous Methane Liquid Methane Propane
Component Leak Size
µ σ median µ σ median µ σ median µ σ median µ σ median
0.01% -2.3 0.3 1.0×10−1 999 999 ∞ -1.7 1.0 1.9×10−1 999 999 ∞ 0.8 1.3 2.2×10+0
0.1% -4.1 0.5 1.7×10−2 999 999 ∞ -3.6 0.8 2.8×10−2 999 999 ∞ -2.2 1.0 1.1×10−1
Compressor 1% -5.4 0.8 4.6×10−3 999 999 ∞ -5.5 0.6 4.1×10−3 999 999 ∞ -5.3 1.0 5.2×10−3
10% -8.8 0.7 1.5×10−4 999 999 ∞ -7.4 0.6 6.0×10−4 999 999 ∞ -8.3 0.7 2.6×10−4
100% -11.1 1.2 1.5×10−5 999 999 ∞ -9.3 0.7 8.8×10−5 999 999 ∞ -11.3 1.2 1.2×10−5
0.01% -13.5 0.7 1.4×10−6 -7.3 1.8 6.5×10−4 -3.6 1.2 2.6×10−2 -7.6 1.1 4.8×10−4 -0.4 1.3 6.6×10−1
0.1% -13.6 0.6 1.2×10−6 -8.9 2.6 1.4×10−4 -4.8 0.9 7.8×10−3 -8.9 2.2 1.4×10−4 -3.9 1.0 2.0×10−2
Vessel (Tank/Cylinder) 1% -14.1 0.6 7.9×10−7 -10.5 2.1 2.8×10−5 -6.1 0.7 2.3×10−3 -10.1 1.9 3.9×10−5 -7.4 0.8 6.4×10−4
10% -14.6 0.6 4.5×10−7 -12.1 2.7 5.7×10−6 -7.3 0.6 6.8×10−4 -11.4 2.4 1.1×10−5 -10.9 0.7 1.9×10−5
100% -15.3 0.6 2.3×10−7 -13.7 3.1 1.2×10−6 -8.5 0.9 2.0×10−4 -12.7 3.2 3.1×10−6 -14.3 0.7 6.1×10−7
0.01% -5.2 1.7 5.3×10−3 999 999 ∞ -1.2 0.9 3.0×10−1 999 999 ∞ -5.2 1.7 5.3×10−3
0.1% -5.3 1.3 5.1×10−3 999 999 ∞ -2.2 0.8 1.1×10−1 999 999 ∞ -5.3 1.3 5.1×10−3
Filter 1% -5.3 1.3 4.8×10−3 999 999 ∞ -3.2 0.6 4.0×10−2 999 999 ∞ -5.3 1.3 4.8×10−3
10% -5.4 0.7 4.6×10−3 999 999 ∞ -4.2 0.6 1.5×10−2 999 999 ∞ -5.4 0.7 4.6×10−3
100% -5.4 0.8 4.4×10−3 999 999 ∞ -5.2 0.6 5.5×10−3 999 999 ∞ -5.4 0.8 4.4×10−3
0.01% -3.9 1.5 2.0×10−2 -3.9 1.5 2.0×10−2 -2.4 1.2 8.7×10−2 -10.1 0.7 4.2×10−5 -3.9 1.5 2.0×10−2
0.1% -6.1 1.1 2.2×10−3 -6.1 1.1 2.2×10−3 -4.7 0.9 8.7×10−3 -10.7 1.2 2.3×10−5 -6.1 1.1 2.2×10−3
Flange 1% -8.3 2.1 2.4×10−4 -8.3 2.1 2.4×10−4 -7.0 0.7 8.8×10−4 -11.2 2.4 1.4×10−5 -8.3 2.1 2.4×10−4
10% -10.5 0.7 2.7×10−5 -10.5 0.7 2.7×10−5 -9.3 0.6 8.8×10−5 -11.7 2.8 8.6×10−6 -10.5 0.7 2.7×10−5
100% -12.7 1.7 2.9×10−6 -12.7 1.7 2.9×10−6 -11.6 0.7 9.2×10−6 -12.2 2.9 5.2×10−6 -12.7 1.7 2.9×10−6
0.01% -7.5 0.4 5.8×10−4 -7.5 0.4 5.8×10−4 -10.5 1.2 2.8×10−5 -13.4 0.7 1.5×10−6 2.5 1.3 1.3×10+1
0.1% -8.5 0.6 2.0×10−4 -8.5 0.6 2.0×10−4 -9.3 0.9 9.1×10−5 -11.7 0.6 7.9×10−6 0.3 0.9 1.4×10+0
Hose 1% -8.7 0.6 1.6×10−4 -8.7 0.6 1.6×10−4 -8.2 0.7 2.9×10−4 -10.1 4.2 4.1×10−5 -1.8 0.8 1.6×10−1
10% -8.8 0.6 1.5×10−4 -8.8 0.6 1.5×10−4 -7.0 0.6 9.1×10−4 -8.5 0.9 2.1×10−4 -4.1 0.7 1.7×10−2
100% -9.7 1.0 6.2×10−5 -9.7 1.0 6.2×10−5 -5.8 0.7 2.9×10−3 -6.8 3.6 1.1×10−3 -6.2 1.4 2.0×10−3
0.01% -10.3 0.2 3.5×10−5 -10.3 0.2 3.5×10−5 0.5 1.1 1.6×10+0 10.5 2.2 3.5×10+4 -0.6 1.3 5.3×10−1
0.1% -12.3 0.9 4.7×10−6 -12.3 0.9 4.7×10−6 -1.5 0.8 2.3×10−1 6.2 1.7 4.8×10+2 -2.3 1.0 1.0×10−1
Joint 1% -11.8 0.5 7.9×10−6 -11.8 0.5 7.9×10−6 -3.4 0.6 3.2×10−2 1.9 1.1 6.5×10+0 -4.0 1.0 1.8×10−2
10% -11.8 0.6 7.5×10−6 -11.8 0.6 7.5×10−6 -5.4 0.5 4.6×10−3 -2.4 0.7 8.8×10−2 -5.6 0.6 3.5×10−3
100% -12.0 0.7 6.4×10−6 -12.0 0.7 6.4×10−6 -7.3 0.6 6.6×10−4 -6.7 0.6 1.2×10−3 -7.4 0.7 6.3×10−4
0.01% -11.7 0.7 8.0×10−6 -11.7 0.7 8.0×10−6 -2.5 1.2 8.1×10−2 -12.8 1.3 2.7×10−6 -7.9 1.0 3.6×10−4
0.1% -12.5 0.7 3.7×10−6 -12.5 0.7 3.7×10−6 -4.2 0.9 1.5×10−2 -13.4 1.4 1.4×10−6 -9.7 0.8 6.2×10−5
Pipe 1% -13.9 1.3 9.6×10−7 -13.9 1.3 9.6×10−7 -5.9 0.9 2.7×10−3 -14.1 1.2 7.9×10−7 -11.4 1.5 1.1×10−5
10% -14.6 1.2 4.6×10−7 -14.6 1.2 4.6×10−7 -7.6 0.6 5.0×10−4 -14.7 1.4 4.2×10−7 -13.2 1.3 1.9×10−6
100% -15.7 1.8 1.5×10−7 -15.7 1.8 1.5×10−7 -9.3 0.9 9.1×10−5 -15.3 1.8 2.3×10−7 -14.9 2.1 3.2×10−7
0.01% -5.9 0.2 2.9×10−3 -5.9 0.2 2.9×10−3 -3.0 1.1 5.1×10−2 -9.4 0.7 8.4×10−5 -4.5 1.0 1.2×10−2
0.1% -7.4 0.4 5.9×10−4 -7.4 0.4 5.9×10−4 -3.9 0.8 2.0×10−2 -10.1 1.0 4.2×10−5 -6.3 0.8 1.9×10−3
Valve 1% -9.8 1.1 5.4×10−5 -9.8 1.1 5.4×10−5 -4.9 1.4 7.8×10−3 -10.7 1.2 2.2×10−5 -8.1 1.5 3.1×10−4
10% -10.6 0.6 2.5×10−5 -10.6 0.6 2.5×10−5 -5.8 0.6 3.0×10−3 -11.3 1.9 1.2×10−5 -9.9 0.6 5.3×10−5
100% -12.2 1.4 4.8×10−6 -12.2 1.4 4.8×10−6 -6.8 1.2 1.2×10−3 -11.9 1.9 6.5×10−6 -11.7 1.4 8.5×10−6
0.01% -7.4 0.7 6.2×10−4 999 999 ∞ -7.3 0.7 6.9×10−4 999 999 ∞ 999 999 ∞
0.1% -8.5 0.8 2.0×10−4 999 999 ∞ -8.1 0.6 3.0×10−4 999 999 ∞ 999 999 ∞
Instrument 1% -9.1 0.9 1.1×10−4 999 999 ∞ -8.9 0.6 1.3×10−4 999 999 ∞ 999 999 ∞
10% -9.2 1.1 1.0×10−4 999 999 ∞ -9.8 0.5 5.7×10−5 999 999 ∞ 999 999 ∞
100% -10.2 1.5 3.7×10−5 999 999 ∞ -10.6 0.7 2.5×10−5 999 999 ∞ 999 999 ∞
0.01% 999 999 ∞ 999 999 ∞ 0.6 1.3 1.8×10+0 -6.1 1.0 2.3×10−3 999 999 ∞
0.1% 999 999 ∞ 999 999 ∞ -1.1 1.0 3.5×10−1 -7.0 1.3 8.9×10−4 999 999 ∞
Heat Exchanger 1% 999 999 ∞ 999 999 ∞ -2.7 1.1 6.7×10−2 -8.0 1.4 3.2×10−4 999 999 ∞
10% 999 999 ∞ 999 999 ∞ -4.4 0.6 1.3×10−2 -9.1 2.3 1.2×10−4 999 999 ∞
100% 999 999 ∞ 999 999 ∞ -6.0 0.9 2.4×10−3 -10.1 1.6 4.2×10−5 999 999 ∞
0.01% 999 999 ∞ 999 999 ∞ 999 999 ∞ -4.8 2.5 8.1×10−3 999 999 ∞
0.1% 999 999 ∞ 999 999 ∞ 999 999 ∞ -3.6 1.9 2.6×10−2 999 999 ∞
Vaporizer 1% 999 999 ∞ 999 999 ∞ 999 999 ∞ -2.5 1.2 8.4×10−2 999 999 ∞
10% 999 999 ∞ 999 999 ∞ 999 999 ∞ -1.3 0.7 2.7×10−1 999 999 ∞
100% 999 999 ∞ 999 999 ∞ 999 999 ∞ -0.1 0.7 8.8×10−1 999 999 ∞
0.01% 999 999 ∞ 999 999 ∞ 999 999 ∞ -1.6 3.0 2.0×10−1 999 999 ∞
0.1% 999 999 ∞ 999 999 ∞ 999 999 ∞ -4.4 2.1 1.2×10−2 999 999 ∞
Transfer Arm 1% 999 999 ∞ 999 999 ∞ 999 999 ∞ -7.2 1.9 7.5×10−4 999 999 ∞
10% 999 999 ∞ 999 999 ∞ 999 999 ∞ -10.3 1.0 3.3×10−5 999 999 ∞
100% 999 999 ∞ 999 999 ∞ 999 999 ∞ -12.7 3.4 3.0×10−6 999 999 ∞
22
2.4.3. Frequency of Dispenser Releases
The annual frequency of other releases ( fOther Releases ) deals with failures that can happen at a
dispenser, rather than random leaks from individual components [4, 14]. The probability of an
accident can be high for several reasons: fueling typically involves direct human interaction to
operate the fueling dispenser, temporary connections rather than hard-plumbed lines, and
inadvertently broken connections because the vehicle is not a permanent part of the system. On
the other hand, releases from fueling can only occur when fueling occurs, and so different
systems that involve different numbers of fueling events can be impacted by these releases to
varying degrees. It is assumed that a dispenser failure would result in a large release of fuel, and
so this frequency is only used in the largest (100%) leak size.
The annual frequency of other releases ( fOther Releases ) is calculated using Equation 15, in which
fFueling Demands is the annual frequency of fueling demands (i.e., the number of times a dispenser
is used to refuel a vehicle in a year) and pDispenser Releases is the probability of a release from a
dispenser during fueling [4]. This implementation assumes that the causes in the fault tree are
mutually exclusive; this is because a hydrogen leak in a system can result in a shutdown of the
system itself, thereby precluding other releases5 .
fOther Releases = fFueling Demands × pDispenser Releases (15)
The annual frequency of fueling demands ( fFueling Demands ) is given by Equation 16, where
NVehicles is the number of vehicles at the facility, NFuelings per Day is the average number of times
each vehicle is fueled per day, and NOperating Days per Year is the number of operating days in a
year [4].
fFueling Demands = NVehicles × NFuelings per Day × NOperating Days per Year (16)
Dispenser failures are categorized in HyRAM+ as Accidents (in which the vehicle tank
overpressurizes or a drive-off occurs) or Shutdown Failures (in which the system fails to shut
down after a release from the nozzle) [4]. The probability for these types of releases are estimated
using a fault tree as shown in Figure 2-6.
5 While simultaneous releases are still possible, depending on how quickly a system shutdown can occur, treating the
events as mutually exclusive will lead to a higher (and therefore more conservative) release probability compared
to treating leak events as independent.
23
100% Release
From Accidents
and Shutdown
Failures
OR
Shutdown
Accidents
Failure
OR AND
Rupture Manual Solenoid

Release Due Nozzle
During ValveCom
Fails to Valves Fail to
to Drive-Off Release
Fueling Close Close
AND AND OR OR
Overpressure Pressure Breakaway Solenoid Common

Drive-Off Coupling Nozzle Nozzle Fails
Com
During Com
Relief Fails to Com Com Com Com Com
Valve Fail to Cause Failure
Com
Fails to Close Ejection to Close
Fueling Open Close (x3) to Close
Figure 2-6 Fault tree for Other Releases from a dispenser [4].
The probability of a release from a dispenser during fueling (pDispenser Releases ) is given by
Equation 17, where pAccidents is the probability of an accident during fueling and pShutdown Failure
is the probability of a shutdown failure during fueling [4].
pDispenser Releases = pAccidents + pShutdown Failure (17)
The probability of an accident during fueling (pAccidents ) is given by Equation 18, where
pRupture During Fueling is the probability of a rupture that occurs during fueling and
pRelease Due to Drive−Off is the probability of a release occurring due to a vehicle drive-off [4].
pAccidents = pRupture During Fueling + pRelease Due to Drive−Off (18)
24
The probability of a rupture during fueling is given by Equation 19, in which
pOverpressure During Fueling is the probability of an overpressure occurring during fueling (e.g., the
dispenser over-fills the vehicle tank) and pPRD FTO is the probability of the dispenser pressure
relief device failing to open on demand [4].
pRupture During Fueling = pOverpressure During Fueling × pPRD FTO (19)
The probability of a release occurring due to a vehicle drive-off (pRelease Due to DriveOff ) is given by
Equation 20, where pDriveOff is the probability of a vehicle driving off while still attached to the
dispenser during fueling and pBreakaway FTC is the probability of the breakaway coupling failing to
close on demand [4].
pRelease Due to DriveOff = pDriveOff × pBreakaway FTC (20)
The probability of a shutdown failure during fueling (pShutdown Failure ) is given by Equation 21,
where pNozzle Release is the probability of the dispensing nozzle releasing fuel, pManual Valve FTC is
the probability of the manual shutoff valve failing to close on demand, and pSolenoid Valves FTC is
the probability of the automated solenoid valves on the dispenser failing to close on
demand [4].
pShutdown Failure = pNozzle Release × pManual Valve FTC × pSolenoid Valves FTC (21)
The probability of the dispensing nozzle releasing fuel (pNozzle Release ) is given by Equation 22, in
which pNozzle Ejection is the probability of the dispenser nozzle being ejected during fueling, and
pNozzle FTC is the probability of the dispenser nozzle failing to close on demand [4].
pNozzle Release = pNozzle Ejection + pNozzle FTC (22)
The probability of the automated solenoid valves on the dispenser failing to close on demand
(pSolenoid Valves FTC ) is given by Equation 23, where pSolenoid Valve FTC is the probability of any one
automated solenoid valve failing to close on demand and pCommon Cause FTC is the probability of
something causing all of the solenoid valves to fail to close on demand (e.g., loss of connection to
sensors) [4].
pSolenoid Valves FTC = [pSolenoid Valve FTC ]3 + pCommon Cause FTC (23)
It should be noted that this fault tree implementation assumes that there are 3 solenoid valves and
that all of them need to fail in order for fuel to be released; thus, the probability for any single
valve failing is cubed.
The probabilities pOverpressure During Fueling , pPRD FTO , pDriveOff , pBreakaway FTC , pManual Valve FTC ,
pNozzle Ejection , pNozzle FTC , pSolenoid Valve FTC , and pCommon Cause FTC can each be specified as a
specific expected value from 0.0 to 1.0, or can be specified as a probability distribution such as
25
beta or lognormal distributions [4]. If a probability distribution is specified, the mean (or median
for a lognormal distribution) value will be calculated and used in the above calculations. Default
values for these probabilities are given in Section 2.4.4.
Any of the probabilities in this section can be used to estimate an annual frequency of any of the
events in question. This can be done by multiplying the probability for event of interest A (pA ) by
the annual number of fueling demands ( fFueling Demands ), as shown in Equation 24 [4].
fA = pA × fFueling Demands (24)
2.4.4. Default Dispenser Failure Probabilities
The default failure probabilities in HyRAM+ were estimated from generic data from the offshore
oil, process chemical, and nuclear power industries as part of a risk assessment for indoor
refueling of hydrogen-powered forklifts [14]. Table 2-3 shows the default probability
distributions and parameters for various types of component failures and Table 2-4 shows the
default accident occurrence probability distributions and parameters for different types of
accidents that are described in Section 2.4.3.
Table 2-3 Default probability distributions for component failure.
Component Failure Mode Distribution Type Parameters

Nozzle Pop-off Beta (α, β) α = 0.5, β = 610415.5
Nozzle Failure to close Expected value 0.002
Breakaway coupling Failure to close Beta (α, β) α = 0.5, β = 5031
Pressure relief valve Premature open Beta (α, β) α = 4.5, β = 310288.5
Pressure relief valve Failure to open Lognormal (µ, σ) µ = −11.74, σ = 0.67
Failure to close
Manual valve Expected value 0.001
(human error)
Solenoid valve Failure to close Expected value 0.002
Common cause failure
Solenoid valves (3 valves, beta factor Expected value 1.28 × 10−4
method)
Table 2-4 Default probability distributions for accident occurrence.
Accident Distribution Type Parameters

Drive-off Beta (α, β) α = 31.5, β = 610384.5
Overpressure during fueling Beta (α, β) α = 3.5, β = 310289.5
26
See Equation 14 for the calculation of the geometric mean (median) for a lognormal distribution.
For a beta distribution with parameters α and β, the arithmetic mean is calculated using
Equation 256 .
α
mean = (25)
α+β
2.5. Consequence Models
The consequences of a leak scenario (cScenario,k for k = 0.01%, 0.1%, 1%, 10%, and 100%) in
HyRAM+ are the estimated numbers of fatalities from each of the j occupants of the facility (see
Section 2.5.1), as calculated in Equation 26.
cScenario,k = ∑ cScenario,k, j (26)

j
The consequences for each of the leak scenarios are estimated by the probability of a fatality for
each of the occupants in the facility, as described in the following sub-sections.
2.5.1. Facility Occupants
The harm and fatalities of interest in HyRAM+ are assumed to happen to facility
occupants [4, 14]. General risk contours or risk at a specific location (such as a building or lot
line) are not calculated explicitly. The risk for the entire facility is a summation of fatality risk for
each of the user-specified occupant locations.
The occupant positions and number of occupants are defined by user input. For each dimension
(x, y, z) for each occupant, the user may assign a position deterministically or may specify a
probability distribution. Probability distributions will be randomly sampled to assign the positions
using user-specified inputs to a uniform or normal distribution. The occupant positions are all
defined relative to the leak point; i.e., the leak occurs at the "origin" (0, 0, 0) and extends in the
positive-x direction, so the occupant positions (x, y, z) are based on that point of reference. The x
and z coordinates are horizontal to the ground, while the y coordinate is height above the ground.
The occupant locations are only sampled (and locations assigned) once per QRA calculation; this
means that if any occupant location dimensions are specified with a probability distribution, the
resulting risk value may differ between calculation runs. Instead, if occupant locations are
specified deterministically for all dimensions, then the same risk value result will occur for each
calculation run.
By default, HyRAM+ includes a set of 9 occupants that are meant to provide an example of
workers within a station or facility. The locations of these default occupants are assumed to be
6 Currently,
only the mean value from the beta distribution is used in risk calculations; future versions of HyRAM+
may use additional information from the probability distribution in uncertainty propagation.
27
distributed with a uniform distribution in the x-direction between 1–20 m, a constant height of
0 m (i.e., same height as the leak itself), and distributed with a uniform distribution in the
z-direction between 1–12 m. These default occupants are assumed to have 2,000 exposed hours
per occupant per year. The number, location specifications, and exposed hours for these default
occupants can be edited by the user, and additional groups with occupants with separate number,
locations, and exposed hours can be specified by the user as well.
2.5.2. Detection and Isolation Scenario Consequences
The scenario in which a leak is safely detected and isolated is assumed to result in no fatalities for
all of the occupant positions [14], as shown in Equation 27.
cIsolated,k, j = 0 (27)
2.5.3. No Ignition Scenario Consequences
The scenario in which a leak does not ignite is assumed to result in no fatalities for all of the
occupant positions [14], as shown in Equation 28.
cNoIgnition,k, j = 0 (28)
2.5.4. Jet Fire Scenario Consequences
The consequences of a jet fire on facility occupants are calculated using Equation 29, in which
pfatal,jetfire,k, j is the probability of a fatality from a jetfire for the leak size k and occupant position
j.
cJetfire,k, j = pfatal,jetfire,k, j (29)
The probability of a fatality from a jetfire is described in Section 2.6, which uses physical effect
modeling as described in Section 3. Specifically, the flame model described in Section 3.4.2 and
multi-point radiative heat flux model described in Section 3.4.3 are used for this scenario. These
models are coupled to the developing flow models described in Section 3.2. The heat flux
calculation is performed at each of the occupant locations, and the resulting values are then used
to estimate the probability of a fatality at each of the occupant locations using the fatality probits
in Section 2.6.1.
28
2.5.5. Overpressure Scenario Consequences
The consequences of an explosion on facility occupants are calculated using Equation 30, in
which pfatal,explosion,k, j is the probability of a fatality from an explosion for the leak size k and
occupant position j.
cExplosion,k, j = pfatal,explosion,k, j (30)
The probability of a fatality from an explosion is described in Section 2.6, which uses physical
effect modeling as described in Section 3. The unconfined overpressure model described in
Section 3.4.5 is used for this scenario. The overpressure calculation is performed at each of the
occupant locations, and the resulting values are then used to estimate the probability of a fatality
at each of the occupant locations using the fatality probits in Section 2.6.2.
2.6. Harm and Loss Models
Probit models are used in HyRAM+ to estimate the probability of a fatality for a given
exposure [6–9, 14]. The probit model is a linear combination of predictors that model the inverse
cumulative distribution function associated with the normal distribution [6, 14]. The probability
of a fatality is given by Equation 31, which evaluates the standard normal cumulative distribution
function (Φ)7 at the value established by the appropriate probit model (Y , see Sections 2.6.1 and
2.6.2) [6, 14].
pfatal = F(Y |µ = 5, σ = 1) = Φ(Y − 5) (31)
2.6.1. Thermal Harm
For thermal radiation, the harm level is a function of both the heat flux intensity and the duration
of exposure. Harm from radiant heat fluxes is often expressed in terms of a thermal dose unit (V )
which combines the heat flux intensity (I, in W/m2 ) and exposure time (t, in seconds) using
Equation 32 [6, 14]. The default thermal exposure time used in HyRAM+ is 30 s8 , but users may
modify this value. This default value was selected based on multiple literature sources that
include the ability of a person to move away from a heat source [21, 22].
V = I (4/3) × t (32)
Table 2-5 lists the thermal probit models that are encoded in HyRAM+ [6, 14]. The probability of
a fatality is evaluated using the probit value resulting from the equations in Table 2-5 with
Equation 31.
7 The standard normal cumulative distribution function (Φ(x)) is the case in which the normal cumulative distribution
function (F(x)) has parameters of µ = 0 and σ = 1.
8 The default value in HyRAM+ version 4.1 and earlier was 60 s [6, 14]
29
Table 2-5 Probit models used to calculate fatality probability as a function of thermal dose (V ).
Reference Fatality Model Notes

Eisenberg [23] Y = −38.48 + 2.56 × ln(V ) Based on population data from nuclear
blasts at Hiroshima and Nagasaki (ultra-
violet radiation)
Tsao & Perry [24] Y = −36.38 + 2.56 × ln(V ) Eisenberg model, modified to account
for infrared radiation
TNO [21] Y = −37.23 + 2.56 × ln(V ) Tsao and Perry model modified to ac-
count for clothing
Lees [25] Y = −29.02 + 1.99 × ln(0.5V ) Accounts for clothing, based on porcine
skin experiments using ultraviolet
source to determine skin damages.
Uses burn mortality information.
LaChance et al. [26] recommended using either the Eisenberg and the Tsao & Perry probit
models for hydrogen-related applications. Therefore, the HyRAM+ default is the Eisenberg
probit model for heat flux. This recommendation is based on the fact that hydrogen flames are
less radiative than hydrocarbon flames, and so the Tsao & Perry probit model may overpredict
due to the inclusion of infrared radiation. By contrast, the Tsao & Perry probit may be more
relevant for hydrocarbon fuels like methane and propane. The TNO and Lees probit accounts for
clothing, which may be appropriate for some situations, though may be less conservative than the
probits that do not account for clothing.
Structures and equipment can also be damaged by exposure to radiant heat flux. Some typical
heat flux values and exposure times for damage to structures and components were provided by
LaChance et al. [26]. However, because the exposure times required for damage is long
(>30 min), the fatality risk due to thermal radiation from fires on structures and equipment is not
generally significant since personnel are able to evacuate the building before significant structural
damage occurs [6, 14]. This is only noted because structural damage can cause physical harm
(fatalities) to humans, which is the risk metric of interest within HyRAM+.
2.6.2. Overpressure Harm
There are several probit models available in HyRAM+ to predict harm and loss from
overpressures [6, 14]. These models generally differentiate between direct and indirect effects of
pressure. Significant increases in pressure can cause direct damage to pressure-sensitive organs
such as the lungs and ears. Indirect effects include the impact from fragments and debris
generated by the overpressure event and collapse of structures. Large explosions can also carry a
person some distance resulting in injury from collisions with structures or from the resulting
violent movement.
The probit models for the effects of overpressures that are included in HyRAM+ are provided in
Table 2-6 [6]. The probability of a fatality is evaluated using the probit value resulting from the
30
equations in Table 2-6 with Equation 31. In this context, the overpressure is defined as the
pressure above ambient.
Table 2-6 Probit models to calculate fatality probability from exposure to overpressures, where Ps is peak
overpressure (Pa) and i is the impulse of the shock wave (Pa·s).
Reference Fatality Model

Eisenberg - Lung hemorrhage [27] Y = −77.1 + 6.91 ln(Ps )
HSE - Lung hemorrhage [28, 29] Y = 5.13 + 1.37 ln(Ps × 10−5 )
TNO - Head impact [21] Y = 5 − 8.49 ln[(2430/Ps ) + 4.0 × 108 /(Ps i)]
TNO - Structure collapse [21] Y = 5 − 0.22 ln[(40000/Ps )7.4 + (460/i)11.3 ]
LaChance et al. [26] recommended the use of the TNO probit models, and suggested that indirect
effects from overpressure events represent the most important concern for people. The
overpressures required to cause fatal lung damage are significantly higher than the values required
to throw a person against obstacles or to generate missiles that can penetrate the skin. In addition,
a person inside a structure would more likely be killed by the facility collapse than from lung
damage [6, 14]. However, a person located outdoors would not be at risk of a structure collapse;
for this reason, the HyRAM+ default is the TNO - Head Impact probit model for
explosion/overpressure effects. It should be noted that some of the unconfined overpressure
models in Section 3.4.5 only estimate peak overpressure, not impulse values, and so cannot be
used with probit models that include an impulse term (i.e., the TNO models).
31
32
3. PHYSICS MODELS
HyRAM+ includes a physics mode, which provides models relevant to the behavior, hazards, and
consequences of releases of the different fuels. Jet flames, concentration profiles for unignited
jets/plumes, overpressure from the delayed ignition of a plume, and indoor accumulation with
delayed ignition causing overpressure can all be investigated from the physics mode. A subset of
these models is used in the QRA mode to calculate the consequences from a given release
scenario, as described in Section 2.5. Several basic property calculations (e.g., the
thermodynamic equation of state) are necessary to numerically simulate the release scenarios.
3.1. Properties of the Fluids
The formulations in this section describe the thermodynamic properties of unignited and ignited
hydrogen, methane, propane, and blends, which are needed to calculate different aspects of
dispersion and combustion. They are described here in detail, and then referred to in subsequent
sections.
3.1.1. Equation of State
Description: HyRAM+ utilizes the CoolProp library [30], called through its Python interface to
perform several thermodynamic calculations. The property calculations are based on a Helmholtz
energy function, and account for the real gas behavior at high pressures and at liquid (which can
be cryogenic) temperatures, for liquids, gases, and two-phase mixtures. CoolProp [30] can be
used to calculate the properties of hydrogen, methane, propane, air, or other fluids. For hydrogen,
the relationships and energy functions are detailed in Leachman et al. [31], for methane, in
Setzmann and Wagner [32], and for propane, in Lemmon et al. [33]. CoolProp handles blends
following the work of Kunz et al. [34, 35] and Lemmon et al. [36–38]. The mixing parameters for
the fluids available in the front-end of HyRAM+ are from Kunz et al. [34, 35]. These
thermodynamic calculations are used to calculate leak rates and are used in mass, momentum, and
energy balances in regions close to the leak point. As an example, for hydrogen, the relationships
between pressure9 , temperature, density, enthalpy, and entropy are plotted in Figure 3-1. In some
regions of the models, the ideal gas equation of state is used, as described in other sections.
9 HyRAM+ calculation inputs use absolute pressure, not gauge pressure
33
20.0 900
8 50 100
10
18 25
60
65
14
45
40 3555 45 50
70 800
95
17.5
12
16
20 30
70
6
700 80
100
15.0 60
70
90
75
60
12.5 50 600 40 35
5 4 500 60
P [bar]
P [bar]
[kg/m3]
[kg/m3]
10.0 40 30
85
400
65
7.5 30 25 40
300
80
5.0 2 20 200 20 15
20
2.5 1 10 10
100
75
3 5
55
0.0 1
20 30 40 50 60 70 50 100 150 200 250 300
T [K] T [K]
300 g
critical point J/k
P = 13.0 bar, T = 33.1 K 0k
400 102
250
3 00 bbar
3000 kJ/kg
20000 barar
5700
100 bar
101
4600bbar r
200
ba
20 b ar
bar
ar bar 5 bar 10 ba r
r
a
900
[kg/m3]
T [K]
2000 kJ/kg
150 100
1500 kJ/kg
100
1b 2 10 1
1000 kJ/kg
50 700 kJ/kg
J/kg
300 k 400 kJ/kg 500 kJ/kg 10 2
0 10 20 30 40 50
S [kJ/kg-K]
Figure 3-1 Graphical representations of state points, calculated using CoolProp [30] which uses the Leachman
et al. [31] equation of state for hydrogen. Top plots show shading and iso-contours of density as a function of
temperature and pressure. Bottom plot shows shading of density as a function of entropy and temperature,
with iso-contours of pressure and enthalpy. The thick black line shows the liquid/two-phase/vapor boundary,
and the black dots mark the triple point and critical points.
Applicability: The fundamental equation of state described by Leachman et al. [31] is valid for
hydrogen at pressures up to 2000 MPa and between 14 K and 1000 K. The equation of state for
methane described by Setzmann and Wagner [32] is valid from 90 K to 620 K at pressures up to
1000 MPa. The relationships described by Lemmon et al. [33] are valid for propane from 85.5 K
to 650 K and for pressures up to 1000 MPa. The most accurate range of validity for
mixtures/blends is for temperatures from 90 K to 450 K and pressures up to 35 MPa, although
limited data extends validity from 60 K to 700 K and up to 70 MPa [35]. Note that there is no
check that the equation of state is being used within the stated validation limits and HyRAM+ can
calculate outputs for temperatures and pressures outside the range of validity using the same
equation of state.
3.1.2. Combustion
Description: HyRAM+ flame calculations are based on the work of Ekoto et al. [39] and rely
on several underlying properties of burned fuel, namely the stoichiometric mixture fraction, fs ,
34
the heat of combustion, ∆Hc , along with the temperature, molecular weight and density of
combustion products for a given mixture fraction.
Assumptions: Combustion is only assumed to occur in expanded fuel at atmospheric pressure.

Because combustion occurs at ambient pressure, the ideal gas equation is used to calculate the
density of the product mixture (ρ) based on the molecular weight of the mixture (MWmixture ), the
temperature (T ), and the gas constant (R):
P(MWmixture )
ρ= . (33)
RT
These combustion calculations assume that there are no losses, that the mixture is thermally
perfect with the local enthalpy, and the pressure of the products is the same as the pressure of the
reactants.
Relationships: It is assumed that there are fuels and inerts reacting with pure air, and complete
combustion drives the products to water and carbon dioxide. For each mole of carbon as a
reactant, 1 mole O2 is needed as a reactant and 1 mole of CO2 will be produced. For each mole of
hydrogen as a reactant, 1/4 mole of O2 is needed as a reactant to produce 1/2 mole of H2 O. For
each mole of oxygen in the reactants (for example, in CO2 ), 1/2 mole less of O2 is needed as a
reactant. Therefore, the moles of oxygen needed as a reactant is
nH nO
νO2 = nC + − (34)
4 2
where n is the moles of each species (subscripts C - carbon, H - hydrogen, O - oxygen) in the fuel.
For example, hydrogen (H2 ) has 2 moles of hydrogen and requires 2/4 = 1/2 mole of O2 ,
methane (CH4 ) has 4 moles of hydrogen and 1 mole of carbon and requires 1 + 4/4 = 2 moles of
O2 , and propane (C3 H8 ) has 3 moles of carbon and 8 moles of hydrogen and therefore requires
3 + 8/4 = 5 moles of O2 for complete combustion. For blends, the moles of carbon, hydrogen,
and oxygen are often non-integers.
During combustion, 242 kJ is released for every mole of gaseous water produced, and 394 kJ is
released for every mole of carbon dioxide produced [40]. Therefore, when hydrogen is the fuel,
the heat of combustion is 242 kJ/molH2 or 120 MJ/kgH2 (using the lower heating value) [41, 42].
For methane and propane, the heats of combustion are 50 MJ/kg and 46.4 MJ/kg, respectively.
The values can be calculated for other fuels in a similar manner. The heat of combustion from a
blend is weighted based on the mass fraction of each reactant in the fuel.
From the moles of oxygen required for complete combustion (Eq. 34), the stoichiometric mixture
fraction ( fs ), which is the same as the mass fraction of fuel, can be calculated as:
MWfuel
fs = . (35)
MWfuel + νO2 (MWO2 + 3.76MWN2 )
35
The stoichiometric mixture fractions for hydrogen, methane, and propane are 0.02852, 0.05519,
and 0.06034 respectively. If incomplete combustion occurs (a mixture fraction other than
stoichiometric), there will be excess air or fuel as a reactant, with the stoichiometry given by
(fuel) + ηνO2 (O2 + 3.76N2 ) → max(0, 1 − η)(fuel)
nH
+ · min(1, η)H2 O + nC · min(1, ηn)CO2
2
+ max (0, (η − 1)νO2 ) O2 + 3.76ηνO2 N2 , (36)
where η, which specifies the moles of air, can vary from 0 to ∞. In this case, the mixture fraction
is equal to
MWfuel MWfuel
f = Yfuel +YH2 O +YCO2 , (37)
MWH2 O MWCO2
where Y is the mass fraction of products. HyRAM+ uses CoolProp [30] to calculate the mass
fraction weighted enthalpy (h) of the fuel along with the enthalpy of H2 O, CO2 , O2 , and N2 as a
function of temperature and then solves for the temperature of products assuming an isenthalpic
reaction, i.e.,
∑ Yi,reac hi,reac (Treac , Preac ) = ∑ Yi,prod hi,prod (Tprod , Pprod )

i=fuel,O2 ,N2 i=fuel,O2 ,N2 ,CO2 ,H2 O
MWfuel
+YH2 O,prod ∆Hc . (38)
(n + 1)MWH2 O
The calculation of the adiabatic flame temperature (Tad ) and density of the products of 298 K,
101,325 Pa pure fuels are shown in Figure 3-2.
2000 H2, Tfs = 2214 K

Tproducts [K]
CH4, Tfs = 2139 K

C3H8, Tfs = 2238 K
1000
fs = 0.02852
fs = 0.05519
fs = 0.06034
[kg/m3]
1.5
1.0
product mixture
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
mixture fraction (fs)
Figure 3-2 Temperature and density of products for the combustion of 298 K, 101,325 Pa fuels as a function of
mixture fraction.
Applicability: These combustion calculations are applicable in atmospheric pressure regions

where heat losses are negligible.
36
3.2. Developing Flow
Several engineering models are used in HyRAM+ to develop boundary conditions for the integral
models of jets/plumes and diffusion flames. These engineering models describe the flow through
an orifice, how the fluid expands to atmospheric pressure (if necessary), how the fluid warms to a
level that the equation of state is valid (if necessary), and finally how the flow develops into the
Gaussian profiles that serve as the boundary conditions to the models described in Sections 3.3.1
and 3.4.2.
3.2.1. Orifice Flow
HyRAM+ assumes that fluids flow isentropically through an orifice. CoolProp [30] is used to
calculate the entropy (s0 ) and enthalpy (h0 ) of the fluid upstream of an orifice, using the specified
pressure and temperature (or phase if a saturated vapor or saturated liquid is specified).
CoolProp [30] is then used to calculate the enthalpy (h) and density (ρ) of a fluid at a given
pressure with the same entropy as the upstream fluid (s0 ). An isenthalpic expansion requires
v2
+ h = h0 (39)
2
which can be solved for the velocity, v, and the mass flux can be calculated as
m˙′′ = ρv. (40)
A maximum mass flux is sought between the ambient and upstream pressures [43–45] using a
bounded solver. If the maximum mass flux occurs at atmospheric pressure, the flow is unchoked,
while if the maximum mass flux is at a pressure above atmospheric, the flow is choked. In the
case of choked flow, the velocity through the orifice will be the speed of sound for the given throat
conditions. The choked flow speed of sound for gases is the same as that calculated by
CoolProp [30], but this algorithm also works for two-phase and liquid flows through the throat,
for which the speed of sound is ill-defined.
Orifices in HyRAM+ are assumed to be circular, characterized by their diameter, d, and a
coefficient of discharge, Cd . When the velocity and density of the fluid at the orifice is known, the
mass flow rate is calculated as:
π π
ṁ = d 2 m˙′′Cd = d 2 ρvCd . (41)
4 4
3.2.2. Notional Nozzles
Notional nozzles are used to calculate the effective diameter, velocity, and thermodynamic state
after the complex shock structure of an under-expanded jet. In HyRAM+, a notional nozzle
model is used if the pressure at the orifice is above atmospheric pressure. They are not necessarily
a physical description of the phenomena, but a jet with the diameter, velocity and state
(temperature and atmospheric pressure) of the notional nozzle would lead to the same dispersion
37
characteristics as the underexpanded jet. There are five different notional nozzle models in
HyRAM+, with each model conserving mass between flow through the real orifice and flow
through the notional nozzle. This means that
ρeff veff Aeff = ρthroat vthroat AthroatCD (42)
where ρ is the density, v is the velocity, A is the cross-sectional area, CD is the discharge
coefficient, the subscript "throat" denotes the choke point (at the orifice, see Section 3.2.1), and
the subscript "eff" denotes effective (after the shock structure and the pressure has returned to
atmospheric).
The default notional nozzle model in HyRAM+ is based on the work of Yüceil and Ötügen [46].
In this case, mass (Equation 42), momentum, and energy are conserved. Conservation of
momentum is written as
ρeff v2eff Aeff = ρthroat v2throat AthroatCD + Athroat (Pthroat − Pambient ) (43)
where P is the pressure. Simultaneous solution of Equations 42 and 43 yields a solution for the
velocity at the notional nozzle
Pthroat − Pambient
veff = vthroatCD + (44)
ρthroat vthroatCD
and the effective area of the notional nozzle
ρthroat v2throat AthroatCD2
Aeff = 2
(45)
CD2

ρeff Pthroat − Pambient + ρthroatVthroat
The effective area calculation in Equation 45 requires the effective density, which can be
calculated using the conservation of energy (assuming isentropic expansion), where
v2eff v2
+ h(ρeff , Pambient ) = throat + hthroat . (46)
2 2
CoolProp [30] is used to calculate the enthalpy and Equation 46 is iteratively solved to determine
the effective density.
Alternative to using Equation 46, the second notional nozzle model follows the work of Birch et
al. (1987) [47]. In this work, the effective density is calculated by assuming that the temperature
of the notional nozzle is the same as the temperature of the stagnant gas, or
ρeff = ρ(T0 , Pambient ) (47)
where T0 is the temperature of the stagnant gas (storage temperature) and CoolProp [30] is used to
calculate the density.
Three other notional nozzle models do not conserve momentum (Equation 43), but rather assume
that the notional nozzle velocity is at the speed of sound, as follows. The third notional nozzle
model follows the work of Birch et al. (1984) [48]. For this model, it is assumed that the
38
temperature at the notional nozzle is the same as temperature of the stagnant gas, the density (see
Equation 47) and velocity at the notional nozzle can be calculated
veff = a(T0 , Pambient ) (48)
where a is the speed of sound, calculated using CoolProp [30]. The conservation of mass,
Equation 42, along with Equations 47 and 48, can be used to specify the notional nozzle
conditions.
Alternatively, Ewan and Moody [49] use the assumption that the temperature at the notional
nozzle is the same as the temperature at the throat, or
ρeff = ρ(Tthroat , Pambient ) (49)

veff = a(Tthroat , Pambient ). (50)
Finally, Molkov et al. [50] specifies that mass and energy are conserved between the orifice and
the notional nozzle and that the notional nozzle is at the speed of sound, i.e., Equation 42 along
with the simultaneous solution of the equations,
v2eff v2
+ h(ρeff , Pambient ) = throat
2 + hthroat (51)
2
veff = a(ρeff , Pambient ) (52)
where h and a are calculated using CoolProp [30].

To summarize, the 5 different notional nozzles available in HyRAM+ solve the equations:
• (default) Yüceil and Ötügen [46]: Equations 44, 45, and 46
• Birch et al. (1987) [47]: Equations 44, 45, and 47
• Birch et al. (1984) [48]: Equations 42, 47, and 48
• Ewan and Moody [49]: Equations 42, 49, and 50
• Molkov et al. [50]: Equations 42, 51, and 52
3.2.3. Initial Entrainment and Heating
The models in HyRAM+ are valid for hydrogen, methane, propane, and blends, including
saturated vapor and saturated liquid releases. Specifically for cryogenic hydrogen, there are
challenges calculating properties in regions of the flow where oxygen and nitrogen from the
entrained air would condense due to the extremely low temperatures, as noted by Houf and
Winters [51]. To account for this, conservation of mass, energy and momentum can be applied
until the temperature of the mixture (still assumed to be a plug-flow) is above a specified
temperature (Tmin ). However, it should be noted that the default minimum temperature is 0 K, so
initial entrainment and heating is currently unused in the GUI implementation and can only be
used in the Python implementation. If the temperature of the notional nozzle (or at the orifice, if
39
the flow is unchoked) is below Tmin , the state after initial entrainment and heating is specified as
the fuel at Tmin , and simultaneous solution to the momentum and energy balances yields the mass
fraction of fuel (Y ) when the mixture has warmed to Tmin , i.e.,
1 −Y
ṁout = ṁin + ṁin (53)
Y
ṁin
vout = vin (54)
ṁout
v2out
hout = (1 −Y )hair (Tmin , Pambient ) +Y hH2 (Tmin , Pambient ) + (55)
2
ṁout hout = ṁin hin + (ṁout − ṁin )hair (Tambient , Pambient ). (56)
Once the mass fraction (Y ) is known, conservation of mass is used to yield the diameter of the
plug flow at the end of the zone of initial entrainment and heating,
1
ρout = 1−Y Y
(57)
ρair (Tmin ,Pambient ) + ρH(T ,P )
2 min ambient
s
πṁout
dout = (58)
4ρout vout
and the momentum driven entrainment rate (see Equation 80) is used to calculate the length of
this zone,
(1 −Y )(ṁout − ṁin )
S= . (59)
ρambient (Tambient , Pambient )Emom
3.2.4. Establishment of a Gaussian Profile
The flows described in Sections 3.2.1, 3.2.2, and 3.2.3 are all assumed to be plug flows, where the
properties (e.g., velocity, density) are constant across the entire cross-section of the flow.
However, jets, plumes, or flames from a pure source are well-known to have Gaussian profiles of
their properties (e.g., velocity, density, mixture fraction) in the downstream regions [39, 52]. The
final model for developing flow describes the transition from plug flow to the Gaussian profile
that is used as an input to a one-dimensional system of ordinary differential equations that
describes unignited dispersion or a diffusion flame (see Sections 3.3.1 and 3.4.2). Following the
work of Winters [53], the centerline velocity of the Gaussian flow is assumed to be equivalent to
the plug flow velocity, the jet is characterized by a half-width, B, where the velocity drops to half
of the center-line value, and a spreading ratio, λ, the ratio of density spreading relative to velocity.
The center-line (denoted with a cl subscript) mass-fraction is related to λ via the relationship,
λ2 + 1 Ycl −Yambient
2
= . (60)
2λ Yplug −Yambient
Then the center-line molecular weight can be calculated,

1
MWcl = . (61)
Ycl /MWCn H2n+2 + (1 −Ycl )/MWair
40
The heat capacity of the fluid and ambient air are determined from CoolProp [30] and used to
calculate the individual and mixture enthalpies as h = c p · T , where
c p,plug = c p,Cn H2n+2 Yplug + c p,air (1 −Yplug ), (62)

c p,cl = c p,plugYcl + c p,air (1 −Ycl ), (63)
λ2 + 1 hcl − hambient
2
= . (64)
2λ hplug − hambient
From these equations, the center-line temperature can be calculated, and the center-line density is
calculated using the ideal gas equation of state, where
MWcl P
ρcl = . (65)
RTcl
The length of the developing flow region is taken from Abraham [54], where S/d = 6.2, which
assumes that the Froude number (Equation 84) is greater than the square root of 40 for these high
speed, low density jets.
3.3. Unignited Releases
3.3.1. Gas Jet/Plume
For a jet or plume, HyRAM+ follows the one-dimensional model described by Houf and
Winters [51]. While the model only considers one dimension, this dimension is along the
streamline, and the jet/plume can curve due to buoyancy effects (or wind, although this aspect is
not currently included). The reduction in dimension comes from the assumption that the mean
profiles of the velocity (v), density (ρ), and product of density and mass fraction (Y ) of fuel are
Gaussian, as
2
r
v = vcl exp − 2 (66)
B
r2

ρ = (ρcl − ρamb ) exp − 2 2 + ρamb (67)
λ B
r2

ρY = ρclYcl exp − 2 2 (68)
λ B
where B is a characteristic half-width, λ is the ratio of density spreading relative to velocity, the
subscript cl denotes the centerline, the subscript amb denotes ambient, and r is perpendicular to
the stream-wise direction. Gravity acts in the negative y-direction, and the plume angle, θ is
relative to the x-axis (horizontal), as shown in Figure 3-3.
41
y S
θ
φ
r
y0 θ0
x0 x
z
Figure 3-3 Sketch of plume model coordinates. Gravity acts in the negative y-direction.
The derivatives of the spatial dimensions are therefore

dx
= cos θ, (69)
dS
dy
= sin θ. (70)
dS
The conservation equations can be written as follows:
continuity:
d
Z ∞
(ρv)2πrdr = ρamb E, (71)
dS 0
x-momentum:
d
Z ∞
ρv2 cos θ 2πrdr = 0,

(72)
dS 0
y-momentum:
d
Z ∞ Z ∞
2

ρv sin θ rdr = (ρamb − ρ)grdr, (73)
dS 0 0
species continuity:
d
Z ∞
(ρvY ) 2πrdr = 0, (74)
dS 0
42
energy:
v2

d
Z ∞
ρv h + − hamb 2πrdr = 0. (75)
dS 0 2
Similar to Houf and Winters [51], HyRAM+ assumes that h = c p T and the ideal gas equation of
state is used for these ambient pressure mixtures. The mixture molecular weight, heat capacity,
and product of density and enthalpy all vary with respect to the radial coordinate (due to the fact
that Y and ρ vary radially, see Equation 68) according to the following expressions:
MWamb MWCn H2n+2
MW = , (76)
Y (MWamb − MWCn H2n+2 ) + MWCn H2n+2
c p = Y (c p,Cn H2n+2 − c p,amb ) + c p,amb , (77)

P
ρh = . (78)
c p MW
The Gaussian profiles in Equations 66-68 are plugged into the governing equations, and with the
exception of the energy equation can be integrated analytically. For the energy equation,
Equation 75, the numeric integration to infinity is estimated by evaluation to 5B. This results in a
system of 7 first order differential equations where the independent variable is S and the
dependent variables are vcl , B, ρcl , Ycl , θ, x, and y. This system of equations is integrated from the
starting point to the distance desired using an explicit Runge-Kutta method of order (4)5.
The entrainment model follows Houf and Schefer [52], where there is a combination of
momentum and buoyancy driven entrainment,
E = Emom + Ebuoy , (79)
where !1/2
2 ρ v2
πdexp exp exp
Emom = 0.282 , (80)
4 ρ∞
where the "exp" subscript denotes after the notional nozzle and zone of initial entrainment and
heating (if used; see Sections 3.2.2 and 3.2.3), and
a
Ebuoy = (2πvcl B) sin θ, (81)
Frl
where the local Froude number,
v2cl
Frl = . (82)
gD(ρ∞ − ρcl )/ρexit
In these equations, a is empirically determined:
(
a = 17.313 − 0.116665Frden + 2.0771 × 10−4 Frden 2 , Frden < 268
(83)
a = 0.97, Frden ≥ 268,
43
with the densimetric Froude number
vplug
Frden = p , (84)
gdplug |ρ∞ − ρplug |/ρplug
where the subscript “plug” denotes the plug flow at the exit of the orifice, notional nozzle, or zone
of initial entrainment and heating, as applicable (see Sections 3.2.1, 3.2.2, and 3.2.3).
As the jet/plume becomes buoyancy-dominated (as opposed to momentum-dominated), the
non-dimensional number
E
α= (85)
2πBvcl
will increase. When α reaches the limiting value of α = 0.082, α is held constant and the
entrainment value becomes:
E = 2παBvcl = 0.164πBvcl . (86)
3.3.2. Tank Mass
The mass of fuel in a storage container (m) is calculated using the density of the fuel (ρ) and fixed
volume of the container (V ) using Equation 87. The density of the fuel (ρ) is calculated using the
specified temperature and pressure using the CoolProp library [30] as described in
Section 3.1.1.
m = ρV (87)
3.3.3. Tank Emptying
In the case of a storage tank with a given volume, the transient process of the tank emptying
(blowdown) can also be calculated by HyRAM+. In this case, energy and mass are conserved,
following the work of Hosseini et al. [55]. The mass flow rate, ṁ is calculated as described in
Section 3.2.1, whether the flow is choked or not. Energy is conserved in the tank volume, where
du ṁ(h − u) + q
= , (88)
dt m
where u and h are the specific energy and specific enthalpy (J/kg), respectively, of the fuel in the
tank (calculated using CoolProp [30]), m is the mass of fuel in the tank (kg, see Section 3.3.2),
and q is the heat flow into the tank (J/s, q = 0 if adiabatic). This equation and the equations
describing the mass flow rate (which are functions of the pressure and temperature inside the
tank) are integrated until the mass or pressure in the tank reaches the desired stopping point (e.g.,
the tank pressure reaches ambient).
44
3.3.4. Accumulation in Confined Areas/Enclosures
When a release occurs in an enclosure, a stratified mixture of fuel and air can accumulate. For
hydrogen or methane, the accumulated mixture will be near the ceiling due to buoyancy.
The release inside an enclosure is assumed to come from some tank with a fixed volume.
Therefore, the flow from the tank follows Section 3.3.3. At each point in time for the blowdown,
the jet/plume is modeled as described in Section 3.3.1. When these releases occur indoors, the
plumes could impinge on a wall. Currently, should this impingement happen, the trajectory of the
jet/plume is modified such that the fuel will travel vertically upwards along the wall, rather than in
the horizontal direction, with the same features (e.g., half-width, centerline velocity). Note that
this deflection upwards will occur regardless of the fuel and is a poor assumption for heavier fuels
such as propane.
Accumulation occurs following the model of Lowesmith et al. [56], where a layer forms along the
ceiling. Conservation of mass requires that
dVl
= Qin − Qout , (89)
dt
where Vl is the volume of gas in the layer, and Q is the volumetric flow rate, with subscript “in”
referring to the flow rate of fuel and air entrained into the jet at the height of the layer, and “out”
referring to flow out the ventilation. Species conservation requires that
d(χVl )
= Qleak − χQout , (90)
dt
where χ is the mole or volume fraction of fuel in the layer and Qleak is the leak rate of the fuel.
Expanding the derivative and substituting Equation 89 yields
dχ
Vl = Qleak − χQin . (91)
dt
Qin is solved for by modeling a jet/plume within the enclosure to calculate the jet half-width (B)
and centerline velocity (vcl ), as described in Section 3.3.1 at the height of the bottom of the layer.
The volumetric flow rate (Qin ) is calculated as:
Qin = πB2 vcl . (92)
Flows out of the enclosure are driven by buoyancy, and potentially wind or a fan. Buoyancy
driven flow is calculated as p
Qb = Cd Av g′ Hl , (93)
where Cd is a coefficient of discharge, Av is the area of the upper (outlet) vent, Hl is the height of
the layer (between the bottom of the layer and the center-point of the outlet vent), and g′ is
reduced gravity:
ρair − ρl
g′ = g , (94)
ρair
45
where g is the gravitational constant, ρl is the density of the layer, and ρair is the density of air (at
the temperature and pressure of the enclosure). The density in the layer (ρl ) is calculated from the
density of air (ρair ) and the density of the fuel (ρCn H2n+2 ), both at the temperature and pressure of
the enclosure, as:
ρl = χρCn H2n+2 + (1 − χ)ρair . (95)
Wind (or mechanical ventilation) is assumed to drive the flow at a rate of:
Cd AvUw Cd Qvent
Qw = √ = √ , (96)
2 2
where Cd is a coefficient of discharge, Av is the area of the lower (inlet) vent, Uw is the velocity of
the wind (or mechanical ventilation) through the lower (inlet) vent, and Qvent is the volumetric
flow rate through the lower (inlet) vent (Qvent = AvUw ). Note that Qvent is a user input.
The total flow out of the enclosure, accounting for buoyancy driven flow and flow from the vent,
is calculated as q
Qout = Qleak + Q2b + Q2w . (97)
3.4. Ignited Releases
3.4.1. Flame Correlations
As noted by Houf and Schefer [57], a non-dimensional flame length, defined as

L f
L∗ = p vis s (98)
d j ρ j /ρamb
collapses onto a single curve for a range of fuels (hydrogen, methane, and propane), where Lvis is
the visible flame length (from the orifice, including any liftoff distance), fs is the mass fraction of
fuel in a stoichiometric mixture of fuel and air (Equation 35), and d j , ρ j , and ρamb are the orifice
diameter, density of fuel at the orifice, and density of air, respectively. It is assumed that this
curve will also be valid for blends. The curve is given by
13.5Fr2/5
(
∗ 2 1/5 , Fr < 5,
L = (1+0.07Fr ) (99)
23, Fr > 5,
which is a function of the Froude number (Fr), which is the ratio of buoyancy to momentum
forces. The Froude number is defined as
3/2
u j fs
Fr = q p , (100)
gd j (Tad − Tamb )/Tamb ρ j /ρamb
where g is the gravitational constant, u j is the velocity of the jet at the orifice, Tad is the adiabatic
flame temperature (for a stoichiometric mixture, see section 3.1.2), and Tamb is the ambient
temperature. The flame width is constant at W f = 0.17Lvis [58, 59].
46
The total emitted radiative power from a flame, Srad , is related to the total energy in the flame by
the radiant fraction,
Srad = Xrad ṁfuel ∆Hc , (101)
where Xrad is the radiant fraction, ṁfuel is the mass flow rate of fuel, and ∆Hc is the heat of
combustion (120 MJ/kg for hydrogen, 50 MJ/kg for methane, and 46.4 MJ/kg for
propane [41, 42]). The radiant fraction (Xrad ) varies with the flame residence time (τ f ); the
relationship is [60]
Xrad = 9.45 × 10−9 (τ f a p Tad
4 0.47
) , (102)
where a p is the Planck-mean absorption coefficient for an optically thin flame (0.23 for
hydrogen [61]). The Plank-mean absorption coefficient should be updated for methane, propane,
and blends, but the current value of 0.23 is used for all fuels. The flame residence time can be
calculated as
ρ f W f2 Lvis fs
τf = , (103)
3ρ j d 2j u j
where ρ f is the flame density. The flame density is calculated as the density at the adiabatic flame
temperature:
pambWmix
ρf = , (104)
RTad
where pamb is the ambient pressure, Wmix is the mean molecular weight of the stoichiometric
products of combustion in air, and R is the universal gas constant.
The transmissivity, which can reduce the radiated heat flux is calculated to account for the
absorption from water vapor and CO2 , using a correlation from Wayne [62]:
τ = 1.006 − 0.001171 (log10 XH2 O ) − 0.02368 (log10 XH2 O )2

− 0.03188 (log10 XCO2 ) + 0.001164 (log10 XCO2 )2 , (105)
where XH2 O and XCO2 is proportional to the amount of water vapor or CO2 in the path
(dimensionless). These values are calculated by:
273 cCO2
XCO2 = L · · , (106)
T 335
2.88651 × 102
XH2 O = RH · L · Smm · . (107)
T
In these relationships, L is the path length (m) through which the radiative light must travel, T is
the ambient temperature (K), cCO2 is the concentration of CO2 in the atmosphere (ppm, assumed
to be 400 ppm), RH is the fractional relative humidity (ranges from 0–1), and Smm is the saturated
water vapor pressure (mm Hg), estimated by the relationship:

5132
Smm = exp 10.386 − . (108)
T
47
3.4.2. Jet Flame with Buoyancy Correction
A similar model to the jet/plume model described in Section 3.3.1 is also used to describe a flame.
The model is described by Ekoto et al. [39]. The major difference between the jet/plume model
and the flame model is that rather than the mole fraction, the mixture fraction, shown in
Equation 37, is a conserved scalar. Similar assumptions are made for Gaussian profiles of the
velocity and mixture fraction:
2
r
v = vcl exp − 2 , (109)
B
r2

f = fcl exp − 2 2 , (110)
λ B
with the conservation equations written as
x-centerline:
dx
= cos θ, (111)
dS
y-centerline:
dy
= sin θ, (112)
dS
continuity:
d
Z ∞
2π ρvrdr = ρamb E, (113)
dS 0
x-momentum:
d
Z ∞
2π ρv2 cos θrdr = ρamb Evwind , (114)
dS 0
y-momentum:
d
Z ∞ Z ∞
2
ρv sin θrdr = (ρamb − ρ)grdr, (115)
dS 0 0
mixture fraction:
d
Z ∞
2π ρV f rdr =0. (116)
dS 0
Note that energy conservation is not included in this formulation, but rather the mixture is
assumed to be thermally perfect, with combustion calculations shown in Section 3.1.2. These
calculations assume that the mixture is always in equilibrium (neglecting heat-losses), which
48
results in a calculation of flame temperature and density as functions of mixture fraction similar to
Figure 3-2 (with slight variations depending on the reactant temperature).
Equation 114 does not explicitly set the right-hand side of the x-momentum equation to 0, as is
done in the unignited jet/plume model (Equation 72). While the Windows GUI version of
HyRAM+ sets vwind = 0, the Python backend enables specification of a cross-wind velocity, vwind .
The Gaussian profiles in Equations 109 and 110 are numerically evaluated out to 5B (an estimate
of ∞), along with the radial profiles of the density (based off of the mixture fraction), which can
be plugged into Equations 111–116 and numerically integrated. This results in a system of 6 first
order differential equations where the independent variable is S and the dependent variables are
vcl , B, θ, fcl , x, and y. This system of equations is integrated from the starting point to the distance
desired using an Adams/BDF method with automatic stiffness detection and switching. Typically,
the integration distance is the visible flame length, calculated using the correlations in
Equations 98 and 99.
Similar to the nonreacting jet, entrainment in the jet flame is modeled as the sum of momentum
and buoyancy contributions (Equation 79). However, the contributions have modified empirical
parameters and calculation methods. The momentum driven entrainment is calculated as
!1/2
2 ρ v2
πdexp exp exp
Emom = 0.0342 , (117)
4 ρ∞
and buoyancy driven entrainment is calculated as

R∞
0 (ρamb − ρ)dr
Ebuoy = 2παbuoy g sin θ (118)
Bvcl ρexit
where αbuoy = 5.75 × 10−4 [39].
3.4.3. Radiation From a Curved Flame
The radiative heat flux from the buoyancy corrected, curved flame is calculated by a weighted
multi-source model, similar to that described by Hankinson and Lowesmith [63]. The heat flux at
a point along the flame is calculated as
VF
q = τSrad , (119)
Af
where Srad is calculated according to Equations 101–103, VF is the view-factor, proportional to the
heat flux transmitted to the observer, τ is the transmissivity, calculated by Equation 105, and A f is
the surface area of the flame. Contributions to the total heat flux are broken up into many (N,
generally 50) points along the length of the curved flame, and the weighted average proceeds as
N
VF wi cos βi
τ =∑ τi , (120)
A f i=1 4πD2i
49
where the emitter strength weighting parameter
(
iw1 i ≤ 0.75N
wi = n−1
(121)
n − N−n−1 (i − (n + 1)) w1 i > 0.75N,
with the constraint that 1 ≤ n ≤ N and ∑N

i=1 wi = 1. In these equations, D and β are the distance
and angle, respectively between the observer and unit normal to the point emitter.
3.4.4. Overpressure in Enclosures
If a confined mixture ignites, significant overpressures can develop within the enclosure or
confinement area. Overpressure is calculated assuming that the cause of overpressure is the
volume change on combustion pressurizing the enclosure. Within HyRAM+, the models
described in Section 3.3.4 are used to calculate the volume of fuel within the layer and entire
enclosure. It is assumed that all of the fuel within the flammability limits (in both the jet/plume
and the accumulated layer) reacts, and the overpressure is calculated, following Bauwens and
Dorofeev [64] as
VT +Vstoich (σ − 1) γ

VT +VCn H2n+2
∆p = p0 −1 , (122)
VT VT
where p0 is the initial pressure, VT is the total volume of the enclosure, VCn H2n+2 is the expanded
volume of pure fuel following the release, Vstoich is the volume of a stoichiometric mixture of the
consumed fuel, σ is the expansion ratio of a stoichiometric fuel-air mixture, and γ is the specific
heat ratio of air. The expanded volume is given by VCn H2n+2 = mCn H2n+2 /ρCn H2n+2 where mCn H2n+2
is the mass of fuel consumed and ρCn H2n+2 is the density of the fuel at ambient conditions. Vstoich
is VCn H2n+2 divided by the stoichiometric mole fraction of fuel.
3.4.5. Unconfined Overpressure
If an unconfined mixture ignites after a release has been flowing for some time, an overpressure
can be observed as the initial mixture burns. In this context, the overpressure is defined as the
pressure above ambient. HyRAM+ has three different methods for calculating the overpressure,
each of which requires the calculations from an unconfined and unignited jet/plume as described
in Section 3.3.1. Each method calculates an overpressure and possibly impulse as a function of
distance, R. The origin for the overpressure/impulse (distance from which R is measured) is the
point at which the fuel concentration is assumed to be halfway between the lower and upper
flammability limits along the jet streamline. This is assumed to be a reasonable origin location,
given the overpressure blast wave will originate from the flammable mixture10 .
10 This is also roughly consistent with the suggested ignition concentration by Jallais et al. [65] for hydrogen.
50
BST The Baker-Strehlow-Tang (BST) model is based on blast curves that relate overpressure
and impulse to the Mach flame speed [66]. The flammable mass of fuel within the unignited
jet/plume (m f lam ) is first found by volumetrically integrating the product of the mass fraction (Y )
and density (ρ) of the jet/plume that is within the flammability limits (YLFL , YUFL ) along the entire
length of the jet/plume streamline coordinate (S), or
!
Z Z S=∞ rY =YLFL
mflam = ρY 2πrdr dS. (123)
S=0 rY =YUFL
The energy within the unignited flammable mixture (Eflam ) is related to the flammable mass
through the relationship
Eflam = kreflection mflam ∆Hc (124)
where kreflection is a ground reflection factor (assumed to be 2) and ∆Hc is the heat of combustion
of the fuel [66]. The scaled distance is related to the energy through the relationship [66]
R
R∗BST = . (125)
(Eflam /Pambient )1/3
The scaled overpressure and impulse are related to the scaled distance, based on the Mach flame
speed, as shown in Figure 3-4 [66]. The Mach flame speed selection is an important choice for
BST model results, and can depend on the fuel being combusted, confinement and congestion of
the flammable mixture, and other factors [66]. The default value in HyRAM+ for the Mach flame
speed is 0.3511 , based on overpressure observations for unconfined releases of hydrogen [65]; this
value can be changed by the user.
* ) [Pa m/(J/Pa)1/3]
Flame Speed
Scaled Peak Overpressure (P * )
100 (Mach)
5.2
101 4.0
10 1 3.0
2.0
Scaled Impulse (IBST
10 1
10 2 1.4
1.0
0.7
10 3 10 3 0.35
10 1 100 101 10 2 10 1 100 101 0.2
Scaled Distance (RBST
* ) [m/(J/Pa)1/3] Scaled Distance (RBST
* ) [m/(J/Pa)1/3]
Figure 3-4 Mapping of scaled distance to scaled overpressure (left) and scaled impulse (right) for the BST
unconfined overpressure model [2].
11 The default Mach flame speed value was 5.2 for HyRAM+ versions 4.0 and 4.1, which is used for detonations;
however, experimental observations of unconfined deflagrations seem more applicable for these models.
51
The scaled overpressure (P∗ ) is the peak overpressure (Ps ) relative to ambient pressure
(Pambient ),
P∗ = Ps /Pambient , (126)
and the impulse (I) is scaled by the flammable energy (Eflam ), ambient pressure (Pambient ), and
speed of sound for air (aair = 340 m/s) as
∗ Iaair
IBST = 2
. (127)
(Eflam Pambient )1/3
TNT The TNT equivalence method is based on finding the mass of TNT that contains the same
energy as the fuel being combusted [66]. The flammable mass of fuel within the jet/plume
(m f lam ) is found by volumetrically integrating the product of the mass fraction and density of the
jet/plume that is within the flammability limits (see Equation 123). The flammable mass is scaled
by an equivalence factor (Fequiv , 3% by default [66]), and the equivalent mass of TNT (mT NTequiv )
is calculated as
mflam ∆Hc
mTNTequiv = Fequiv , (128)
∆Hc,TNT
where and ∆Hc,TNT = 4.68 MJ/kg is the equivalent specific blast energy of TNT12 . The scaled
distance (R∗TNT ) is related to the equivalent mass of TNT (mTNTequiv ) through the relationship
R
R∗TNT = 1/3
, (129)
mTNTequiv
and the scaled overpressure (P∗ ) and scaled impulse (ITNT

∗ ) resulting from combustion are related
to the scaled distance as shown by Figure 3-5.

* ) [Pa s/kg1/3]
Scaled Peak Overpressure (P * )
104 100
102 10 2
Scaled Impulse (ITNT
100 10 4
10 2 10 6
10 2 10 1 100 101 102 10 2 10 1 100 101 102
Scaled Distance (RTNT
* ) [m/kg1/3] Scaled Distance (RTNT
* ) [m/kg1/3]
Figure 3-5 Mapping of scaled distance to scaled overpressure (left) and scaled impulse (right) for the TNT
equivalence unconfined overpressure model [2].
12 Previous versions of HyRAM+ (e.g., [1]) had a separate Engineering Toolkit calculation that used a slightly different
value for the TNT blast energy.
52
The scaled overpressure (P∗ ) is the overpressure (Ps ) relative to ambient pressure (see
Equation 126), and the impulse (I) is scaled by the third root of the TNT mass (mTNTequiv ), or
∗ I
ITNT = 1/3
. (130)
mTNTequiv
Bauwens The Bauwens method for unconfined overpressure calculation is based on the work
of Bauwens and Dorfeev [67, 68]. In this method, the detonable mass within the unconfined
jet/plume is calculated and then the overpressure is based on detonation of that mass of fuel. Due
to the dependence of this model on detonation cell size, it has not been updated to work with
mixtures/blends yet.
The detonation cell size (λdet ) is calculated at each point in the jet/plume, based on the mass
fraction field (Y , see section 3.3.1) and fits to data from the detonation database [69]. The fits to
the data are based on a polynomial fit in log-log space [1]:
ln λdet = a + b ln φ + c(ln φ)2 + d(ln φ)3 + e(ln φ)4 , (131)
where the parameters a–e are given in Table 3-1.
Table 3-1 Detonation cell size fitted parameters
Fuel a b c d e
H2 2.94771698 -0.16536739 2.2608031 -1.18064551 0.45823461
CH4 5.768321 1.13938677 113.36802963 0 0
C3 H8 4.44856885 -0.73108257 5.50526263 0 0
The fits and data for different fuels are shown in Figure 3-6.
1000 103
H2
800 CH4
C3H8
Cell Size [mm]
600
102
400
H2
200 CH4
C3H8
0 101
0 1 2 3 4 5 6 10 1 100 101
Equivalence Ratio Equivalence Ratio
Figure 3-6 Detonation cell size data (points) from the detonation database [69] and fits to the data (lines) on a
linear (left) and logarithmic (right) scale.
53
Once the detonation cell size is calculated, the gradient in the radial direction of the detonation
cell size (dλdet /dr) is found numerically. In addition, the number of cells that fit within the layer
(nλdet ) is numerically calculated as
Z r
dr
nλdet = . (132)
0 λdet (r)
Detonations are presumed to propagate in areas that are within the flammability limits, with a
detonation cell size gradient less than 0.1 (dλdet /dr < 0.1), and where there are at least 5 cells
within the layer (nλdet ≥ 5). The mass of fuel within the jet/plume that meets these constraints is
calculated as the detonable mass (mdet ). In equation form,
Z ∞ Z ∞
mdet = ρY 2πrdr dS, where Ylean ≤ Y ≤ Yrich , dλdet /dr < 0.1, and nλdet ≥ 5. (133)
S=0 r=0
The dimensionless distance (R∗Bauwens ) from the center of the detonable region is calculated as
1/3
Pambient
R∗Bauwens =R , (134)
Edet
where the energy of detonable fuel (Edet ) is calculated as
Edet = mdet ∆Hc . (135)
Finally, the scaled overpressure (P∗ ) is calculated as

0.34 0.062 0.0033
P∗ = + + . (136)
(R∗ )4/3 (R∗ )2 (R∗ )3
There is currently no calculation of impulse for this model. This means that the overpressure
probit models in Section 2.6.2 that use impulse cannot be used with this model.
54
4. SUMMARY OF NUMERICAL METHODS
4.1. Python Calculation Methods
The Python modules in HyRAM+ utilize the NumPy and SciPy packages [70–72]. NumPy
provides support for multi-dimensional arrays, mathematical functions of arrays, and some
numerical linear algebra routines. SciPy provides a variety of numerical method routines
including support for statistical distributions, numerical linear algebra, integration, interpolation,
optimization, root-finding, ordinary differential equation solvers, and others. Plots in HyRAM+
are made using Matplotlib [73].
4.2. Leak Frequency Computations
Almost all of the computations for HyRAM+ are done in the Python code modules. However, the
HyRAM+ QRA mode uses statistical distributions from Math.NET [74] for calculating the
component release frequency mean and variance (see Section 2.4). This allows the leak frequency
values to update quickly in the front-end without calling Python.
4.3. Unit Conversion
HyRAM+ enforces an immutable link between values and the units that define them. Input values
are stored in the International System of Units (SI). Conversions are performed implicitly by the
system. Therefore, the application is able to present data in units preferred by the user or more
relevant in problem context, while being able to pass data to the Python calculation algorithms in
the expected SI units. Table 4-1 contains the convertible units currently available in HyRAM+.
Note that all pressure units are assumed to be absolute pressure, not gauge pressure.
Table 4-1 HyRAM+ convertible units.
Unit Type Units Available

Distance m, cm, mm, in, ft, yd, mi, au
Area m2 , cm2 , mm2 , ft2 , in2 , yd2
cm3 , dm3 , dam3 , m3 , km3 , mm3 , µm3 , ft3 , in3 ,
Volume
yd3 , mi3 , L, µL, mL, dL, daL, kL, ML
Angle radians, degrees
Energy J, kWh, BTU
Time s, ms, min, hr
Pressure Pa, kPa, MPa, psi, atm, bar, J/m3
Temperature Celsius, Fahrenheit, Kelvin
Speed m/s
Volumetric Flow Rate m3 /s
55
56
REFERENCES
[1] B. D. Ehrhart and E. S. Hecht, “Hydrogen Plus Other Alternative Fuels Risk Assessment
Models (HyRAM+) Version 4.1 Technical Reference Manual,” Tech. Rep.
SAND2022-5649, Sandia National Laboratories, April 2022.
[2] E. S. Hecht, B. D. Ehrhart, and K. M. Groth, “Hydrogen Plus Other Alternative Fuels Risk
Assessment Models (HyRAM+) Version 4.0 Technical Reference Manual,” Tech. Rep.
SAND2021-14813, Sandia National Laboratories, November 2021.
[3] B. D. Ehrhart, E. S. Hecht, and K. M. Groth, “Hydrogen Risk Assessment Models
(HyRAM) Version 3.1 Technical Reference Manual,” Tech. Rep. SAND2021-5812, Sandia
National Laboratories, May 2021.
[4] B. D. Ehrhart, E. S. Hecht, and K. M. Groth, “Hydrogen Risk Assessment Models
(HyRAM) Version 3.0 Technical Reference Manual,” Tech. Rep. SAND2020-10600, Sandia
National Laboratories, September 2020.
[5] K. M. Groth, E. S. Hecht, J. T. Reynolds, M. L. Blaylock, and E. E. Carrier, “Methodology
for assessing the safety of Hydrogen Systems: HyRAM 1.1 technical reference manual,”
Tech. Rep. SAND2017-2998, Sandia National Laboratories, March 2017.
[6] K. M. Groth, E. S. Hecht, and J. T. Reynolds, “Methodology for assessing the safety of
Hydrogen Systems: HyRAM 1.0 technical reference manual,” Tech. Rep.
SAND2015-10216, Sandia National Laboratories, November 2015.
[7] K. M. Groth and A. V. Tchouvelev, “A toolkit for integrated deterministic and probabilistic
assessment for hydrogen infrastructure,” in Proceedings of the Probabilistic Safety and
Management Conference (PSAM 12), (Honolulu, HI (USA)), June 22-27 2014.
[8] K. M. Groth and E. S. Hecht, “HyRAM: a methodology and toolkit for quantitative risk
assessment of hydrogen systems,” in Proceedings of the International Conference on
Hydrogen Safety (ICHS 2015), (Yokohama (Japan)), October 19-21 2015.
[9] K. M. Groth and E. S. Hecht, “HyRAM: A methodology and toolkit for quantitative risk
assessment of hydrogen systems,” International Journal of Hydrogen Energy, vol. 42,
no. 11, pp. 7485–7493, 2017.
[10] B. D. Ehrhart, E. S. Hecht, and J. A. Mohmand, “Validation and comparison of HyRAM
physics models,” Tech. Rep. SAND2021-5811, Sandia National Laboratories, May 2021.
[11] S. Egbert, X. Li, M. L. Blaylock, and E. Hecht, “Mixing of liquid methane releases.,” Tech.
Rep. SAND2018-13757R, Sandia National Laboratories, 2018.
[12] M. L. Blaylock, E. Hecht, and C. Jordan, “Validation of the altram physics models for use
with compressed natural gas,” Tech. Rep. SAND2019-13408, Sandia National Laboratories,
June 2020.
[13] J. G. Shum, E. S. Hecht, and M. L. Blaylock, “Validation of the HyRAM+ physics models
for use with propane,” Tech. Rep. SAND2021-3244, Sandia National Laboratories, 2021.
57
[14] K. M. Groth, J. L. LaChance, and A. P. Harris, “Early-stage quantitative risk assessment to
support development of codes and standard requirements for indoor fueling of hydrogen
vehicles,” Tech. Rep. SAND2012-10150, Sandia National Laboratories, November 2012.
[15] A. V. Tchouvelev, “Knowledge gaps in hydrogen safety: A white paper,” tech. rep.,
International Energy Agency Hydrogen Implementing Agreement Task 19, January 2008.
[16] J. LaChance, W. Houf, B. Middleton, and L. Fluer, “Analyses to support development of
risk-informed separation distances for hydrogen codes and standards,” Tech. Rep.
SAND2009-0874, Sandia National Laboratories, March 2009.
[17] A. Glover, A. Baird, and D. Brooks, “Final report on hydrogen planthazards and risk
analysissupporting hydrogen plant sitingnear nuclear power plants,” Tech. Rep.
SAND2020-7946, Sandia National Laboratories, July 2020.
[18] D. M. Brooks, B. D. Ehrhart, and C. LaFleur, “Development of liquid hydrogen leak
frequencies using a Bayesian update process,” in 2021 International Conference on
Hydrogen Safety, September 2021.
[19] D. Brooks, A. Glover, and B. D. Ehrhart, “Compressed natural gas component leak
frequency estimation,” Tech. Rep. SAND2022-14164, Sandia National Laboratories,
October 2022.
[20] G. W. Mulcahy, D. M. Brooks, and B. D. Ehrhart, “Using Bayesian methodology to estimate
liquefied natural gas leak frequencies,” Tech. Rep. SAND2021-4905, Sandia National
Laboratories, April 2021.
[21] TNO, “Methods for the determination of possible damage,” Tech. Rep. CPR 16E, The
Netherlands Organization of Applied Scientific Research (TNO), 1992.
[22] P. K. Raj, “A review of the criteria for people exposure to radiant heat flux from fires,”
Journal of Hazardous Materials, vol. 159, no. 1, pp. 61–71, 2008.
[23] N. A. Eisenberg, C. J. Lynch, and R. J. Breeding, “Vulnerability model. A simulation system
for assessing damage resulting from marine spills,” Tech. Rep. SA/A-015 245, U.S. Coast
Guard, 1975.
[24] C. K. Tsao and W. W. Perry, “Modifications to the vulnerability model: a simulation system
for assessing damage resulting from marine spills,” Tech. Rep. ADA 075 231, U.S. Coast
Guard, 1979.
[25] F. P. Lees, “The assessment of major hazards: a model for fatal injury from burns,” Process
Safety and Environmental Protection, vol. 72, no. 3, pp. 127–134, 1994.
[26] J. LaChance, A. Tchouvelev, and A. Engebø, “Development of uniform harm criteria for use
in quantitative risk analysis of the hydrogen infrastructure,” International Journal of
Hydrogen Energy, vol. 36, pp. 2381–2388, February 2011.
[27] Center for Chemical Process Safety (CCPS), Guidelines for Chemical Process Quantitative
Risk Analysis. American Institute of Chemical Engineers, 1999.
58
[28] HSE, Major hazard aspects of the transport of dangerous substances. UK Health and Safety
Executive, 1991.
[29] UK Health and Safety Executive, “Methods of approximation and determination of human
vulnerability for offshore major accident hazard assessment,” tech. rep., UK Health and
Safety Executive, 2010.
[30] I. H. Bell, J. Wronski, S. Quoilin, and V. Lemort, “Pure and Pseudo-pure Fluid
Thermophysical Property Evaluation and the Open-Source Thermophysical Property
Library CoolProp,” Industrial & Engineering Chemistry Research, vol. 53, pp. 2498–2508,
February 2014.
[31] J. W. Leachman, R. T. Jacobsen, S. G. Penoncello, and E. W. Lemmon, “Fundamental
equations of state for parahydrogen, normal hydrogen, and orthohydrogen,” Journal of
Physical and Chemical Reference Data, vol. 38, pp. 721–748, September 2009.
[32] U. Setzmann and W. Wagner, “A new equation of state and tables of thermodynamic
properties for methane covering the range from the melting line to 625 K at pressures up to
100 MPa,” Journal of Physical and Chemical Reference Data, vol. 20, no. 6, pp. 1061–1155,
1991.
[33] E. W. Lemmon, M. O. McLinden, and W. Wagner, “Thermodynamic properties of propane.
III. A reference equation of state for temperatures from the melting line to 650 K and
pressures up to 1000 MPa,” Journal of Chemical & Engineering Data, vol. 54, no. 12,
pp. 3141–3180, 2009.
[34] O. Kunz, R. Klimeck, W. Wagner, and M. Jaeschke, The GERG-2004 wide-range equation
of state for natural gases and other mixtures. VDI Verlag GmbH, Jul 2007.
[35] O. Kunz and W. Wagner, “The GERG-2008 wide-range equation of state for natural gases
and other mixtures: An expansion of GERG-2004,” Journal of Chemical & Engineering
Data, vol. 57, no. 11, pp. 3032–3091, 2012.
[36] E. W. Lemmon and R. T. Jacobsen, “A generalized model for the thermodynamic properties
of mixtures,” International Journal of Thermophysics, vol. 20, pp. 825–835, 1999.
[37] E. W. Lemmon, R. T. Jacobsen, S. G. Penoncello, and D. G. Friend, “Thermodynamic
properties of air and mixtures of nitrogen, argon, and oxygen from 60 to 2000 k at pressures
to 2000 mpa,” Journal of Physical and Chemical Reference Data, vol. 29, no. 3,
pp. 331–385, 2000.
[38] E. W. Lemmon and R. T. Jacobsen, “Equations of state for mixtures of r-32, r-125, r-134a,
r-143a, and r-152a,” Journal of Physical and Chemical Reference Data, vol. 33, no. 2,
pp. 593–620, 2004.
[39] I. W. Ekoto, A. J. Ruggles, L. W. Creitz, and J. X. Li, “Updated jet flame radiation modeling
with buoyancy corrections,” International Journal of Hydrogen Energy, vol. 39,
pp. 20570–20577, December 2014.
[40] National Institute of Standards and Technology, “NIST Chemistry WebBook, SRD 69.”
https://webbook.nist.gov/chemistry/.
59
[41] S. Mannan, ed., Lees’ Loss Prevention in the Process Industries. Elsevier Science &
Technology, 3 ed., 2005.
[42] M. J. Hurley, D. T. Gottuk, J. R. H. Jr., K. Harada, E. D. Kuligowski, M. Puchovsky, J. L.
Torero, J. M. W. Jr., and C. J. Wieczorek, eds., SFPE Handbook of Fire Protection
Engineering. Springer, New York, NY, 5 ed., 2016.
[43] J. Travis, D. Piccioni Koch, and W. Breitung, “A homogeneous non-equilibrium two-phase
critical flow model,” International Journal of Hydrogen Energy, vol. 37, no. 22,
pp. 17373–17379, 2012.
[44] T. Spicer and D. Miller, “Quantifying the mass discharge rate of flashing two phase releases
through simple holes to the atmosphere,” Process Safety Progress, vol. 37, no. 3,
pp. 382–396, 2018.
[45] A. G. Venetsanos, “Homogeneous non-equilibrium two-phase choked flow modeling,”
International Journal of Hydrogen Energy, vol. 43, no. 50, pp. 22715–22726, 2018.
[46] K. B. Yüceil and M. V. Ötügen, “Scaling parameters for underexpanded supersonic jets,”
Physics of Fluids, vol. 14, pp. 4206–4215, October 2002.
[47] A. D. Birch, D. J. Hughes, and F. Swaffield, “Velocity decay of high pressure jets,”
Combustion Science and Technology, vol. 52, no. 1-3, pp. 161–171, 1987.
[48] A. D. Birch, D. R. Brown, M. G. Dodson, and F. Swaffield, “The structure and concentration
decay of high pressure jets of natural gas,” Combustion Science and Technology, vol. 36,
no. 5-6, pp. 249–261, 1984.
[49] B. C. R. Ewan and K. Moodie, “Structure and velocity measurements in underexpanded
jets,” Combustion Science and Technology, vol. 45, no. 5-6, pp. 275–288, 1986.
[50] V. Molkov, D. Makarov, and M. Bragin, “Physics and modelling of under-expanded jets and
hydrogen dispersion in atmosphere,” in Proceedings of the 24th International Conference on
Interaction of Intense Energy Fluxes with Matter, March 1-6 2009.
[51] W. Houf and W. Winters, “Simulation of high-pressure liquid hydrogen releases,”
International Journal of Hydrogen Energy, vol. 38, pp. 8092–8099, June 2013.
[52] W. Houf and R. Schefer, “Analytical and experimental investigation of small-scale
unintended releases of hydrogen,” International Journal of Hydrogen Energy, vol. 33,
pp. 1435–1444, February 2008.
[53] W. S. Winters, “Modeling leaks from liquid hydrogen storage systems,” Tech. Rep.
SAND2009-0035, Sandia National Laboratories, January 2009.
[54] G. Abraham, “Horizontal jets in stagnant fluid of other density,” Journal of the Hydraulics
Division, vol. 91, no. 4, pp. 139–154, 1965.
[55] M. Hosseini, I. Dincer, G. Naterer, and M. Rosen, “Thermodynamic analysis of filling
compressed gaseous hydrogen storage tanks,” International Journal of Hydrogen Energy,
vol. 37, pp. 5063–5071, March 2012.
60
[56] B. Lowesmith, G. Hankinson, C. Spataru, and M. Stobbart, “Gas build-up in a domestic
property following releases of methane/hydrogen mixtures,” International Journal of
Hydrogen Energy, vol. 34, pp. 5932–5939, July 2009.
[57] W. Houf and R. Schefer, “Predicting radiative heat fluxes and flammability envelopes from
unintended releases of hydrogen,” International Journal of Hydrogen Energy, vol. 32,
pp. 136–151, January 2007.
[58] S. R. Turns and F. H. Myhr, “Oxides of nitrogen emissions from turbulent jet flames: Part
i—fuel effects and flame radiation,” Combustion and Flame, vol. 87, no. 3, pp. 319–335,
1991.
[59] R. Schefer, W. Houf, B. Bourne, and J. Colton, “Spatial and radiative properties of an
open-flame hydrogen plume,” International journal of hydrogen energy, vol. 31, no. 10,
pp. 1332–1340, 2006.
[60] P. P. Panda and E. S. Hecht, “Ignition and flame characteristics of cryogenic hydrogen
releases,” International Journal of Hydrogen Energy, vol. 42, pp. 775–785, January 2017.
[61] A. Molina, R. W. Schefer, and W. G. Houf, “Radiative fraction and optical thickness in
large-scale hydrogen-jet fires,” Proceedings of the Combustion Institute, vol. 31,
pp. 2565–2572, January 2007.
[62] F. D. Wayne, “An economical formula for calculating atmospheric infrared transmissivities,”
Journal of Loss Prevention in the Process Industries, vol. 4, pp. 86 – 92, January 1991.
[63] G. Hankinson and B. J. Lowesmith, “A consideration of methods of determining the
radiative characteristics of jet fires,” Combustion and Flame, vol. 159, pp. 1165–1177,
March 2012.
[64] C. Bauwens and S. Dorofeev, “CFD modeling and consequence analysis of an accidental
hydrogen release in a large scale facility,” International Journal of Hydrogen Energy,
vol. 39, pp. 20447–20454, December 2014.
[65] S. Jallais, E. Vyazmina, D. Miller, and J. K. Thomas, “Hydrogen jet vapor cloud explosion:
A model for predicting blast size and application to risk assessment,” Process Safety
Progress, vol. 37, no. 3, pp. 397–410, 2018.
[66] Center for Chemical Process Safety, Guidelines for Vapor Cloud Explosion, Pressure Vessel
Burst, BLEVE, and Flash Fire Hazards. John Wiley & Sons Ltd., 2010.
[67] C. R. L. Bauwens and S. B. Dorofeev, “Modeling detonation limits for arbitrary
non-uniform concentration distributions in fuel–air mixtures,” Combustion and Flame,
vol. 221, pp. 338–345, 2020.
[68] C. R. L. Bauwens and S. B. Dorofeev, “Quantifying the potential consequences of a
detonation in a hydrogen jet release,” in Proceedings of the International Conference on
Hydrogen Safety (ICHS 2019), (Adelaide (Austrailia)), pp. 612–623, September 24 - 26
2019.
61
[69] California Institute of Technology and J. E. Shepherd, “Detonation Database.”
https://shepherd.caltech.edu/detn_db/html/db_10.html.
[70] T. E. Oliphant, A guide to NumPy, vol. 1. Trelgol Publishing USA, 2006.
[71] S. Van Der Walt, S. C. Colbert, and G. Varoquaux, “The NumPy array: a structure for
efficient numerical computation,” Computing in Science & Engineering, vol. 13, no. 2,
p. 22, 2011.
[72] P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau,
E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson,
K. Jarrod Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. Carey,
İ. Polat, Y. Feng, E. W. Moore, J. VanderPlas, D. Laxalde, J. Perktold, R. Cimrman,
I. Henriksen, E. A. Quintero, C. R. Harris, A. M. Archibald, A. H. Ribeiro, F. Pedregosa,
P. van Mulbregt, and SciPy 1.0 Contributors, “SciPy 1.0: Fundamental Algorithms for
Scientific Computing in Python,” Nature Methods, vol. 17, pp. 261–272, 2020.
[73] J. D. Hunter, “Matplotlib: A 2D graphics environment,” Computing in Science &
Engineering, vol. 9, no. 3, pp. 90–95, 2007.
[74] “Math.NET Numerics.” https://numerics.mathdotnet.com/.
62
DISTRIBUTION
Email—Internal
Name Org. Sandia Email Address
Ethan Hecht 8367 [email protected]
Kristin Hertz 8367 [email protected]
Myra Blaylock 8751 [email protected]
Brian Ehrhart 8854 [email protected]
Chris LaFleur 8854 [email protected]
Ben Schroeder 8854 [email protected]
Technical Library 1911 [email protected]
Email—External
Name Company Email Address Company Name
Laura Hill [email protected] DOE EERE HFTO
Michael Laughlin [email protected] DOE EERE VTO
Mark Smith [email protected] DOE EERE VTO
63
Sandia National Laboratories is a
multimission laboratory managed
and operated by National
Technology & Engineering
Solutions of Sandia LLC, a wholly
owned subsidiary of Honeywell
International Inc., for the U.S.
Department of Energy’s National
Nuclear Security Administration
under contract DE-NA0003525.

SAND2022 16425 HyRAMplus 5.0 Technical Reference Manual

Uploaded by

Copyright:

Available Formats

SAND2022 16425 HyRAMplus 5.0 Technical Reference Manual

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SAND2022 16425 HyRAMplus 5.0 Technical Reference Manual

Uploaded by

Copyright:

Available Formats

SANDIA REPORT

Hydrogen Plus Other Alternative Fuels Risk

Telephone: (865) 576-8401

Telephone: (800) 553-6847

2. Quantitative Risk Assessment Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4. Summary of Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

Table 2-1. Default ignition probabilities for different fuels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

AIR average individual risk

CFD computational fluid dynamics

DOE U.S. Department of Energy

EERE Office of Energy Efficiency and Renewable Energy

ESD event sequence diagram

FAR fatal accident rate

FTC failure to close

FTO failure to open

HFTO Hydrogen and Fuel Cell Technologies Office

HSE U.K. Health and Safety Executive

HyRAM Hydrogen Risk Assessment Models

HyRAM+ Hydrogen Plus Other Alternative Fuels Risk Assessment Models

NFPA National Fire Protection Association

PHMSA Pipeline and Hazardous Materials Safety Administration

PLL potential loss of life

PRD pressure relief device

QRA quantitative risk assessment

SI International System of Units

TNO Netherlands Organisation for Applied Scientific Research

VTO Vehicle Technologies Office

1.1. About HyRAM+ and This Report

1.2. Design Goals and Limitations

1.3. Summary of HyRAM+ Outputs

1.4. Summary of Changes Made

2.1. Quantitative Risk Assessment Methodology Overview

Risk = ∑(pi × ci ) (1)

2.2. Risk Metrics Calculations

AIR = H × FAR × 10−8 (4)

2.3. Scenario Models

fIsolated = fRelease × pIsolated (5)

fUnignited = fRelease × (1 − pIsolated ) × (1 − pImmed. Ignite − pDelayed Ignite ) (6)

fJetfire = fRelease × (1 − pIsolated ) × pImmed. Ignite (7)

fExplosion = fRelease × (1 − pIsolated ) × pDelayed Ignite (8)

2.3.2. Default Ignition Probabilities

Table 2-1 Default ignition probabilities for different fuels.

(a) Hydrogen (b) Methane and Propane

Release Ignition Probability Release Ignition Probability

2.4. Frequency of a Release

fRelease, k=100% = fRandom Releases, k=100% + fOther Releases (11)

2.4.1. Frequency of Random Leaks

Heat Loading Component

fRandom Releases,k = ∑ NComponenti × fLeak,i,k (12)

2.4.2. Default Component Leak Frequencies

fLeak,i,k ∼ Lognormal(µ, σ2 ) (13)

0.01% Leak Area

0.01% Leak Area

fOther Releases = fFueling Demands × pDispenser Releases (15)

Rupture Manual Solenoid

Overpressure Pressure Breakaway Solenoid Common

pDispenser Releases = pAccidents + pShutdown Failure (17)

pAccidents = pRupture During Fueling + pRelease Due to Drive−Off (18)

pRupture During Fueling = pOverpressure During Fueling × pPRD FTO (19)

pRelease Due to DriveOff = pDriveOff × pBreakaway FTC (20)