Review Article

Proc IMechE Part G:

J Aerospace Engineering
A review of aircraft environmental control 2023, Vol. 237(11) 2453–2467
© IMechE 2023

system simulation and diagnostics

Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/09544100231154441
journals.sagepub.com/home/pig
Shafayat H Chowdhury , Fakhre Ali  and Ian K Jennions

Abstract
The aircraft Environmental Control System (ECS) enables the aircraft to maintain a comfortable and safe environment for
its passengers throughout its operating envelope. The Pressurised Air Conditioner (PACK) is the heart of the ECS, and is
composed of multiple sub-systems: heat exchangers, valves, compressor, turbine, and a water separator. The PACK’s
principle function is to enable conditioning of the hot, high pressure bleed air from the engine or APU, for temperature,
pressure and humidity against the cabin requirements. The operation of the PACK is governed by a control system which
has the ability to mask degradation in its component during operation until severe degradation or failure results. The
required maintenance is then both costly and disruptive. The PACK has been reported as major driver of unscheduled
maintenance by the operators. The aviation industry is currently proactively exploring innovative health management
solutions that aid the maintenance of aircraft key systems based on predictive based maintenance approaches using online
condition monitoring techniques. This paper presents a comprehensive review of the simulation and diagnostic meth-
odologies applicable to fault diagnostics of the ECS PACK. The existing literature suggests that model-based and data-driven
methods are effective for conducting fault detection and isolation of the PACK system. The conceived findings indicate that
the model-based diagnostic approach have been extensively employed to conduct PACK diagnostics at component level
only. Their successful implementation requires robust experimental verification and validation against the actual data under
healthy and faulty conditions. Although a substantial amount of work has been reported on developing first principles based
simulation models and diagnostic strategies for the ECS, the acquired findings suggest that there is a compelling need for
a verified and validated ECS simulation model to enable accurate PACK system-level diagnostics based on single and
multiple component level degradation scenarios. It has also been identified that the existing literature lacks the evaluation of
humidity regulation and the effect of the control system on the PACK performance characteristics. Finally, a taxonomy of
diagnostic techniques and simulation models is compiled based on the available literature.

Keywords
diagnostics, environmental control system, heat exchangers, passenger air conditioner, simulation

Date received: 28 July 2022; accepted: 10 January 2023

Introduction of PACK failure affect the pressurisation of the cabin. The

PACK functionality is also crucial for maintaining hu-
The Environmental Control System (ECS) of an aircraft midity level in the cabin as overheating of excess moisture
provides conditioned air to pressurise the cabin and can cause misting in the cockpit, and the cabin, which can
regulate the temperature and humidity throughout cabin. It hinder the safety of passengers and crew.4 The operations
also supplies the cooling air into the avionics bay to of the PACK are governed by the control system which has
prevent the electronic equipment from overheating. It the ability to mask degradation in its component during
contains two Passenger Air Conditioners (PACK), which operation until severe degradation or failure results. The
are the primary system for conditioning the airflow within required maintenance is then both costly and disruptive.
the ECS.1 Although the system is quite robust, there have The PACK has been reported as major driver of
been a few incidents reported due to system failure. A
simple ECS component fault occurrence can escalate and
result in system failure. For example, a Ram air actuator IVHM Centre, Cranfield University, Bedfordshire, UK
failure can result in reduction of cold mass flow over the
Corresponding author:
heat exchanger causing an instant PACK shutdown due to
Fakhre Ali, Propulsion Centre, Cranfield University, Building 52, College
overheating.2 Overheating can also occur due to a change Road, Cranfield MK43 0AL, UK.
of the bleed air source at the PACK input.3 These incidents Email: f.ali@cranfield.ac.uk
2454 Proc IMechE Part G: J Aerospace Engineering 237(11)

unscheduled maintenance by the operators. The aviation enable accurate diagnostics, there is a need to develop
industry is currently proactively exploring innovative system-level simulation models. The fault detection and
health management solutions that aid the maintenance of prediction analysis can dictate the supply-chain and lo-
aircraft key systems based on predictive based mainte- gistics to best prepare for scheduled maintenance, creating
nance approaches using online condition monitoring the new manufacturing paradigm supporting predictive
techniques. maintenance 12[p. 4].
The safety of the crew and operational cost was a major
concern when NASA began human space flight.5 After the
Apollo success, the concept of Integrated Vehicle Health
Environmental control system
Management (IVHM) began to surface in the 70s to ad- Air conditioning systems are designed to provide con-
dress the safety and cost factors. In 1992, NASA defined ditioned air ensuring human comfort. They are applicable
the IVHM goals and objectives in a report.6 These ob- in aircraft, automobile, offices, industries and various
jectives can be summarised as the capability to facilitate other places. During the mid-1940s, the pioneering
optimised maintenance actions, operational cost re- Boeing 307 Stratoliner was the first aircraft equipped with
duction, improved readiness of the fleet and enhanced air conditioning system, ECS, for cabin pressurisation.
safety.7 This enabled high altitude flights within thin air and sub-
Zhang et al.8 suggests that civil aviation maintenance zero temperature regions above the weather and
costs are estimated to make up 10% of airline operating turbulence.13,14
cost, with spending reported to be around US$83 billion ECS is a generic term for the subsystems and equip-
in 2019. Considering the industry efforts to build and ment associated with ventilation, heating, cooling, con-
recover from the Corona Virus 2019 (COVID19), these tamination control, and pressurization in the occupied
costs are anticipated to reach US$115 billion by 2030. compartments, cargo bays and electronic racks. As il-
This motivates airliners to take the IVHM concept de- lustrated in Figure 1, the overall environmental control
veloped by NASA further by integrating health moni- system of a typical civil aircraft is composed of several
toring technologies into their asset.9 It helps in subsystems, which are the Bleed Air System (BAS), the
monitoring the critical components of a system of their Anti-Icing System (AIS), the Pressurized Air Conditioner
asset and cuts down on unscheduled maintenance and (PACK) and the Cabin Pressure Control System (CPCS).
downtime.10 Thus, the use of IVHM toolsets is enabling The bleed system provides the pressurised air to the PACK
a shift from preventive maintenance over to the pre- for conditioning. Part of the bleed mass flow before en-
dictive maintenance. tering the PACK is taken out to drive the Ram turbofan to
The data collected from the aircraft are instrumental in drive cold ambient air over the heat exchangers in the
conducting diagnostics and prognostics using the model- PACK to enable pre-cooling of the bleed air. After con-
based or data-driven technique. A vehicle level reasoning ditioning of the bleed air in the PACK, the air is mixed
system can be used to identify faults considering the in- with trim and recirculating air from the cabin in the mixing
teraction between different systems.11 In order to un- manifold before it is distributed to different zones of the
derstand interdependencies between components, and to aircraft.

Figure 1. ECS schematic.15

Chowdhury et al. 2455

The PACK is the primary system for conditioning the Integrated vehicle health management
airflow within the ECS. Figure 2 illustrates the schematic
of a single PACK with HPWS in a B737-400 aircraft. It After the introduction of IVHM by NASA, the aviation
consists of Valves, Heat Exchangers, an Air-Cycle Ma- industry started utilising it in accordance with the new
chine (ACM) and a High-Pressure Water Separator Product Service System (PSS) business model. It helps in
(HPWS). There are other ECS systems, such as electri- monitoring the critical components of a system of an asset
cally driven ECS (used in modern aircraft)16 [p. 3], and cuts down on unscheduled maintenance and down-
pneumatic ECS with low-pressure water separator time, defining the company’s profit margin.10 Un-
(LPWS) (generally used in older generation aircraft),17 scheduled maintenance plays a key role in increasing
vapour cycle ECS (used for light turboprop aircraft ap- maintenance costs. Deploying IVHM tools on expensive
plication)18 and membrane-based dehumidifier (used in assets such as the aircraft, raises the profit margin through
spacecraft).19 the reduction in downtime during maintenance and as
The PACK consists of a PACK valve (PV) and Tem- a result enhancing fleet availability.
perature Control Valves (TCV) which regulate the hot Figure 3 distinguishes between preventive and pre-
dictive maintenance. The manufacturing sector of the
mass flow through the system and the core. The Primary
aviation industry typically follows the preventive main-
Heat Exchanger (PHX) and Secondary Heat Exchanger
tenance approach where the aircraft is grounded fre-
(SHX) use the cold Ram air as a heat sink. In between the
quently for routine maintenance regardless of fault
heat exchangers sits the refrigeration unit of the PACK.
occurrences. This increases downtime as all critical
The ACM drives the air through the core and contributes components go through thorough inspection. The idea of
to cabin pressurisation. Towards the end of the PACK sits IVHM is to monitor the health condition of the critical
the HPWS consisting of Reheater (RHX), Condenser systems using sensor technology to detect and predict fault
(CHX) and Water Separator (WS) which regulates air occurrences. At the fault prognostics stage shown in
temperature to enable condensation and extraction of Figure 3, the remaining useful life of components is es-
water from the air to regulate humidity. The PACK has timated, based on that maintenance actions are planned.
a control system that governs the opening of the valve by The early detection of failure and prognostics helps cut
using temperature sensors at the WS outlet (to regulate the down on unscheduled maintenance and also ensures the
TCVopening) and Compressor outlet (to regulate the Ram safety of passengers and crew.
air mass flow modulating door) to meet cabin demand The sensor data collected from the aircraft is in-
temperature. In addition to that, there are three temperature strumental in conducting diagnostics and prognostics
switches: at the compressor outlet, WS outlet and PACK using a model-based or data-driven technique. In order to
outlet, which shut down the PV to prevent any damage to understand interdependencies between components to
the system in the case of overheating. enable accurate diagnostics it requires development of

Figure 2. Schematic of the 737–400 ECS PACK with HPWS.

2456 Proc IMechE Part G: J Aerospace Engineering 237(11)

Figure 3. Maintenance flow chart: Preventive (left) and Predictive (right).

system-level simulation model. Then a vehicle level

reasoning system is used to identify faults considering the
interaction between different systems.11 This use of
vehicle level reasoning proposed by Cordelia et al.,11 is a
novel way of using reasoning to enable condition-based
maintenance. The overall analysis of aircraft health
condition and interdependencies between major
systems can then be used to well prepare the maintenance
actions.12 [p.4]

Scope of present work

This paper provides a literature survey on the ECS towards
its diagnosis which includes the following: Figure 4. Typical fin-and-plate type cross-flow heat
exchanger.24
(I) Identification of diagnostic methodologies appli-
cable for the ECS
(II) A review on system-level simulation models for the
Model-based method
ECS The model-based method relies on a system model with
(III) Experimental investigation on PACK performance the mathematical equations defining the input/output
conditions.28 This technique involves installing sensors
at the optimised locations to acquire health defining pa-
Diagnostic methodologies rameters and comparing them against the model output.
Heat exchangers within the PACK have been extensively This sensor set is unique for each system and can be
investigated by researchers.20–23 The aircraft ECS heat established by studying the thermodynamic PACK per-
exchangers are made of fin and plate assemblies as shown formance characteristics. This study of PACK perfor-
mance characteristic under heat exchanger fouling is
in Figure 4, which uses cold Ram air as a heat sink to cool
demonstrated by Jennions and Ali.29 The heat exchanger
the hot bleed air. The ram air often contains debris which
fouling has been simulated in the model by manipulating
accumulates at the cold-side of the heat exchanger re-
the heat exchanger effectiveness. Their results suggest that
sulting in a drop in heat exchanger effectiveness.24 Due
the PACK performance under PHX and SHX degradation
to drop in effectiveness, the heat transfer across the hot
are very different to each other. The deviation between the
and cold side will be impacted and the system will model data (healthy baseline under given operating
overheat imposing a risk of immediate PACK shutdown. conditions) and test data can then be used as an indication
Therefore, the heat exchangers were considered to be of fault occurrences.[p.19]30 Figure 5 demonstrates
critical components of the aircraft ECS system which a typical model-based approach that relies on residual
requires appropriate diagnostic rules.25 The diagnostic evaluation. The difference, or residual, is usually gener-
methodology that has been implemented can be classified ated by one of three different methods: (i) Parameter
as: (i) Model-based method,26 (ii) Data-Driven method9 Estimation, (ii) State-observer and (iii) Parity Space.31
and (iii) Hybrid method.27 The hybrid method is
a combination of both model-based and data-driven Parameter estimation and state-observer. Fouling in heat
methods. exchanger is due to particulate accumulation. This can
Chowdhury et al. 2457

Figure 5. Concept of model-based diagnostic approach.30 [p. 19]

occur in both the hot and cold side. Research indicates the understand the input-output behaviour of the system.37
parameters that need to be monitored for diagnosing heat For this, experimental component performance analysis is
exchanger fouling are the heat transfer coefficient, the required.
pressure drop, the mass flow rates, the temperature var- The heat exchanger fouling can also be detected by
iation and the weight of the plates.32 Deviation in the analysing the PACK control system. The PV and TCV
estimated parameters over time is used as an indication of regulate mass flow through the system to match cabin
fault occurrences. To estimate the values of model pa- demand temperature. The variation in hot and cold mass
rameters using parameter estimation method, a dynamic flow affects the heat transfer coefficient. The valves are
model of the system is required. An example of such regulated by a control signal which drives the motor to
dynamic model in the form of physical state space model rotate a butterfly valve. Given the ram mass flow is
representation of a heat exchanger proposed in Reference constant, estimating the deviation of valve command from
32, where the hot inlet temperature (Th, 1 ), hot outlet fault free condition can also help identify heat exchanger
temperature (Th, 2 ), cold inlet temperature (Tc, 1 ) and cold fault occurrence. Shang and Liu24 proposed that the fault
outlet temperature (Tc, 2 ) are the model state. The model detection of the heat exchanger can be monitored by
parameters are α, β, τ h and τ c .A detailed discussion of the estimating the deviation of valve control command using
derivation and parameterisation can be found in the ref- Kalman filter. The proposed methodology as shown in
erence article.20 Shah et al.33 and Jonsson et al.32 used Figure 6 is based on the error (ε) between the output
extended Kalman filter (EKF) on such state space model temperature (Tload ) and simulated output temperature
of heat exchanger to estimate the required parameters. The b load ) and control command ( b
(T ubypass ).
EKF is most commonly used to diagnose heat exchanger The valves regulating the mass flows in the PACK are
fouling using the model-based technique.32–35 regulated using series of temperature sensors installed at
The traditional EKF used to estimate the state pa- various locations across the ECS. These sensors can be
rameters for detecting degradation or fouling can poten- faulty which can affect mass flow through the core of the
tially lose its monitoring capability in case of abrupt PACK. Wang and Wang38 developed a fault diagnosis
changes in the process states. This problem was resolved strategy by employing a genetic algorithm technique for
using two Kalman filters, simultaneously forming a dou- parameter estimation which can be used to detect sensor
ble model filter23 which is used to trace fault-free and faults.
faulty state of a non-linear system. For both the methods, Another approach to heat exchanger fouling detection
the selection of the initialisation value for the parameter was proposed and experimentally validated by Palmer
can be difficult, for which an adaptive filter such as strong et al.,22 where a manually initiated built-in-test (iBIT)
tracking filter (STF) is developed by Ma et al.,34 to im- method was used. Within parameter estimation method
prove the existing EKF that can be used to estimate the Rajarman.S et al.,39 used Kharitonov’s theory instead of
fault parameters. The STF is an extension of the EKF the EKF, due to its suitability in handling multiplicative
which has the capability to perform online state estimation fault detection.31 Furthermore, Matthew D. et al.,40,41
regardless of the system reaching steady-state.36 The developed a physics-based approach in estimating RUL,
authors also used a modified Bayes classification algo- where particle filter based on the parameter estimation
rithm to detect and isolate fault.34 method was implemented for predicting the life-cycle of
Similarly, the state-observer method uses model input a pneumatic valve with limited sensing conditions.
data to estimate state variables, based on suitable esti-
mation method. Once the system output is reconstructed Parity-space. As opposed to the parameter estimation and
using the estimated state variables, error between the state-space observer approach, the parity space approach
simulation and actual output is to be calculated which will works with the state-space representation of dynamic
be zero if the system is fault free, and vice versa.30 For systems. Without requiring detailed knowledge in ad-
these two methods to be effective, it is essential to vanced control theory of the system, the parity-space
2458 Proc IMechE Part G: J Aerospace Engineering 237(11)

Analysis (PCA) approach within the MSA framework is

popular due to its efficiency in minimising mean squared
distance between high dimensional and complex data
points to analyse data variability (i.e. deviation between
the healthy case and faulty data sets).49,50
The heat exchanger of the ECS is a complex system in
which the output is dependent on both the inlet temper-
ature on the hot-side and the mass-flow rate and tem-
perature of the Ram air in the cold-side. Najjar et al.51
adapted PCA technique for extracting features to analyse
the output temperature of the heat exchanger. From the
Figure 6. Heat exchanger fouling detection method based on
extracted principal components of the heat exchanger,
TCV control command.24 faults were classified using Support Vector Machine
(SVM) and k-Nearest Neighbor (k-NN) methods to rec-
ognise the pattern within the data set. Similarly, the PCA
diagnostic method works by developing a simple math- approach has been proven to enhance fault detection and
ematical algorithm equating the input–output state vector isolation (FDI) capability through automatic identification
and the past state vector.[p.17]37 Due to the use of simple of the diagnostic signals obtained.52,53
algebraic equations with state-space equations, the parity
space is deemed to be simpler compared to observer Artificial-intelligence
approach in a study by Yu et al.42 Expert system. An expert system is a rule-based
Mostofi et al.43 used this parity-space method to de- decision-making technique which is often used in the
velop a diagnostic algorithm to isolate the actuator and realm of data-driven diagnostics. It first emerged in the
sensor faults in an automobile engine. The throttle angle 1970s and was one of the successful forms of AI tech-
has been taken as an input and engine revolutions, niques.54 Bruton et al.55 in their literature review sug-
manifold air-pressure and temperature as the state varia- gested the expert system to be relatively easy to develop,
bles as outputs. To avoid the complexity of a nonlinear however, the disadvantage it that it requires complex rules
state-space model, the Jacobian Method was used for for complex systems. It is also suggested that rule-based
linearization, and a parity matrix was formed which was expert systems FDI method have excellent potential for
then used to design a weight matrix satisfying the residual market deployment due to its relatively straightforward
matrix. The proposed diagnostic method has been vali- development process.
dated using real engine data. Similarly, Varrier et al.,44 Fuzzy logic. Fuzzy logic, first introduced in 1965, works
conducted fault detection within a linear parameter by defining partial truth values ranging between 0 and 1.56
varying systems where parity space approach was used, These fuzzy sets help to represent imprecise information
including the time-varying parameters in the parity matrix. in terms of numerical values. Once fuzzy variables are
The parity-space approach has the capability to detect obtained, if-then rules can be developed for fault
additive faults and to check the sensor deviations effec- classification.57
tively, however, it is sensitive to noise. Hwang and Huh,45 The fuzzy-model-based control scheme was developed
developed a diagnostic method coupling the parity-space to predict performance for the control system of a building
and observer approaches to achieve effective diagnostic air conditioning system with existing water-side and air-
capability. Numerous other studies exist that used parity- side fouling on a heat exchanger.58 Fuzzy model and rule-
space for detecting faults when noise has been injected based methodologies have limitations after a certain point
into the system. These include using the parity-space of accuracy. As highlighted by Najafi et al.,59 as the
approach for handling nonlinear systems and system complexity of the system develops, it requires substantial
behavioural indicative matrix for power trains.46–48 effort in adjusting and tuning fuzzy sets either manually or
through complex algorithms. Subsequently, the com-
Data-driven method plexity in the rule-based method increases with the rise in
complexity of the system failure (to the point when there is
The data-driven method for diagnostics can be funda- a higher level of uncertainties and measurement errors).
mentally categorised into the following: (i) Statistical, and For those circumstances, the machine learning approach
(ii) Artificial-Intelligence. was taken where a behavioural pattern was to be generated
from previous fault knowledge and comparing that with
Statistical. The data-driven method is solely dependent on the observed behaviour of the system.59 This method
the quality and quantity of the data being collected from heavily depends on the amount of data available for de-
the model without having prior domain knowledge on the veloping a base-line behavioural pattern for the system.
physics governing the functionality of the system. The Artificial neural network. The ANN is a branch of the
output of systems such as the ECS is influenced by more machine learning technique in diagnostics of system
than one parameter, which necessitates multivariate failure. The ANN method was developed by depicting
statistical analysis (MSA). The Principal Component biological neural networks with neurons and the
Chowdhury et al. 2459

connection between them. The ANN model consists of are deemed applicable to the aircraft ECS system is
three layers – input, hidden, and output – of which the highlighted in Figure 7.
weights and biases need to be parameterised to fit the data
set.60 This method does not require domain knowledge of Finding 2. The open literature lacks comprehensive study
the system, but it requires data sets to train the algorithm.61 on the ACM and the HPWS. The ACM can impact the
It can be used to conduct behavioural analysis to develop overall performance of the PACK functionality as dem-
patterns which are then used for FDI.62 The ANN tech- onstrated by Santos et al.72 where the authors conducted
nique was applied in the ECS for system-level fault de- a thermodynamic analysis on a 3 wheel bootstrap ACM.
tection and isolation by developing diagnostic tree This study validates the need for investigating the ACM in
defining the entire system as the input node with the greater depth. Furthermore, an in-depth study on the
components as the branches.63 The neural network model humidity regulation and the diagnostic of HPWS is
and algorithm was constructed to isolate faults in a top- missing in the open literature.
down fashion.
Allen et al.64 combined the fuzzy model and ANN
pattern recognition techniques to perform fault detection System-level simulation of ECS
and classification. This method was suggested to be ef-
Complex systems such as the ECS requires system-level
fective due to its capability of solving complicated non-
analysis, as there are components such as valves which can
linear systems. To further improve the fault detection
mask any fault occurrence and fault can also propagate
capability, a dual-neural network structure coupled with
through the system. There are only a few full-scale
subtractive clustering analysis technique was developed.65
simulation models available for the ECS. They were
Another approach to machine learning is to implement
mostly developed for assisting in designing a new com-
Deep Neural Network (DNN), which uses hidden layers
ponent. These simulation models are cost-effective ways
between the input and output layers.66 Data-driven
of testing a model before it goes into manufacturing and
techniques are now prevalent in the industry and acade-
tests. As these simulation models provide a detailed un-
mia; however, the model-based approach for fault de-
derstanding of the system, they can be used for diagnostic
tection has been popular in the past and still used by
purposes. This section discusses the existing models of the
researchers.67–69
ECS and their functionality.

Hybrid method Functional model library of the environmental

A model can either be static or dynamic.70 The dynamic control system
models are deemed to be more robust, as it can reduce the
difference between the model and the actual system output FLECS is an ECS modelling library programmed using
by taking into consideration the degradation of the MATLAB/Simulink software packages.73 It is a project
component through time.28 These types of models are conducted by Airbus, Hamburg University of Applied
mostly a set of differential equations. The static models are Science and CeBeNetwork GmbH to construct a system-
developed using polynomial equations, which depends on level model for the ECS. The library contains healthy
the physical process constraints (i.e. effectiveness, re- models of different component blocks from BAS, PACK,
sistance and flow rate). These parameters can be partially valves, mixing manifold and cabin. Each of the sub-
known or unknown and need to be estimated and con- components was taken as an individual module and were
verged to eliminate noise from the residual outcome. Yang designed based on one dimensional thermodynamic
X. et al.,71 suggested further research on using a hybrid equations and control system.74
model combining both model-based and data-driven A simulation model of the ECS system was built within
methods which are predicted to be able to isolate fault FLECS consisting of ducts and cabin model. After per-
under noise conditions. forming a comparison study with other existing simulation
The variation of the physical process constraints such data, the model was verified based on temperature data
as the heat exchanger effectiveness, resistance and flow collected at the supply duct inlet, PACK outlet and cabin
rate, giving a converged residual value from model data inlet, during a test flight of Airbus A340-600. The ac-
and actual system data can indicate the faulty compo- quired temperature data from the cabin inlet demonstrated
nent.70 However, this method is difficult to implement on good match.74 The verification did not include pressure
dynamic models in which the fault propagates with time. and mass flow data.

Finding 1. The FDI capability can be fundamentally ob- Flowmaster

tained through either model-based or data-driven or both
techniques combined. It is noticed that for the aircraft ECS Flowmaster software is commonly used in the aerospace
system, the model-based approach is more popular than industry for modelling. It is a 1-D thermo-fluid systems
the data-driven approach. This is due to the scarcity of simulation that linearises model coefficients to achieve
data available from the aircraft. The classification of the good convergence properties. Tu, Y. and Lin, G. P.75
FDI methods and their advantages/disadvantages which developed an ECS model using FlowmasterV7, which
2460 Proc IMechE Part G: J Aerospace Engineering 237(11)

Figure 7. Classification of Diagnostic methodologies.

is programmed in C++ language, and the sub-block pa- Square Twin Support Vector Machine (LSTSVM) method
rameter method was used for mathematical modelling of was used for leakage detection.
the components. The work focussed mostly on developing
the cabin temperature control system using expert
Proportional-Integral-Derivative (PID) controller, which
Easy5
uses the fuzzy algorithm to adjust the PID parameters.1 Back in 1976, The Boeing Company developed Easy4
The effect of the specific humidity of the flow to model the which later became Easy5. This software was developed
heat exchanger (wet and dry) and ACM was considered. for the need of having one single simulation modelling
Both steady-state and dynamic analysis was validated platform that can be used to create both nonlinear and
experimentally with a maximum and minimum deviation linearised analysis of dynamic systems. The library was
of 4.3°C and 0.2°C.75 It is to be noted that the verification developed based on the SysteMMS a modular modelling
did not involve pressure and mass flow readings. system. Easy5 also supports additional components cre-
Although the simulation models were primarily for ated using Fortran. It is only required for the user to or-
design and development purposes rather than fault de- ganise the components to be placed according to the
tection and isolation, Flowmaster simulation software has desired model and the interconnection between the
been used for fault detection for another system. Lang components are formed by the built-in Model Generation
et al.76 researched to detect pipe leakage based on the Program.15 This includes defining the physical driving
feature extracted from the pressure signal. Wavelet quantities such as forces, flow rates, velocities and other
analysis was conducted for noise removal, and Least variables depending on the input parameters defined by
Chowdhury et al. 2461

the user. It also includes an Analysis Program which both dry and wet simulation analysis. In comparison to
permits dynamic, static, linear and nonlinear analysis of Boeing’s previous simulation model Easy5, SESAC can
the model after all the required input parameters are set. be used for fault simulation analysis. This model has been
This software can run frequency response analysis, steady- validated against Easy5 data provided by Boeing.85
state analysis and optimal controller synthesis.77 This ECS simulation model can be used for di-
Easy5 was later used by Boeing for developing an agnostic purposes using model-based method by
aircraft ECS model for simulation analysis. Following comparing model data and test data. The fault simu-
that, Hoffman15,78 reported the use of Easy5 for simulation lation analysis will provide knowledge of PACK per-
of F-14F fighter aircraft ECS. In another study, Gulfstream formance under faulty condition which will
Aerospace used this software for developing the ECS subsequently support identification of optimised sensor
model for G500 and G550 passenger jets.15,79 SAAB location for fault detection. Diego,86 has used the
Group collaborated with Linköping Institute of Tech- SESAC and Thermodynamic model to propose five
nology to model ECS for their fighter jet JAS39 Gripen. sensor location shown in Table 1 for identifying fault
Fault modes identified from the historical data focusing modes tabulated in Table 2.
mostly on the valve jamming was used to develop a state-
observer model-based diagnostic method.15,80
Finding 3. The list of models identified above is sum-
marised in Figure 8. With the exception of SESAC, the
Dymola simulation models have been primarily used for design
and analysis purposes. Also, in order to be used for di-
SAAB Group later migrated to Dymola as the ECS re-
agnostic purposes the models require experimental veri-
quired major modifications. S-ECS system was modelled
fication and validation at a system level.87 It is noted that
using the Modelica modelling language, which is then
the verification and validation of the models have not been
simulated in the simulation platform Dymola.81 As sug-
done comprehensively.
gested by Steinkellner82 in a study, the model requires
experimental verification and validation.
Finding 4. The emphasis on the impact of control system
failure (including valve malfunction) on PACK per-
MATLAB/Simulink formance at a system level is missing in the open
literature.
For more than a decade, simulation modelling was based
on arithmetic operations, integrators and transfer func- Table 1. Optimal sensor set.86
tions.[p.7]83 After Easy5 simulation platform, Mathworks
developed the simulation environment called Simulink Variable Sensor name
around 1991, which was a major shift to the paradigm of
Temperature Temperature at the primary heat exchanger
using ordinary differential equation (ODE).
outlet
Romani and De Goes programmed a cabin temperature
Temperature Temperature at the secondary heat exchanger
control model using ordinary differential equations
outlet
(ODEs). The model has been validated by cabin tem-
Temperature Temperature at the turbine outlet
perature recorded under three different conditions: aircraft
Pressure Pressure at the temperature control valve marge
cooling in flight, aircraft heating in flight and pull-down
outlet
scenarios.84 The verification did not include Pressure and
Pressure Pressure at the turbine inlet
Mass flow readings. Although their results matched well
with the data, their study lacks the modelling of the dy-
namic behaviour of the valves that have the potential to
mask fault symptoms.15 Furthermore, humidity has not
been considered for simulation modelling. Table 2. List of faults in the ECS PACK provided by the
industrial partner.86

Simscape ECS simulation for all conditions Name Failure mode

Cranfield University IVHM Centre on a funded project by PACK valve Incorrect valve position
the Boeing Company developed a computationally effi- Temperature control valve Incorrect valve position
cient and robust simulation framework for the ECS Primary heat exchanger Fouling or blockage
called – Simscape ECS Simulation for All Condition Secondary heat exchanger Fouling or blockage
(SESAC).15 The components were constructed with Reheater Fouling
simple algebraic equations based on their thermodynamic Condenser Blockage cold side
properties. SESAC is a component library which enables
Water separator Malfunction
drag-and-drop component list that can be used to construct
Air cycle machine Compressor or turbine
ECS models. This model can be used for analysing PACK
malfunction
performance characteristics under different operating
Ram air inlet deflector door Malfunction
conditions (i.e. in flight and on ground). It also supports
2462 Proc IMechE Part G: J Aerospace Engineering 237(11)

Figure 8. Existing ECS simulation models.

Experimental work on the

ECS diagnostics
The simulation models used for developing diagnostic
rules need to be experimentally verified and validated. In
addition, the sensors used for condition monitoring also
require appropriate calibration and validation. A faulty
model and a misplaced sensor can produce a wrong di-
agnostic result, which can lead towards wrong mainte-
nance action resulting in accidents and increased
maintenance costs.
Honeywell-SECAN conducted an experimental in-
vestigation to validate the performance characteristics of
a new LPWS design.88 In contrast to the classical single
stage water separator, a new two stage system with Figure 9. Rig set-up for LPWS Coalescer testing.88
a coalescer followed by water separating device was
designed and tested. For experimental testing, a rig was
developed with a compressor air inlet, water injection input and output of each component. The data collected
from the rig was used to validate the 1D thermodynamic
nozzle, turbine, diffuser and the coalescer under test. The
rig set-up is shown in Figure 9. During the experimen- model by comparing the ECS coefficient of performance
tation, the air flow rate was set, water injected to vary the and system heat rejection. To test the diagnostic capa-
humidity value and drop size distribution and pressure bilities, Ram air inlet (to the SHX) was blocked, and the
drop at coalescer outlet was measured. While this was to data collected was analysed. It was highlighted that a re-
test the coalescer, for testing the separating device, an duced Ram mass flow predominantly reduces the co-
additional aspiration system was added to the rig to collect efficient of performance of the ECS.
the separated water. The amount of water recovered at the This rig was used to simulate faults and to conduct
WS outlet, in the turbine and in the air stream was diagnoses at a system-level. The first failure mode analysis
measured. Data collected were used to calculate the was on the blockage in the bleed air duct. It was identified
overall efficiency of the WS for design validation.88 from the research that the blockage does not drastically
Childs et al.89 developed a rig for testing BAE Systems affect its performance until cycle mass flow falls below
fast-jet Hawk military aircraft ECS at Loughborough 75% of its unblocked state. The TCV position testing
University. The two-wheel bootstrap cycle with LPWS showed that TCVs would comfortably mask any failure
ECS was set up in a lab environment. The rig was in- within the system.90 Similarly, the rig was used to analyse
strumented with temperature and pressure sensor, at the the effect of humidity on system performance.90 The
Chowdhury et al. 2463

This paper has identified different diagnostic methodol-

ogies that are applicable to the ECS system. A taxonomy
of the diagnostic methodologies, including their advan-
tages and disadvantages, are demonstrated Figure 7.
In an integrated and complex system like the ECS, fault
propagates through the system and diagnosing faults
correctly requires knowledge of component inter-
dependencies. To study the interdependencies between
components within a system, a system-level simulation
model is required. Furthermore, the control system gov-
erning the valves needs to be well designed as they can
lead to misleading PACK performance simulation anal-
ysis. The emphasis on control system is missing in the
Figure 10. Experimental set-up on the B737-200.92 open literature.
Finally, it was found that the models available for
performing ECS simulation were used primarily for de-
results indicated that humid air impacts the ECS perfor- sign and analysis purposes. These models have been
mance characteristics particularly at the ACM.91 verified with very limited data and have been mostly
In collaboration with Boeing, Esperon-Miguez et al.92 dependant on only temperature measurements. Therefore,
developed a thermodynamic model for the ECS capable of it is deduced that there is a compelling need for a full-scale
simulating faults in heat exchangers, valves and WS. The experimental facility to perform ECS analysis under
model has been validated using test data collected from different operating conditions and to use temperature,
a ground-based B737-200 aircraft which has a LPWS in pressure and mass flow data to verify the models.
the ECS. The ECS PACK was instrumented with ther-
mocouples. The valve angles were measured using po-
tentiometers. The rig set-up is shown in Figure 10. Two Appendix
different types of experiments were conducted: healthy
conditions and fault injected cases. Heat exchanger inlets
were blocked using aluminium plates to reduce the ef- Acronyms
fectiveness of the heat exchangers. Valves were regulated ACM Air cycle machine
with potentiometer to fully open and fully closed sce- AHU Air handling unit
narios. The coalescer bag was clogged without exceeding AI Artificial intelligence
pressure differential. Using the healthy and fault injected AIS Anti ice system
temperature data collected the thermodynamic model was ANN Artificial neural network
verified.92 BAS Bleed air system
CHX Condenser
Finding 5 CPCS Cabin pressure control system
DMF Double model filter
There is a need to develop a civil aircraft in situ ECS GTF DNN Deep neural network
targeting temperature, pressure, mass flow and humidity ECS Environmental control system
data collection. FCD Fractal correlation method
FDI Fault detection and isolation
FLECS Functional model library of the ECS
Conclusion
HPWS High-pressure water separator
The aircraft ECS has been reported as one of the major HVAC Heating, ventilation and air
drivers of unscheduled maintenance by the operators. ECS iBIT Built-in-test
failure, resulting in sudden cabin depressurisation, has IVHM Integrated vehicle health management
caused a number of aircraft to make emergency landings. k-NN k-Nearest Neighbor
The existing literature suggests that there is a compelling LPWS Low-pressure water separator
need to advance the level of scientific understanding of the MRO Maintenance repair and overhaul
ECS system operating under healthy and degraded con- MSA Multivariate statistical analysis
ditions. An area of interest for the manufacturers is to ODE Ordinary differential equation
investigate diagnostic approaches for detecting and iso- PACK Passenger air conditioner
lating faults within the ECS system. PCA Principal component analysis
It has been identified that heat exchanger degradation PHX Primary heat exchanger
has been extensively researched based on the im- PID Proportional integral derivative
plementation of model-based diagnostic techniques. There PM Predictive maintenance
is very limited research available on the air cycle machine PSS Product service system
and high pressure water separator component degradation. PV Pack valve
2464 Proc IMechE Part G: J Aerospace Engineering 237(11)

RHX Reheater 9. Lightfoot H and Greenough RM. State-of-the-art in in-

RUL Remaining useful life tegrated vehicle health management. Proc Inst Mech Eng G
SESAC Simscape ECS simulation under all J Aerosp Eng 2016; 223: 157–170. DOI: 10.1243/
conditions 09544100JAERO446 (March): 1–14.
10. Grubic T, Jennions I and Baines T. The Interaction of PSS
SHS Stochastic hybrid system
and PHM - a mutual benefit case. United States: Annual
SHX Secondary heat exchanger
Conference of the PHM Society, 2009, pp. 1–10.
STF Strong tracking filter
11. Ezhilarasu CM, Skaf Z and Jennions IK. Progress in aerospace
SVM Support vector machine sciences the application of reasoning to aerospace integrated
TCV Temperature control valve vehicle health management (IVHM): challenges and oppor-
WS Water separator tunities. Prog Aerospace Sci 2019; 105: 60–73. DOI: 10.1016/j.
paerosci.2019.01.001 Elsevier Ltd December2018.
Acknowledgement 12. Mobley RK. An introduction to predictive maintenance.
The Boeing Company, as part of their collaboration with United States: Elsevier, 2002, p. 4.
Cranfield University’s IVHM Centre, funded this work; the 13. Pomata AE. Boeing 307 stratoliner pressurized airliner.
authors would like to thank them for their support of this project. Available at: https://www.historylink.org/File/3598 (2001,
accessed 14 September 2020).
Declaration of conflicting interests 14. Cooper JB. Altitude conditioning of aircraft cabins. SAE Int
1941; 36/49: 240–248 0148-7191.
The author(s) declared no potential conflicts of interest with respect
15. Jennions I, Ali F, Miguez ME, et al. Simulation of an aircraft
to the research, authorship, and/or publication of this article.
environmental control system. Appl Therm Eng 2020 May;
172: 114925. DOI: 10.1016/j.applthermaleng.2020.114925
Funding 16. Vega Diaz R. Analysis of an electric environmental control
The author(s) disclosed receipt of the following financial support system to reduce the energy consumption of fxed-wing and
for the research, authorship, and/or publication of this article: rotary-wing aircraft. Cranfield: Cranfield University, 2011.
This work was supported by the Boeing; IVHH-Cranfield. Available at: https://dspace.lib.cranfield.ac.uk/bitstream/
handle/1826/7422/Rolando_Vega_Diaz_Thesis_2011.pdf?
sequence=1&isAllowed=y
ORCID iDs
17. Devadurgam H, Rajagopal S and Munjulury RC. Analytical
Shafayat H Chowdhury  https://orcid.org/0000-0002-1757-3599 desgin and estimation of conventional and elctrical aircraft
Fakhre Ali  https://orcid.org/0000-0002-9822-2482 environmental control system. CADLab 2019; 1–14, DOI:
10.31224/osf.io/gu4pz
References 18. Genty A, Saunders A, Cimino JP, et al. Environmental
1. Tu Y and Lin GP. Dynamic simulation of aircraft envi- control system for light turbo prop. United States: SAE
ronmental control system based on flowmaster. J Aircraft International, 2003, pp. 1–9.
2011; 48(6): 2031–2041. DOI: 10.2514/1.C031433 19. Shang Y, Wang J, Bi Y, et al. Air dehumidification by
2. Skybrary. B737 en-route. Glen Inne, NS, Australia: 2017. membrane with cold water for manned spacecraft envi-
Available at: https://www.skybrary.aero/index.php/B737_en- ronmental control. MATEC Web of Conf 2017; 114: 1–7.
route (accessed 26 June 2019). DOI: 10.1051/matecconf/201711402011
3. Skybrary. B735, en-route, SE. Kushimoto Wakayama, Ja- 20. Jonsson G and Palsson OP. An application of extended
pan: 2006. Available at: https://www.skybrary.aero/index. kalman filtering to heat exchanger models. J Dynamic Syst
php/B735,_en-route (accessed 26 June 2019). Meas Control Trans ASME 1994; 116(2): 257–264. DOI: 10.
4. Skybrary. B738. Glasgow UK: 2012. Available at: https:// 1115/1.2899218
www.skybrary.aero/index.php/B738 (accessed: 26 June 2019). 21. Najjar N, Gupta S, Hare J, et al. Optimal sensor selection and
5. Aaseng GB. Blueprint for an integrated vehicle health fusion for heat exchanger fouling diagnosis in aerospace
management system. In: DASC. 20th digital avionics sys- systems. IEEE Sensors J 2016; 16(12): 4866–4881. DOI:
tems conference (Cat. No.01CH37219), Daytona Beach, FL, 10.1109/JSEN.2016.2549860
USA, 14–18 October 2001, 3C1/1–3C1/11. DOI: 10.1109/ 22. Palmer KA, Hale WT, Such KD, et al. Optimal design of
DASC.2001.963350 IEEE. tests for heat exchanger fouling identification. Appl Therm
6. NASA. Research And technology goals and objectives for Eng 2016; 95: 382–393. DOI: 10.1016/j.applthermaleng.
integrated vehicle health management (IVHM). Washingtin, 2015.11.043 Elsevier Ltd.
DC: NASA, 2019. Report No: NASA-CR-192656. Avail- 23. Pang L, Qu H and Dong S. Fault diagnosis algorithm for heat
able at: https://ntrs.nasa.gov/citations/19930013844 exchanger of aircraft environment control system based on
7. Science BA. Integrated vehicle health management tech- double model filter. Hangkong Xuebao/Acta Aeronaut As-
nology and its applications in commerical aviation. In: 2017 tronaut Sin 2008; 29(3): 548–553.
International conference on sensing, diagnostics, prog- 24. Shang L and Liu G. Aircraft heat exchanger fouling de-
nostics, and control, Shanghai, China, 16–18 August 2017. tection based on valve control command. J Aerospace. 2012;
DOI: 10.1109/SDPC.2017.145 226(7): 740–750. DOI: 10.1177/0954410011416058
8. Zhang G, Wang J, Lv Z, et al. A integrated vehicle health 25. Wright S, Andrews G and Sabir H. A review of heat ex-
management framework for aircraft. In: IEEE conference changer fouling in the context of aircraft air-conditioning
on prognostics and health management, Austin, TX, systems, and the potential for electrostatic filtering. Appl
USA, 22–25 June 2015, 1–8. DOI: 10.1109/ICPHM. Therm Eng 2009; 29(13): 2596–2609. DOI: 10.1016/j.
2015.7245034 applthermaleng.2009.01.002
Chowdhury et al. 2465

26. Cubillo A. Physics-based approach to detect metal-metal units. Energy and Build 2014; 82: 550–562. DOI: 10.
contact in the hydrodynamic bearing of a planetary 1016/j.enbuild.2014.06.042
transmission. Cranfield: Cranfield University, 2016. 43. Mostofi M, Shamekhi AH and Ziabasharhagh M. De-
27. Eker Ö. A hybrid prognostic methodology and its appli- veloping an algorithm for si engine diagnosis using parity
cation to wellcontrolled engineering systems. Cranfield: relations. Chicago, Illinois, USA: American Society of
Cranfield University, 2015. Mechanical EngineersDynamic Systems and Control Di-
28. Byington CS, Watson M, Edwards D, et al. A model-based vision (Publication) DSC, 2006, March 2014, DOI: 10.1115/
approach to prognostics and health management for flight IMECE2006-13463
control actuators. In: 2004 IEEE Aerospace Conference 44. Varrier S, Koenig D and Martinez JJ. A parity space-based
Proceedings, Big Sky, MT, USA, 06–13 March 2004, fault detection on LPV systems: approach for vehicle lateral
3551–3562. DOI: 10.1109/AERO.2004.1368172 6 dynamics control system. IFAC Proc Volumes (IFAC-Pa-
29. Jennions I and Ali F. Assessment of heat exchanger deg- persOnline) 2012; 45: 1191–1196. DOI: 10.3182/
radation in a boeing 737-800 environmental control system. 20120829-3-MX-2028.00053
J Therm Sci Eng Appl 2021 1 December; 13(6): 061015. 45. Hwang W and Huh K. Fault detection and estimation for
DOI: 10.1115/1.4050324 electromechanical brake systems using parity space ap-
30. Butler S. Prognostic algorithms for condition monitoring proach. J Dynamic Syst Meas Control 2015; 137(1): 014504.
and remaining useful life estimation. National University of DOI: 10.1115/1.4028184
Ireland, Maynooth: 2012. Available at: https://core.kmi. 46. Laamami S, Ben HM and Sbita L. Fault detection for
open.ac.uk/download/pdf/11527019.pdf nonlinear systems: parity space approach. In: 2015 World
31. Isermann R. Model-based fault-detection and diagnosis - congress on Information technology and computer appli-
status and applications. Annu Rev Control 2005; 29(1): cations congress (WCITCA), Hammamet, Tunisia, 11–13
71–85. DOI: 10.1016/j.arcontrol.2004.12.002 June 2015, 1–5. DOI: 10.1109/WCITCA.2015.7367055 (5).
32. Jonsson GR, Lalot S, Palsson OP, et al. Use of extended 47. Zhirabok AN. Parity space approach to fault diagnosis in
kalman filtering in detecting fouling in heat exchangers. nonlinear systems. In: European control conference,
Interantional J Heat Mass Transfer 2007; 50: 2643–2655. ECC 1999 - conference proceedings, Karlsruhe; Ger-
DOI: 10.1016/j.ijheatmasstransfer.2006.11.025 many, 31 August–3 September 1999, 232–235. Available
33. Shah S, Liu G and Greatrix DR. On-line fouling detection at: https://www.scopus.com/inward/record.url?eid=2-s2.
of aircraft environmental control system cross flow heat 0-84930599543&partnerID=tZOtx3y1
exchanger. In:IEEE international conference on mecha- 48. El-Mezyani T, Düştegör D, Srivastava SK, et al. Parity space
tronics and automation, Changchun, China, 09–12 August approach for enhanced fault detection and intelligent sensor
2009, 2940–2945. DOI: 10.1109/ICMA.2009.5246062 network design in power systems. power and energy. In:
ICMA. IEEE PES general meeting, Minneapolis, MN, USA, 25–29
34. Ma J, Lu C and Liu H. Fault diagnosis for the heat exchanger July 2010, 1–8. Available at: https://ieeexplore.ieee.org/
of the aircraft environmental control system based on the xpls/abs_all.jsp?arnumber=5590102
strong tracking filter. PLoS One 2015; 10(3): 1–11. DOI: 10. 49. Yin S and Huang Z. Performance monitoring for vehicle
1371/journal.pone.0122829 suspension system via fuzzy positivistic C-means clustering
35. Mutambara AGO and Al-Haik MSY. EKF based parameter based on accelerometer measurements. IEEE/ASME Trans
estimation for a heat exchanger. In: Proceedings of the Mechatronics. IEEE 2015; 20(5): 2613–2620. DOI: 10.
American control conference, San Diego, CA, USA, 02–04 1109/TMECH.2014.2358674
June 1999, 3918–3922. DOI: 10.1109/acc.1999.786251 50. Yin S, Li X, Gao H, et al. Data-based techniques focused on
6(June). modern industry: an overview. IEEE Trans Ind Electronics.
36. Hu G, Gao S, Zhong Y, et al. Modified strong tracking IEEE 2015; 62(1): 657–667. DOI: 10.1109/TIE.2014.
unscented kalman filter for nonlinear state estimation with 2308133
process model uncertainty. Int J Adaptive Control Signal 51. Najjar N, Hare J, D’Orlando P, et al. Heat exchanger fouling
Process 2015; 29(12): 1561–1577. DOI: 10.1002/acs.2572 diagnosis for an aircraft air-conditioning system. SAE Int
37. Ding SX. Model-based fault diagnosis techniques. London, 2013. DOI: 10.4271/2013-01-2250
UK: Springer, 2013. DOI: 10.1007/978-1-4471-4799-2 52. Littooy B, Loire S, Georgescu M, et al. Pattern recognition
Second and classification of HVAC rule-based faults in commercial
38. Wang S and Wang JB. Robust sensor fault diagnosis and buildings. In: 2016 IEEE international conference on big
validation in HVAC systems. Trans Inst Meas Control 2002; data (big data), Washington, DC, USA, 05–08 December
24(3): 231–262. DOI: 10.1191/0142331202tm030oa 2016, 1412–1421. DOI: 10.1109/BigData.2016.7840749
39. Rajaraman S, Hahn J and Mannan MS. A methodology for 53. Skliros C, Ali F, King S, et al. Aircraft system-level di-
fault detection, isolation, and identification for nonlinear agnosis with emphasis on maintenance decisions. Proc Inst
processes with parametric uncertainties. Ind Eng Chem Res Mech Eng O: J Risk Reliability 2021 26 October; 236:
2004; 43(21): 6774–6786. DOI: 10.1021/ie0400806 1057–1077. DOI: 10.1177/1748006X211054938
40. Daigle MJ and Goebel K. A model-based prognostics ap- 54. Leondes CT. Expert systems: the technology of knowledge
proach applied to pneumatic valves. Int J Prognostics Health management and decision making for the 21st century.
Manage 2011; 2: 1–16. Academic Press, 2002, pp. 1–22.
41. Daigle M and Goebel K. Model-based prognostics under 55. Bruton K, Raftery P and Kennedy B. Review of automated
limited sensing. In: 2010 IEEE aerospace conference, Big fault detection and diagnostic tools in air handling units.
Sky, MT, USA, 06–13 March 2010. DOI: 10.1109/AERO. Energy Eff 2014; 7: 335–351. DOI: 10.1007/s12053-013-
2010.5446822 9238-2
42. Yu Y, Woradechjumroen D and Yu D. A review of fault 56. Zadeh LA. Fuzzy sets. Inf Control 1965 June; 8(3):
detection and diagnosis methodologies on air-handling 338–353. DOI: 10.1016/S0019-9958(65)90241-X
2466 Proc IMechE Part G: J Aerospace Engineering 237(11)

57. Adhikari S, Sinha N and Dorendrajit T. Fuzzy logic based 72. Santos APP, Andrade CR and Zaparoli EL. A thermody-
on-line fault detection and classification in transmission line. namic study of air cycle machine for aeronautical applica-
SpringerPlus 2016; 5: 1002. DOI: 10.1186/s40064-016- tions. Int J Thermodynamics 2014; 17(3): 117–126. DOI: 10.
2669-4 5541/ijot.538
58. Liu XF and Dexter A. Fault-tolerant supervisory control of 73. Scholz D, Giese T and Flecs EC. Functional library of the
VAV air-conditioning systems. Energy Build 2001; 33(4): environmental control system – a simulation tool for the
379–389. DOI: 10.1016/S0378-7788(00)00120-1 support of industrial processes. Hamburg, Germany: AST,
59. Najafi M, Auslander DM, Bartlett PL, et al. Application of 2007.
machine learning in the fault diagnostics of air handling 74. Müller C and Scholz Tg D. Dynamic simulation of in-
units. Appl Energ 2012; 96: 347–358. DOI: 10.1016/j. novative aircraft air conditioning. 1st CEAS European Air
apenergy.2012.02.049 Elsevier Ltd. and Space Conference. German Society for Aeronautics and
60. Teodosiu C, Pastravanu O and Macoveanu M. Neural net- Astronautics, 2007, pp. 869–878.
work models for ultrafiltration and backwashing. Water Res 75. Tu Y and Lin G. Dynamic simulation of humid air envi-
2000; 34(18): 4371–4380. DOI: 10.1016/S0043-1354(00) ronmental control system. In: 40th International Conference
00217-7 on Environmental Systems, Barcelona, Spain, 11 July 2010–
61. Riverol C and Napolitano V. Estimation of the overall heat 15 July 2010, 1–10. DOI: 10.2514/6.2010-6305
transfer coefficient in a tubular heat exchanger under fouling 76. Lang X, Li P, Hu Z, et al. Leak detection and location of
using neural networks. application in a flash pasteurizer. Int pipelines based on LMD and Least squares Twin support
Commun Heat Mass Transfer 2002; 29(4): 453–457. DOI: vector machine. United States: IEEE Access, 2017, 5, pp.
10.1016/S0735-1933(02)00342-1 8659–8668.
62. Riverol C and Napolitano V. Estimation of fouling in a plate 77. Burroughs JD and Hammond RA. Control analysis and
heat exchanger through the application on neural networks. design features of EASY5. San Francisco: American Control
J Chem Tech Biotechnol 2005; 80(5): 594–600. DOI: 10. Conference, 1983, 58–63. DOI: 10.23919/ACC.1983.
1002/jctb.1198 4788072
63. Hare J, Gupta S, Najjar N, et al. System-level fault diagnosis 78. Hoffman G. Environmental control system simulation using
with application to the environmental control system of an EASY5, as applied to the F-14. Intersociety Conf Environ
aircraft. In: SAE 2015 aerotech congress & exhibition, Syst 1985; 8.
Seattle Washington, USA, 22–25 Sep 2015. DOI: 10.4271/ 79. Schooley C. A time-response analysis of an air distribution
2015-01-2583 Copyright. subsystem of an aircraft environmental control system using
64. Allen WH, Member S, Rubaai A, et al. Fuzzy neural MSC easy5 TM. MSC.Software VPD conference, 2006.
network-based health monitoring for HVAC system 80. Karlsson J. Diagnosis of the air distribution system of the
variable-air-volume unit. IEEE transactions on industry JAS39 Gripen environmental control system. Linköping,
applications. IEEE 2016; 52(3): 2513–2524. DOI: 10.1109/ Sweden: Linköping Institute of Technology, 2001.
TIA.2015.2511160 81. Steinkellner S, Andersson H, Gavel H, et al. Modeling and
65. Du Z, Fan B, Jin X, et al. Fault detection and diagnosis for simulation of saab gripen’s vehicle systems. AIAA modeling
buildings and HVAC systems using combined neural net- and simulation technologies conference. Reston, Virigina:
works and subtractive clustering analysis. Building Environ American Institute of Aeronautics and Astronautics, 2009,
2014; 73: 1–11. DOI: 10.1016/j.buildenv.2013.11.021 1–15. DOI: 10.2514/6.2009-6134.
Elsevier Ltd. 82. Steinkellner S. Aircraft vehicle systems modeling and sim-
66. Lee K-P, Wu B-H and Peng SL. Deep-learning-based fault ulation under uncertainty. Linköping, Sweden: Linköping
detection and diagnosis of air-handling units. Build Environ Universuty, 2011.
2019; 157: 24–33. DOI: 10.1016/j.buildenv.2019.04.029 83. Sielemann M. Device-oriented modeling and simulation in
MU in MRRR fault detection and diagnosis of air- aircraft energy systems design. Harburg, Germany: Tech-
handling units _ EER pdfh.-L. nischen Universit€at Hamburg, 2012.
67. Sreedharan P and Haves P. Comparison of chiller models for 84. Romani R and De Goes LC. Cabin temperature control
use in model-based fault detection priya sreedharan and phil model for commercial aircraft. Model Simulation Tech Conf
haves. International Conference for Enhanced Building 2012; 1–15. DOI: 10.2514/6.2012-4637
Operations: Austin, TX, 2001. (July). 85. Escobar IC. Condition monitoring for environmental control
68. Yu B, Van Paassen DHC and Riahy S. General modeling for systems. Cranfield, United Kingdom: Cranfield University,
model-based FDD on building HVAC system. Simulation 2015.
Pract Theor 2002; 9(6–8): 387–397. DOI: 10.1016/S1569- 86. Figuero DV. Diagnostic capability for environmental control
190X(02)00062-X systems. Cranfield, United Kingdom: Universidad Po-
69. Buswell RA, Norford LK, Klaassen CJ, et al. Demonstration litécnica de Madrid, Cranfield Univeristy, 2017.
of fault detection and diagnosis methods for air-handling 87. Scott M. Development of a realistic, automated, & amp;
units. HVAC&R Res 2011; 8(1): 41–71. DOI: 10.1080/ efficient reliability development growth test for a large,
10789669.2002.10391289 distributed, environmental control system. United States:
70. Isermann R. Process fault detection based on modeling and SAE Technical Paper Series, 1999.
estimation methods-a survey. Automatica 1984; 20(4): 88. Fourquet H, Taillepied D, Novo N, et al. Performance
387–404. DOI: 10.1016/0005-1098(84)90098-0 validation of a new water separator concept. 30th in in-
71. Yang XB, Jin XQ, Du ZM, et al. A novel model-based fault ternational conferfence on environmental systems. United
detection method for temperature sensor using fractal cor- States: SAE Technical Paper, 2000.
relation dimension. Building Environ 2011; 46(4): 970–979. 89. Childs T, Jones AB and Chen R. Development of a full scale
DOI: 10.1016/j.buildenv.2010.10.030 Elsevier Ltd. experimental and simulation tool for environmental control
Chowdhury et al. 2467

system optimisation and fault detection 2015; 2015-1196: 91. Childs T, Jones A, Chen R, et al. A study into refrigeration
0–11. DOI: 10.2514/6.2015-1196 cycle working fluids using an air cycle machine environ-
90. Childs TG, Jones AB, Chen R, et al. Evaluating envi- mental control system. 54th AIAA Aerospace Sci Meet 2016;
ronmental control system thermal response to degraded 0: 0–14. DOI: 10.2514/6.2016-2190
operating conditions. Evaluating environmental control 92. Esperon-Miguez M, Jennions IK, Escobar IC, et al. Simulating
system thermal response to degraded operating con- faults in a boeing 737-200 environmental control system using
ditions 2016; AIAA 2016-1151: 0–16. DOI: 10.2514/6. a thermodynamic model. Int J Prognostics Health Manage 2019
2016-1151 1 June; 10(2): 1–12. DOI: 10.36001/ijphm.2019.v10i2.2731