UNCERTAINTIES AND ERRORS OF QRA

Laura Bruce, Ifiok Etukudo, Widya Siswanto, Ugoeze Emmanuella Uzor and Shobha Venkataswamy.

WHAT IS QRA?
QRA(Quantitative Risk Assessment) is a study for determination of expected hazards which could arise from operations in process, and their consequences/environmental impacts. The expected hazards may be fires, explosion, leak of flammable/toxic gases etc.

TYPES OF UNCERTAINTIES IN QRA

Parameter Uncertainty: When the values of the parameters used in the models are not accurately known. It might be stochastic or epistemic uncertainty. Model Uncertainty: Any model, conceptual or mathematical, will inevitably be a simplification of the reality it is designed to represent Completeness Uncertainty: Originates from the fact that not all contributions to risk are addressed in QRA models.

UNCERTAINTIES INTRODUCED AT THE DIFFERENT STAGES OF QRA

THE IDENTIFICATION STAGE

Produce a comprehensive list of possible initiating events To identify priorities between them and make decisions on which of them are to be analysed further

Major uncertainty: Completeness. Have all major hazards and/ or possible accident scenarios been identified? Methods for structured identification (to facilitate completeness): HAZOP, what-if analysis, FMEA, etc.

THE IDENTIFICATION STAGE (CONTD)

Example of accidents with their coverage in standard QRA methodologies.
Accident Mechanism Included in Standards QRA Methodologies ? N Models Available?

Texas City 2005 Tosco Avon 1997 Buncefield 2005 CAI, Danvers 2006

KO drum overflow, spraying vent Hydrocracker reactor runaway

Overflow, N splashing flow, UVCE Confined VCE N

Y (Adapted from Taylor, 2010)

FREQUENCY ESTIMATION

FREQUENCY ESTIMATION (CONTD)

The two main methods of likelihood and frequency estimation are: Historical Data: Questions raised on accuracy and applicability. Data may be inaccurate, incomplete or inappropriate. Issue on parameter uncertainty. E.g.: Leak frequency data for North Sea to elsewhere.

Fault and event tree analysis: Questions raised on completeness, simplification and parameter uncertainty in the model. E.g.: Omission of significant failure mechanisms can lead to erroneous results.

FREQUENCY ESTIMATION (CONT:)

Major sources of uncertainty in frequency estimation: - failure frequency; - leak frequency; - ignition probability; and - explosion probability.

CONSEQUENCE ESTIMATION

The consequence estimation scheme involves three steps: y Accident scenario analysis
  

Identification of initiating events. Generation of accident scenarios for each initiating events. Quantification of accident scenarios.

Identification and classification of losses y Estimation of losses

y

CONSEQUENCE ESTIMATION (CONTD)

Conducted using physical and effect models. y Physical models: flash fire, pool fire, BLEVEs, etc. y Effect models: explosion effects, heat radiation, toxic effects, etc. Major uncertainty: y Model uncertainty in physical and effect models y Parameter uncertainty, such as leak size.

CONSEQUENCE ESTIMATION (CONTD)

Scenario: Piping connecting Separator A with Gas Compressor experience 30 mm leak (9 in piping with 110 m length). Operating pressure is 30 bar and operating temperature is 27oC. Model uncertainty and parameter uncertainty was identified in the consequence estimation.
Leak Diameter (mm) 30 25 30 Difference Full Analysis Spouge (%) Leak Flame Leak Flame in flame (kg/s) length (m) (kg/s) length (m) length 2.87 28.51 2.7 27.8 2.50% 1.99 24.55 1.875 23.94 2.50% 2.51 26.99 2.7 27.8 2.99%

Cd 0.8 0.8 0.7

CONSEQUENCE ESTIMATION (CONTD)

With the same scenario, if SDV is used as safeguard, the consequence estimation presented below.
Time (minutes) 0 5 10 Leak (kg/s) Flame length (m) 2.87 28.51 3.25E-04 0.7 3.60E-08 0.02

Uncertainty relates to the assumption made on the condition of the safeguard, such as full closure (airtight) condition of the valve, time response, etc.

ESTIMATION OF RISK
The final step in the QRA is to generate the actual risk measure. Risk is product of the probability of a certain outcome with the consequence of that particular outcome. Risk indices represented by:

Individual Risk Per Annum (IRPA); y Probable Loss of life (PLL); y Temporary Refuge Impairment Frequency (TRIF); and
y

y Societal Risk (F-N Curve).

ESTIMATION OF RISK (CONTD)

Major uncertainty:
y

Assumption and simplification made in order to decrease the complexity of the analysis Example: assumptions on distribution of ignition sources, population distributions, etc. Relative importance of risk

CONCLUSION
Identification of the contributors to the overall uncertainty is important to improve the quality of the QRA Whenever necessary, increased investment in data collection or model development could significantly reduce uncertainty. However, uncertainty in inherent statistical variability in failure rate data can not be removed.

REFERENCES
Abrahamsson, Marcus. Uncertainty in Quantitative Risk Analysis Characterization and Methods of Treatment. 2002. Lees, Frank P. Loss Prevention in the Process Industries, 3rd Edition. 2004. Taylor, J.R. Accuracy in Quantitative Risk Assessment? 13th International Symposium on Loss Prevention. 2010. Borysiewicz, M.J, et. al. Quantitative Risk Assessment (QRA).

THANK YOU.
ANY QUESTIONS?