Uncertainty aspects in process safety analysis A.S. Markowski*,M.S. Mannan**, A.Bigoszewska* and D. Siuta* *Process and Ecological Safety Division Faculty of Process and Environmental Engineering Technical University of Lodz, Poland **Mary Kay O Connor Process Safety Center Department of Chemical Engineering Texas A&M University, College Station-TX Process and Ecological Safety Division TU Lodz MKOPSC Symosium 2008 College Station,, TX, USA 1 Mary Kay O Connor Process Safety Center TAMU
Motivation 1. to disccuss issue of uncertainty in traditional the Process Safety Analysis (PSA) 2. to propose the general method to handle a PSA taking into account the uncertainty 3. to demonstrate the application of fuzzy logic in PSA, e.g. in consequence analysis 2
Uncertainty Uncertainty is a term used in different ways in a number of fields. Uncertainty applies to predictions of future events or to the unknown. It is essentially the absence of information, information that may or not be obtainable. All engineering calculations are affected by uncertainties In terms of PSA, uncertainty means the possibility of predicting wrong risk index. 3
Dealing with uncertainties (three ways) Neglecting uncertainties may lead to faulty decision bases for dimensioning components and, hence, to components which are too weak. Safety factors (expert opinion) which may lead to an insufficient design, overdesign etc. By modelling which may essentially reduce the uncertainities 4
Advantages of accounting for uncertainties The information base used becomes broader. The meaning of safety factors becomes evident, safety reserves are made explicit. The credibility of the results is increased. Indications of areas are given where models and data should be refined. If the quality of the different input data for treating a problem differs, this fact is propagated through the calculations and reflected in the final result. 5
Uncertainties approaches Uncertainties in PHA Physical variables (objective) Lack of knowledge (subjective) Type of uncertainty Probability theory Variation in the results Expert knowledge Possibility theory Uncertainty measure Statistical approaches Sensitivity analysis Rule-based system Fuzzy sets theory Uncertainty approaches Classical methods Probabilistic methods Sensitivity analysis Expert system Fuzzy logic system Uncertainty techniques Best estimate Probability distribution Correlation coefficients Certainty factor Fuzzy set Uncertainty representation 6
Representation of uncertainity results and their interpretation Risks are normally characterized by indicating the frequency of occurrence of an accident and the corresponding severity of consequence. It is useful to indicate the uncertainty associated with the final result. A distinction is made between: risk index (for semi quantitative methods) - group risk individual risk, depending on method used for RA 7
Uncertainty sources and approaches Type of uncertainty Characteristic Approaches Methods Objective (aleatory) Physical variability Probabilistic Statistic Sensitivity Subjective (epistemic) Lack of knowledge Possibility theory Fuzzy sets All types od uncertainties occur in PSA, especially subjective type 8
Uncertainty For PSA it is convinient to distinguish the following types: completeness uncertainty modeling uncertainty parameter uncertainty 9
PSA framework (traditional) Hazard analysis (quntitative) Risk analysis (semi-quantitative or quntitative) Frequency of RAS Hazard identification Representative accident scenario selection (RAS) Accident scenario logic structure Risk estimation and assessment Severity of consequences Steps of PSA: hazard identification consequence assessment frequency risk evaluation 10
1st step of PSA Hazard identification Main goal Identification of accident scenario Main tool HAZOP PHA FTA ETA Types of uncertainty Completeness Modelling Parameter Incomplete identification of all accident scenarios as well errors in screening of hazards Interaction between different contributors and variables in accident scenario models Imprecision or vagueness in characteristic properties of contributors and variables Completeness uncert.- main component of uncertainity 11
2nd step of PSA Consequence assessment Main goal Main tool Types of uncertainty Completeness Modelling Parameter Heath, property and environmental losses Consequence models Incorrectness in identification of all types of the consequences as well as of all interactions among consequences Complexity phenomena and imprecision of source terms, dispersion and physical effects Inadequacy or vagueness in values for model variable Modelling uncert.- main component of uncertainity 12
3rd step of PSA Risk evaluation Main goal Risk index or risk level Main tool QRA LOPA Types of uncertainty Completeness Modelling Parameter Limited depth of assumptions in: external conditions, number of accident outcome cases and incorrectness in interpretation of results Inadequacy in selection of appropriate risk measures as well as of risk acceptance criteria Lack of real time data for weather conditions and population, for real failure rates and human errors Parameter uncert.- main component of uncertainity 13
Fuzzy logic for process safety PSA is a complex problem as characterized by the presence of diffrent types of uncertainty contained in the variables, models and assumptions. Such a complex system is difficult to precise analysis. Where no precise analysis and ambiguity or vaguiness take place the fuzzy set analysis can help. PSA can be treated as a fuzzy concept because plant safety cannot be strictly classified as a safe or unsafe (because of the existence of inherent hazards); therefore level of safety may belong simultaneously to safe state category and to unsafe state category with some memberships; this can only be realized by fuzzy sets. 14
Fuzzy set Precisly determined boundary Fuzzy boundary State partialy belongs to set Unsafe state Safe state Classical set Unsafe state Safe state Fuzzy set Membership of state to fuzzy set A (x) where: A : X [01, ] A {( x, ( x)); x X} A μ (x) is membership function describing degree of belonging for x in A 15
Fuzzy set 16
Fuzzy logic system (FLS) 17
Apllication for PSA (assumptions( for fpsa A fuzzy logic system is applied to all elements of the PSA All linguistic variables (frequency, severity and risk) are represented by fuzzy logic sets (fuzzification), defined in their own universe of discourse Output fuzzy frequency (F) is calculated on the basis of bow-tie model Output fuzzy severity of consequences (S) is assessed using an expert opinien or applying fuzzy arithmetics to the consequence models (parameter method) Output fuzzy risk index (R) is assessed using FLS where knowledge of rules is provided by a risk matrix Fuzzy Risk Correction Index (RCI) is used to take into account the uncertainties involved in identification of RAS 18
Fuzzy PSA (fpsa( fpsa) PHA & RAS Traditional part Bow-tie model for RAS FLS for F(RAS) FLS 1 2 for S(RAS) Fuzzy part FLS for R O(RAS) 3 FLS 4 5 FLS Risk index R(RAS) 19
Fuzzy risk surface FUZZY RISK MATRIX CLASSICAL RISK MATRIX 20
Consequence analysis 21
FLS for consequence analysis Two methods may be used: 1. simplified method based on the categorization of the severity of consequences into separate categories; further process applies assigning of fuzzy set for that category of release (fuzzification) and this is input data for fuzzy risk matrix, 2. parameter method used for particular consequence model, e.g. BLEVE model and the apllication of fuzzy arithmetic on the model. 22
FLS for consequence analysis parametr method 23
FLS for BLEVE an example Important output data: distance (radius) to hazardous radiation level 24
BLEVE calculation I =E p F view 2 T[kW/m] 25
BLEVE - an example - 600 m3 tank with LPG with the help of PHAST program -three threshold values for thermal radiation - 4, 12.5, 37.5 kw/m2 26
Fuzzification of sensitive parameters -fuzzy sets Membership function 1 0,5 273 283 0 243 240 260 280 313 300 320 T [K] Ambient temperature Membership function 1 0,5 0 0,5 0,6 0 0,1 0,5 0,9 1 Filling degree Membership function 1 70 80 0,5 0 0 20 30 40 60 80 100 120 X [%] Air humidity 27
BLEVE results Range of distance for different radiation levels 28
BLEVE results Type of analysis Range of distance to radiation level [m] 4 kw/m 2 12.5 kw/m 2 37.5 kw/m 2 Non fuzzy 876 506 283 Fuzzy 793 467 264 Overprediction of hazardous zone distance by about 10 % 29
Conclusions Process Safety Analysis (PSA), representing numbers of uncertainties those may lead to important overlooks in the risk assessment of the process plants. One of the promising methods for reduction of the uncertainties in process safety assessment is fuzzy logic, which allows to apply imprecise and approximate data that are typically met in PSA to receive a quite precise output results. The fuzzy PSA model is presented where FLS is built in the particular components of RA. Preliminary tests indicated that fuzzy PSA can produce more precise results concerning both elements of risk (frequency and severity of consequences). It is also possible to include the effect of the quality of PSA on 30 overall risk index assessment by means of fuzzy Risk Correction