International Symposium MKOPSC, 2010 College Station, TX, USA Application of fuzzy logic to explosion risk assessment A.S. Markowski*, M.S. Mannan**, A. Kotynia* * Process and Ecological Safety Division, Technical University of Lodz, Poland ** Mary Kay O Connor Process Safety Center, Department of Chemical Engineering, Texas A&M University, College Station-TX
Content 1. Motivation, ATEX Directives in EU 2. Basic tools for explosion risk assessment ExLOPA 3. Fuzzy logic for explosion risk assessment 3.1. Fuzzy logic system 3.2. Development of fuzzy sets for the variables 3.3. Fuzzy inference engine 4. Case study sampling from gasoline storage tank 5. Conclusions 2
Motivation A very bad statistic concerning an explosion accidents at the work places. The EU introduced special regulations, called ATEX to control explosion risks at work place. These require an explosion risk assessment. The variabilty of data, lack of detail knowledge and assumption required for preceise assessment of explosion risk form number of uncertainities concering the explosion risk. To work-out semiquantitative method applicable to an explosion workplace risk assessment based on LOPA approach which can include the knowledge uncertainity aspects. It may be achieved by application of fuzzy logic (Fuzzy ExLOPA). 3
The ATEX directives ATEX 100 covers equipment and protection systems intended for use in potentially explosive atmospheres. ATEX 137 covers the minimum requirements for improving personel health and safety in hazardous area demanding the assessment of explosion risk at the workplace. In compliance with article 4 of the ATEX 137 (1999/92/EC) the employer shall assess the specific risks arising from explosive atmospheres. 4
Examples of possible process operations Sampling, small leaks Pneumatic transport, dust discharing 5
Examples of process operations Powder charging into solvent Paint spraying Filling of gas bottles Pneumatic conveing system 6
Tank explosion / fire resulting from sampling 7
General characteristic of the explosion systems Fuel and air mixture Vapour-air Type of hazard VCE or FF Major external condition Open space with congestion Open space, no congestion Gas-air Mist-air Dust-air Dust layer-air Hybrid mixtures Gas explosion Mist explosion Dust explosion Dust fire All possible Open space with congestion or closed space (enclosed structure or vessel) As above Closed space (enclosed structure or vessel) As above As above Conclusion: Wide variety of physical effects in terms of nature as well as magnitude! 8
Explosion scenarios 9
Atmospheric explosion risk Ignition source Ex Hazardous explosive atmosphere Failure of safety measures 10
Safety barriers model for atmospheric explosion Flam m able substance B 1 E atm xplosive osphere B 4 O (air) xidant B 3 E xplosion /Fire C andproperty onsequences tow orkers B 5 Ignition source E ignition source fective B 2 P reventionlayer P rotectionandresponselayers B 1-control m easuresofexplosiveatm osphere B 2- controlm easuresofignitionsources B 4-protectionsafetym easures B 5-response safetym easures B 3-procesoperatingcontrol m easures 11
Basic asssumption for explosion risk assessment 1. Explosion risk model is determined on the basis of explosion phenomena and principles of the Layer of Protection Analysis (ExLOPA) 2. The simplified approch based on the categorization of all variables should be applied 3. A fuzzy logic system (FLS) is applied to the explosion risk model 4. All variables should be expressed in form of linguistic fuzzy sets defined in their own universe of discourse (fuzzification process) 5. Inference process generated from engineering knowledge by means of the collection of IF-THEN statements allows for fuzzy risk assessment 6. Output fuzzy set representing the risk level is defuzzified into a crisp value of risk level (defuzzification process) 12
Definition of explosion risk Explosion risk can be expressed as the following product: R n (T EXP ) = f (F atex, P EFI, F SM, SC) F atex P IG F SMn SC T EXP - frequency of the occurrence of an explosive mixture, - probability of the presence of an ignition source, - probability of failure of safety measures, - the consequences of explosion, - time of exposure (work time). 13
Explosion risk evaluation model - ExLOPA 14
Uncertainty - lack of knowledge 1. Explosion risk assessment using ExLOPA is a complex problem as characterized by the presence of knowledge type of uncertainty. It means the possibility of predicting wrong risk index (over or uderestimated). 2. 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. 3. Fuzzy set analysis gives a possibility of better insights into hazards and safety phenomena for each explosion risk scenario. It is not possible to receive such conclusions from the traditional ExLOPA calculation results. However it requires the application of computer-aided analyses which may be partially in conflict with a simplicity of ExLOPA. 15
Fuzzy Logic System ( FLS) 16
FLS for explosion risk assessment 17
Fuzzy sets for the frequency category of explosive atmosphere, K atex Classification of hazardous areas Linguistic term Linguistic category Gas Vapor Dust 0 20 Permanently P 1 21 Occasionally O 2 22 Unlikely U 18
Fuzzy sets for category of ignition sources, K EFI Linguistic term Continuous (certain) Description, example Operational type Linguistic probability category K EFI CO Rare Due to occasional failure of control ignition parameters R Very Rare Very rare failure of control ignition parameters VR 19
Fuzzy sets for category of safety measure, K SM Linguistic term Description, example Linguistic probability category K SM Very high Basic requirements for control of ignition source plus two additional independent explosion protection measures I High Basic requirements plus one additional independent explosion protection measure II Standard Basic requirements for ignition source III 20
Fuzzy sets for category of severity of consequence, K SC Linguistic term Description Linguistic Severity Category,K SC 1 Negligible Very minor N 2 Minor Minor injury MI 3 Medium Single injury M 4 Major Serious injuries MA 5 - Catastrophic Fatality or multiple injuries C 21
Fuzzy sets for risk category, K R Linguistic term Description Linguistic Risk Category,K R A Acceptable No further action 1 TA Tolerable acceptable TNA tolerable not acceptable NA not acceptable ALARP 2 Additional risk control required in due time Additional risk control required immediately 3 4 22
Fuzzy inference system 1. FIS F - Fuzzy IF-THEN rules for frequency category of the explosion, K F on the basis of the K atex,k IG, K SM ). 2. FIS SC - Fuzzy IF-THEN rules for severity of category assessment, K SC, on the basis of amount explosive mass and properties. 3. FIS R - fuzzy IF-THEN rules for risk category assessment, K R, on the basis risk matrix, presented below. 4. All together for one exposure condition determined by exposure time, there are 135 rules used. 23
Case study - sampling from gasoline storage tank No. Possible accident scenario 1. Explosion in head space of the tank due to ignition by electrostatic spark 2. Explosion in head space of the tank due to ignition by lightning 24
Final results K SC K R No. K atex K EFI K SM Traditional Traditional Fuzzy Fuzzy 1. P VR III 5 4.68 TNA 2.77 TA 23% TNA 77% Description MPA most probable accident 2. P VR II 5 4.68 TA 2.00 TA 100% Additional SM 25
Conclusions 1. Explosion risk assessment for workers employed in potentially explosive atmospheres can be successfully approached by the Layer of Protection Analysis (ExLOPA). 2. Risk model estimation comprises a multi-step process in which the relation between all input categories and output category of risk is achieved by the set of inference relations: IF AND THEN. 3. The evaluation of categories and inference relations between them involve a number of uncertainties which may be classified as the lack of knowledge. 4. The subjective and vague problems connected with these input data may be successfully solved by modified ExLOPA named Fuzzy ExLOPA, where the fuzzy set theory is applied. It requires the application of computer-aided analyses which may be partially in conflict with a simplicity of ExLOPA. 5. The case study proves that final results of risk are more realistically determined and offer an advantage with respect to the traditional single-point estimations. It helps the analyst to make better decision if additional safety measures are required. 26