Heterogeneous redundancy optimization for multi-state series-parallel systems subject to common cause failures Chun-yang Li, Xun Chen, Xiao-shan Yi, Jun-youg Tao Reliability Engineering and System Safety Volumes 95 (2010) p.202-207 Advisor: Yung-Sung, Lin Presented by : Hui-Yu, Chung Agenda Introduction Problem Formulation Reliability Estimation of the System The UGF of a Component The UGF of Subsystems without CCFs The UGF of the Subsystem with CCFs The Reliability of the System Genetic Algorithms A multi-state series-parallel system with CCFs Mathematics Model Solution Encoding and Initial Population Creation Individual Evaluation by Fitness Function Selection, Crossover & Mutation New Population Formation and Termination Numerical Example Conclusion 2 Introduction Common Cause Failures (CCFs) The simultaneous failure of multiple components due to a common cause (CC). Exists in many systems composed of redundant components CC Events Environmental loads Errors in maintenance System design flaws. 3 Introduction Multi state system (MSS) Universal generating function (UGF) Compared with Binary Systems Work in different performance levels A technique first stated from Levitin Gregory’s book ”The universal generating function in reliability analysis and optimization” Used to estimate the system reliability Genetic Algorithm (GA) Used to optimize the system structure 4 Agenda Introduction Problem Formulation Reliability Estimation of the System The UGF of a Component The UGF of Subsystems without CCFs The UGF of the Subsystem with CCFs The Reliability of the System Genetic Algorithms A multi-state series-parallel system with CCFs Mathematics Model Solution Encoding and Initial Population Creation Individual Evaluation by Fitness Function Selection, Crossover & Mutation New Population Formation and Termination Numerical Example Conclusion 5 Problem Formulation Four General Assumptions: The components are not repaired The system and components are multi-state Mixing of components of different types in the same subsystem is allowed When any load is beyond the limit of components, all components of this type will fail 6 Problem Formulation 7 A multi-state series-parallel system with CCFS N subsystems connected in series Subsystem I consists of hi different types of components in parallel 8 Mathematics model 9 Agenda Introduction Problem Formulation Reliability Estimation of the System The UGF of a Component The UGF of Subsystems without CCFs The UGF of the Subsystem with CCFs The Reliability of the System Genetic Algorithms A multi-state series-parallel system with CCFs Mathematics Model Solution Encoding and Initial Population Creation Individual Evaluation by Fitness Function Selection, Crossover & Mutation New Population Formation and Termination Numerical Example Conclusion 10 Reliability Estimation of the System 4 approaches to estimate MSS reliability UGF technique Structure function Stochastic process Monte Carlo simulation Here, we use UGF approach System structure, performance & reliability 11 A UGF Component Random performance G Represented by two sets gij & qij g ij {g ij 1, g ij 2 ,..., g ijM } qij {qij 1, qij 2 ,..., qijM } Definition: 12 The UGF of subsystems without CCFs Operations: (SUM) (MAX) 13 The UGF of the subsystem with CCFs Subsystem f composed of identical components of type j: Subsystem f composed of different types of components: 14 The Reliability of the System Usually, the performance of the system is equal to the minimum of performance of subsystems Operation: UGF of the System: 15 The Reliability of the System Define operation : The reliability of the system is: 16 Agenda Introduction Problem Formulation Reliability Estimation of the System The UGF of a Component The UGF of Subsystems without CCFs The UGF of the Subsystem with CCFs The Reliability of the System Genetic Algorithms A multi-state series-parallel system with CCFs Mathematics Model Solution Encoding and Initial Population Creation Individual Evaluation by Fitness Function Selection, Crossover & Mutation New Population Formation and Termination Numerical Example Conclusion 17 Genetic Algorithms Genetic Algorithm An optimization technique based on concepts from Robert Darwin’s “evolution theory” Initialization Selection Reproduction Termination 18 Solution encoding & Initial population creation Representation of chromosomes: Integer strings vk (n11, n12,..., n1H1 , n21, n22,..., n2H2 ,..., nN1, nN 2,..., nNHN ) Population size: pop_size nij Z, nij 0 1 ni nmax 19 Individuals evaluation by fitness function Fitness function The sum of the objective and a penalty function determined by the relative degree of infeasibility Rk: system reliability; R0: acceptable reliability fk Ck K max(0,R0 Rk ) 20 Selection, crossover & mutation Selection fmax fk fk fmax fmin Use roulette wheel selection method to select individuals for reproduction Single-point crossover & mutation Crossover probability: pc Mutation probability: pm 21 New population formation and termination Termination: A pre-defined maximum generation Nmax_gen is reached The best feasible solution has not changed for consecutive Nstall_gen generations. 22 Agenda Introduction Problem Formulation Reliability Estimation of the System The UGF of a Component The UGF of Subsystems without CCFs The UGF of the Subsystem with CCFs The Reliability of the System Genetic Algorithms A multi-state series-parallel system with CCFs Mathematics Model Solution Encoding and Initial Population Creation Individual Evaluation by Fitness Function Selection, Crossover & Mutation New Population Formation and Termination Numerical Example Conclusion 23 Numerical example Multi-state series-parallel system 4 subsystems Formulation: 24 H1=4 H2=5 H3=6 H4=4 25 According to Eq.12, the UGF of subsystem 1 is: 26 Using Genetic Algorithm to solve K=100 =0.001 Nmax_gen=1000 Nstall_gen=500 pop _ size {20,50,100} p {0.5,0.8,1.0} c p {0.01,0.1,0.2} m 27 Using GA & Hybrid GA Hybrid GA: Fuzzy logic controller regulates GA parameters automatically Exploitation around the near optimum solution Longer, but the parameter tuning time is not taken into account. 28 Results of reliability optimization with/without CCFs Optimal result is different Mixing of components of different types enhances the system reliability and controls the cost The effect of CCFs cannot be ignored 29 Agenda Introduction Problem Formulation Reliability Estimation of the System The UGF of a Component The UGF of Subsystems without CCFs The UGF of the Subsystem with CCFs The Reliability of the System Genetic Algorithms A multi-state series-parallel system with CCFs Mathematics Model Solution Encoding and Initial Population Creation Individual Evaluation by Fitness Function Selection, Crossover & Mutation New Population Formation and Termination Numerical Example Conclusion 30 Conclusion CCFs reduce the effect of components redundancy CCFs make the redundancy allocation strategy different Mixing of components of different types is very useful to improve the reliability of MSSs subject to CCFs 31 ~The End~ ★~Thanks for Your Attention~★ 32