Research About Quick Dft-based Reliability Analysis And Risk Vs Cost Multiple Objectives Optimization Of Nuclear Power Plant Systems

The fault tree and event tree risk model based static PSA(Probabilitic Safety Assessment)technique plays a key role in building,operating,risk evaluation and routine management of Nuclear Power Plant(NPP),and also is an inevitable content of submitting a safety reviewe report.However,along with the appearance of complex designing systems and the development of reliability theory,it is found that the traditional static PSA technique is incapable of analyzing reliability of the complex systems with dynamic sequence failure behaviors.The complexity of NPP systems embody in three aspects: the sequential failure behaviors of the systems;the diversity of the system equipment failing;the contradiction of managing the risk and cost of the low-demand standby safety systems of NPP.The first one implies that the failure of systems not only depends on combination logics of equipment but also depends on their failure sequences.The second one means that a failure of equipment can be either caused by a random behavior(i.e.,reliability failure),or caused by its regular test,maintenance activities(i.e.,command failure).To the last one,the routine managers of the NPP hope the strategies they take can keep the low-demand standby safety systems of NPP at higher level of reliability,meanwhile,at less expenditure.Obviously,it is a multiple objective optimizaition problem.As to the reliability evaluations of complex systems with sequence-dependent failure behaviors,they are often modeled and analyzed by dynamic fault trees(DFTs).DFTs are first applied to reliability evalutions of fault-tolerant computer systems,and then,such techniques are gradually applied to systems of aeronautical and space,NPP,chemical process plant.Nowadays,the techniques for solving DFTs are still in development.Especially,for largescale and highly coupled DFTs,the solvability and efficiency are required to be further improved.SDPs(i.e.,sum of disjoint products)are considered one of the most effective methods.However,when applying such approaches to solve large-scale and highly coupled DFTs,three main problems remain to be handled: quantification of a generalized cut sequence;negation of a generalized cut sequence;efficient SDP models building algorithm.As to the first problem,a universal negating formula based on the Morgan theorem is proposed.For the second problem,multi-integration formulas based on a sequential failure region are presented.And to the third problem,according to the scale and coupled forms,the author has put forward three SDP models building algorithms: As to the small scale and highly coupled DFTs,an adapted K.D.Heitdtmann based method is presented;for the medium and large scacle and highly coupled DFTs,especially for DFTs where dynamic logic gates are directly coupled(which are called I-type DFTs in the paper),a dynamic binary tree method is proposed;and for the large scacle and highly coupled DFTs where dynamic gate are located in lower or bottom levels(if dynamic gates are coupled,they are coupled due to repeated basic events),which are called II-type DFTs in the thesis,a dynamic binary decision diagram method is developed.To improve the computing efficiency of these SDP approaches,three corresponding building strategies are developed.Compared with the traditional DFT solving approaches,these methods are much more efficient.In addition,as to numerically simulating of non-repairable large-scale and highly coupled DFTs,The author has also made deep research and presented a ‘minimal cut sequence set(MCSS)plus sequential failure regions' based simulation method which can be viewed as a standard tool for numerical simulation of non-repairable DFTs.As to the reliability assessent of dynamic systems having equipment with multiple failure behaviors,a MCSS-based Monte Carlo simulation method is also presented.Such simulation approach simulates the failure time intervals of the considered systems during their missioin time to obtain the reliability indexes according to the built statistical quantites.With regards to the low-demand standby safety systems of NPP,NPP managers not only concern their reliability,but also concern their maintenance expenditures.It is highly desirable to achieve the maximum level of availability at the least maintenance cost.Obviously,it is a multi-objective optimization problem.To solve this problem,in this paper,a multiplepopulation multi-objective mixed integer particle swarm optimization framework is built by the author.It has solved three problems existing for multiple objective optimization: the problem of fitness assignment for multiple objectives;unpdating problem of the external shared archive;the problem of mixed integer particle swarm optimization.To conclude,our work is helpful to quick reliability analysis of NPP systems with dynamic sequence failure behaviors,and multi-objective optimization of the low-demand standby safety systems of NPPs considering the risk and cost.