A Complex System Dynamic Probabilistic Safety Assessment Method Base On ET-DFT Hierarchical Model

With the rapid development of science and technology, and the establishment of newtype of complex system, all kinds of large equipments and systems become increasinglysophisticated, as the capacity parameters improve constantly, environmental conditions aremore harsh, thus the reliability and such risks become increasingly prominent. Reliabilitytechnology researches are of great importance, which ensure safety in production, optimizeproduct structure and performance, and improve the reliability, security and maintainability ofsystem operation, expedite the development of the manufacturing science. The interactionbetween hardware, software and people exists in complex mechanical system, the state ofsystem and its components are often in a dynamic change process, as a result, when doing riskassessment, not only the interaction on system reliability and risks should be considered, butalso the responsive time of systems and components, the order of events and the effects ofcomponents on the system state. This paper’s research objective is the hydraulic system ofcomputer numerical control machine. Beacause of the shortcomings of the static probabilisticsafety assessment methods, we are building a ET-DFT model, which is setting up bycombining ET (Event Tree) and DFT (Dynamic Fault Tree). In order to make up for theinadequacy of ET-DFT model analysis, applying markov chain and discrete-time bayesianmethod in it. And then, we do the dynamic probabilistic safety assessment in the hydraulicsystem of CNC machine. In finally, it’s more objective and reliable to apply importancedegree analysis and diagnostic analysis in ET-DFT model. This paper is mainly as follows:In order to show how dynamic behavior of complicated system influences its reliabilityand reduces probability calculation on system dynamic fault, we present a dynamicprobabilistic safety assessment method based on ET-DFT model, which uses hierarchicalthoughts, then given the algorithms that the static fault tree and dynamic fault tree in ET-DFTmodel solve by BDD and MC respectively, and the algorithm of the probability of each statein the system. Taking the hydraulic system of CNC machine as research objective, we get theoccurrence probability of the static sub-modules quantitatively and the states’ probabilitycurve of the dynamic sub-modules changes over time, then combines above analyzing results,which makes the DPSA more accurate.Owing to it is existing the state space explosion inevitablely, and it is error-prone whendynamic fault tree coverts to Markov chain under ET-DFT model based on MC, we present adynamic probabilistic safety assessment for ET-DFT model based on discrete-time bayesiannetworks, and given corresponding algorithm that all static and dynamic fault tree in ET-DFT Model convert to Discrete-time Bayesian Networks. Taking the hydraulic system of CNCmachine as research objective, we draw that not only can solve each modules posteriorprobability when pressure is poor or volatile fault is happened of the hydraulic system ofCNC machine, so as to recognize the weakest module in this system, but also can obtain theprobability of failure within each time interval, therefore, the fault early warning can carryconveniently, and the DPSA for the system would more accurate.it’s more objective and reliable to apply importance degree analysis and diagnosticanalysis in ET-DFT model. Taking the hydraulic system of CNC machine as researchobjective, firstly we can utilize discrete time Bayesian networks to get four importancedegrees—FV, RRW, BM and RAW—of each modules when pressure is poor or volatile faultis happened of the hydraulic system of CNC machine, and rank them separately. Secondly,we quantitatively obtain each DIF value under the path2model by applying DFT diagnosticanalysis based on ADORA method in ET-DFT model. In finally, we determine the MinimalEvent Set group by ZBDD qualitative method, then establish DDT to help diagnostic analysis.Therefore, the position for each module in the system can be determined precisely if takingabove methods. Meanwhile, it helps to recognize the weakest module in this system, providethe theory basis for system diagnosis, operation, maintenance and reliable allocation.