Study On Fault Diagnosis For Fuel Cell Engine

As the issues of energy and environment become more and more serious,fuel cell electric vehicle(FCEV) is viewed as an important area all over the world.Fuel cell egine(FCE) is one of the most important component of a FCEV,and performance,reliability,life span and adaptability to environment of the FCE will affect FCEV's application prospect directly.Therefore,the research aspects that accurately diagnose and eliminate fault sources in time and ensure FCE's steady and reliable working are very important.This paper focuses on fault diagnosis of FCEs based on a 60kW FCE developed by our group.The main contents are as follows:First of all,this paper compares and analyzes several fault diagnosis methods such as fault diagnosis based on the analytical model,fault diagnosis based on signal processing and fault diagnosis based on the knowledge.On the basis of FCE,the method of fault diagnosis based on fault tree and Fuzzy Neural Networks(FNN)is put forward.According to the structure of FCE,this papper has discussed sensor faults, actuator faults,controller faults and so on,what is more,analyzed the typical fault symptoms and their technical condition characteristic description parameters of self-checking defeat,start difficulty,power insufficiency and so on.Considering the distinctness of fault tree analysis,this paper takes the undesirable fault symptoms stated above as the top event of all the fault trees,then treats hardware faults,software faults,artificial factor and environment factor as the reason events and sets up the fault tree models of FCE's typical fault discussed above by means of developing them from top to bottom.Besides,all the cut sets and minimal cut sets of each fault tree model are figured out and analyzed qualitatively, at last,the fault rate of both Hydrogen leak and fuel cell stack malfunction and the importance degree of their minimal cut sets are analyzed quantitatively.Under the condition that all the sampled signals from sensors are correct,aiming at common actuator faults and misuse faults,this paper expresss the inputs and outputs with membership functions and establishes the FNN fault diagnosis model by training Artificial Neural Networks(ANN)with fault sample collected,at last tests and confirms the model with simulation.This paper has established the fault tree model and FNN fault diagnosis model of FCE's typical faults and combined them togethrer.The practical results have shown good effcet after using them in actual research work.