Research On Fault Diagnosis Of Internal Combustion Engine System Based On GA-BP Neural Network

Internal combustion engine plays an important role in many fields.For example,large agricultural equipment,construction machinery and equipment,military equipment,ships and other heavy weapons warships,is worthy of the name of the industrial power of internal combustion engine.Tamping machine is a large railway maintenance equipment,is a typical big machine.The utility model is characterized in that the utility model has the advantages of hundreds of thousands of tons of its weight working,and the power supply of the internal combustion engine is completely provided by the internal combustion engine.Therefore,in view of the tamping machine,internal combustion engine system is one of the core equipment.When the engine fails,if cannot be found in time which may cause engine damage sustained to the hidden dangers of social production,and mechanical equipment itself with operation,affecting the mechanical life.It is possible to reduce this phenomenon by regular storage maintenance,but it is not only waste of manpower and material resources,but also reduce the working efficiency.If the internal combustion engine has been broken,and at this time as scheduled not to repair time,mechanical equipment will continue to work.Therefore,it is important and meaningful to diagnose the faults of the internal combustion engine system.In this paper,by analyzing the common methods and methods of the current fault diagnosis,some diagnostic models have been focused on the study of mechanical vibration,while others are simply analyzed by a single algorithm.These are some shortcomings and limitations.Inspired by bionic technology,the research object and research methods are improved.The examination method of the doctor to the patient,the emissions of internal combustion engine as the research object,through analysis and study of the internal combustion engine to exclude carbon oxides,nitrogen oxides,hydrocarbons,and temperature and humidity,emissions discharge rate and other related data.The failure of the internal combustion engine is reflected by the mechanical condition and the internal combustion of the cylinder.For example,the nozzle clogging of the fuel system of an internal combustion engine,will cause the fuel into the combustion chamber is low.which leads to mixed gas,incomplete combustion,increase in emissions is the direct reaction of carbon oxide content.In this paper,the artificial neural network is introduced to improve the efficiency of BP neural network.After analyzing the shortcoming of the neural network which is easy to fall into the local optimum,the paper introduces the improvement of the momentum and adaptive learning rate.The improved BP neural network training is optimized,but it does not eliminate the defects.While the GA has strong global search ability,and integrate the two to use of GA-BP neural network weights and threshold and network structure for global optimization,in the process of training network BP algorithm by backward error correction.the network weights and thresholds obtained.These two algorithms can achieve complementary advantages,give full play to each other's advantages.In the process of modeling and experiments,this paper established the fault diagnosis model based on GA-BP neural network,research on discharge using F12L413F air-cooled diesel engine DEUTZ data,the simulation model of MATLAB for fault diagnosis.GA-BP neural network simulation model compared with the BP neural network and optimized the results show that BP neural network has undergone 342 step training to complete convergence,while the optimized GA-BP neural network after about 60 steps to complete the training convergence.The results show that the optimized GA-BP neural network has a significant improvement in the accuracy of the fault diagnosis of the internal combustion engine system.