Research On Engine Monitoring Technology Driven By Data And Knowledge

To study the engine monitoring system,we must face two major technical problems:engine baseline mining and fault diagnosis.Baseline acquisition is a prerequisite for fault diagnosis.Baseline mining and fault diagnosis models are constructed for these two problems respectively.The specific work is as follows:(1)To improve the accuracy of baseline mining,a baseline mining model combining BAS(Beetle Antennae Search)and Elman is proposed.In order to solve the problem of low efficiency of BAS optimization,the BAS algorithm is grouped;in order to solve the problem of searching accuracy of BAS algorithm,the BAS algorithm is modified with adaptive thought.Finally,the population ADP BAS algorithm(improved BAS)is combined with Elman to establish the population ADP BAS-Elman model.The calculation results show that the convergence speed and accuracy of the group ADP BAS algorithm are higher than those of the traditional optimization algorithm,and the baseline fitting accuracy and learning ability of the group ADP BAS-Elman model are better than those of the traditional model.(2)To improve the accuracy of engine gas path fault diagnosis,a diagnostic model combining PSO(particle swarm optimization)and ELM(Extreme Learning Machine)is proposed.To overcome the shortcoming that PSO is easy to fall into local optimum,a chaotic algorithm is proposed.In order to solve the contradiction between the search accuracy and the search speed of PSO,the PSO algorithm is improved by using adaptive thought.The calculation results show that the chaotic ADP PSO optimization algorithm(improved PSO)has higher convergence speed and accuracy than the traditional optimization algorithm;the hidden layer activation function in ELM is Tanh function,which is more suitable for diagnosis;the diagnostic accuracy and learning ability of the chaotic ADP PSO-ELM diagnosis model are significantly higher than the traditional diagnosis model.(3)In order to establish an accurate CBR(Case-based Reasoning)gas path fault diagnosis model,a PW4056 engine case base was established.In order to simplify thecalculation of CBR,the redundant parameters are eliminated according to the engine principle.Considering the directivity of engine parameters,a cosine-improved grey relational degree(Cosine-IGRA)case similarity calculation function is proposed.The chaotic ADP PSO algorithm is used to find the optimal weights for each parameter.Finally,a CBR diagnosis model based on osine-IGRA function is established.The simulation results show that the model has high diagnostic accuracy.