Research And Development Of Automobile Engine Fault Diagnosis System

In today’s society,cars are closely related to our daily lives.At the same time,the production and sales of cars are also increasing,but with it,the number of car breakdowns is also increasing.Automobile engine is the core part of automobile.Whether the engine can work normally or not directly affects the state of the entire automobile.Therefore,the research on automobile engine fault diagnosis methods is of great significance to the entire automobile industry.The main research content can be divided into:(1)Analyzed the most frequently failed parts of the engine,and studied the structural characteristics and vibration characteristics of these parts.Study the structural characteristics and vibration characteristics of these parts.Use this as a basis to simulate the vibration signal of the engine under normal operating conditions,abnormal valve clearance and single-cylinder misfire.(2)Aiming at the problem that the marginal spectrum obtained by the Hilbert-Huang transform is not easy to observe,an EMD-AR spectrum estimation method is proposed.Perform Hilbert-Huang transform and EMD-AR spectrum estimation transform on the signals obtained from the simulation respectively,It is verified that the image obtained by EMD-AR spectrum estimation transformation is smoother and easier to distinguish.(3)Use EMD-AR spectrum estimation to simulate signals for various working conditions.Accumulate the first four groups of IMF components obtained from EMD decomposition and get the EMD-AR spectrum curve.Sum and normalize the energy of the signal segment and obtain the signal characteristic value of each working condition.(4)Input the obtained feature values into support vector machine(SVM)and extreme learning machine(ELM)for group training prediction.Using particle swarm optimization(PSO)to solve the problem of random selection of hidden layer parameters of extreme learning machine neural network.The shortest running time of the extreme learning machine,but the accuracy is not as good as the other two algorithms.All evaluation indicators of support vector machine are medium,but only suitable for classification predictions with fewer categories.PSO-ELM is stronger in terms of accuracy and stability,but its prediction takes longer.(5)Use MATLAB GUI software to compile the used signal processing and fault diagnosis methods into a complete fault diagnosis system,which makes the user’s signal processing and fault classification and prediction operations more concise and convenient.