Study On Optimization Of Diagnosis Model For Wear Condition Of Piston-cylinder Liner Friction Pair

With the improvement of the thermal efficiency of the diesel engine,the thermal shock borne by various parts of the diesel engine increases,and the reliability of the diesel engine needs to be improved accordingly.The piston cylinder liner friction pair,as a key component of a diesel engine,plays a crucial role in the performance of the engine.The abnormal wear of the piston cylinder liner friction pair can lead to a decrease in diesel engine performance in mild cases and a complete scrapping of the diesel engine in severe cases,posing a safety hazard.Therefore,monitoring and diagnosing the condition of the piston cylinder liner friction pair is of great significance for the normal operation and maintenance of diesel engines.The surface vibration signals of the diesel engine body contain rich state information,and the wear status of the piston cylinder liner can be obtained by extracting the characteristic parameters of the vibration signals.Combined with machine learning algorithms,the recognition of the wear status of the diesel engine piston cylinder liner can be achieved.This article mainly focuses on the optimization of a diagnostic model for abnormal wear status of piston cylinder liner friction pairs in diesel engines.Firstly,a diesel engine test bench was established to conduct tests under different torques,speeds,lubricating oil temperatures,and cylinder clearances.Transient parameters such as surface vibration acceleration,combustion pressure in the cylinder,and gear ring signals were collected.Analyze the surface vibration signals of the engine body using time-frequency analysis method to explore the relationship between combustion excitation,piston knocking excitation,and surface vibration signals of the engine body.Selecting the Variational Mode Decomposition(VMD)algorithm as the decomposition method for vibration signals,15 timedomain and frequency-domain characteristic parameters of vibration signals were selected to provide a basis for selecting input vectors for fault diagnosis models.Secondly,key parameters such as the number of decomposition layers and quadratic penalty terms of the VMD algorithm were determined.The VMD algorithm was used to decompose the surface vibration signal of the aircraft body into multiple modal components and extract the characteristic parameters of each component.By exploring the correlation between torque,speed,lubricating oil temperature,and cylinder clearance with various characteristic parameters,a highly correlated feature parameter set is preliminarily determined.Analyze the above feature parameter set through multiple evaluation criteria to obtain feature parameters with high contribution and low dimensionality as input vectors for the diagnostic model.Finally,research is conducted on the optimization of the diagnostic model for abnormal wear of diesel engine piston cylinder liners.Divide the input vectors into three categories,which are used to characterize the engine operating state,combustion state,and piston knocking state.On this basis,a training and testing set are constructed,and an abnormal wear state diagnosis model is constructed based on Support Vector Machine(SVM)theory.Optimize the key parameters of the diagnostic model through grid search methods and genetic algorithms,train the optimized model with a training set,and evaluate its performance through a test set.The evaluation results show that the accuracy of the diagnostic model optimized by genetic algorithm can reach 97.2%,with a running time of 17.6 seconds.Its accuracy and efficiency are better than those optimized by grid search method.The model optimized by genetic algorithm has good performance.