A Characteristic Parameter Selection Method for Vibration Diagnosis of Internal Combustion Engine

A novel characteristic parameter selection method was proposed.In the method,the class recognition rate of each characteristic parameter under each operation condition was calculated,and the two charateristic parameters with highest class recognition rates in each operation condition were chosen to form the united diagnosis vector.To validate the effectiveness of the proposed method,various valve gap settings were used to simulate the valve train faults,common amplitude domain parameters extracted from the cylinder head vibration signals were analyzed and selected,and neural networks were adopted to recognize the simulatied faults.Statistical results show that when the number of diagnosed characteristic parameter is decreased to its half,the reduction of diagnosis accuracy is less than 1%.