Predictions Studies Of Diesel Engine Combustion Noise Transfer Function And Overall Noise

With the stringent national laws and regulations and the increasing demand for consumer comfort,it is urgent to improve the NVH performance of engines.Due to the complexity of structure and the diversity of transfer path,results of traditional separation methods of combustion noise are limited.Therefore,by introducing modern signal processing technologies,accurate analysis of combustion noise transfer function and mechanical noise can be realized,which is of great significance for accurate prediction of engine noise and noise reduction.Based on the mechanism of combustion noise,the cylinder and sound pressure signals related to combustion section under various working conditions of time window function were used.Then the theory of cylinder pressure decomposition was used to preprocess the cylinder pressure signals,and the least square polynomial fitting was used to eliminate the trend term interference of noise signals.Then,time-frequency domain conversion was performed,and combustion noise transfer functions under different injection strategies were calculated by combining multiple linear regression functions.The calculation accuracy of multi-direction combustion noise transfer function was compared and analyzed,and the influence rule of each injection strategy on the calculation of combustion noise transfer function was summarized.Results showed that combustion noise transfer functions in different working conditions were consistent in all test directions,and the main center frequencies of the combustion noise transfer function in all test directions of the engine were consistent.In injection strategies,adjusting the timing of main injection to calculate the combustion noise transfer function obtained the best consistent results under various working conditions.On the whole,adjusting the parameters related to main injection to calculate the stability was better than adjusting the parameters related to pre-injection.There was a close positive correlation between combustion noise and the maximum rate of pressure rise in cylinderIn addition,an improved genetic algorithm was introduced to realize the calculation of mechanical noise separation and combustion noise transfer function,and the advantages of genetic algorithm in optimization were illustrated by comparing the traditional multiple regression algorithm.According to the improved genetic algorithm model,the combustion noise transfer function obtained by adjusting the main injection timing was calculated,and compared with the calculated value of multi-load combustion noise transfer function,it was found that it had a load consistency rule.Transfer functions of combustion noise and mechanical noise data were calculated by applying the load consistency law of transfer function to realize the accurate prediction of noise.Results showed that the algorithm could achieve accurate prediction of combustion noise transfer functions and mechanical noise with less test data,and the calculation deviation of the test noise could be less than 2%.The transfer function of combustion noise was obtained from the change of engine loads,which was roughly the same as that obtained by adjusting the injection strategies.This provided a simple and effective method for the prediction of engine combustion noise.In order to further simplify the prediction method of engine overall noise under different working conditions,the paper analogized the similarity between the transmission structure of RBF neural network and the transmission of engine’s vibration and noise,and proposed that the network could be studied and trained on the basis of input of cylinder pressure excitation in frequency domain and sound pressure spectrum in test direction.The overall noise of other load conditions could be directly predicted by learning data of different injection strategies of single load,and the overall prediction accuracy could reach above 97%.