Abstract: Diesel engine noise has become an important component of the environmental noises, the research on establishment and evaluation of diesel engine noise is of importance in theory and practice. The Q diesel engine is taken as a sample; the sound quality of the radiated noise varied with loads/speeds is studied so as to improve the radiated noise quality and establish the theoretical foundation for the structure acoustical design. On the basis of studying the objective evaluation of diesel engine radiated noise at home and abroad, the physical acoustics parameters kurtosis, impulsiveness characters and psychoacoustics parameters consisting of loudness, sharpness, roughness and fluctuation strength are chosen to describe the objective characteristics of the diesel engine noise; Considering the characteristics of the diesel engine noise, the experienced listeners and the classified-paired comparison are applied in the subjective evaluation; the prediction model of noise quality is established by the multilayer perceptron neural networks. The prediction model for diesel engine noise quality can reflect the non-linear relationship between objective characteristics and subjection evaluation results accurately.

Key words: Diesel engine, Neural network, Noise quality, Prediction model, Psychoacoustics, Subjective evaluation