The Design Of Diesel Engine Exhaust Muffler On The Basis Of Artificial Neural Network

The noise level of diesel engine has become one of important indicators on evaluation of its quality and performances. Exhaust noise is a large proportion in the overall noise of diesel engine, and the effective method of its control is the application of muffler with good performances. In this paper, ANN was applied to the muffler design with the efforts to reduce the intrinsic error that was caused by the assumption on which the muffler's acoustic model was based. In order to deeply research the effect of structural parameters on the spectrum of exhaust noise, the author designs a system to sample the exhaust noise and 1/3 FIR octave-band filters to calculate the spectrum of the sampling-data. By the means of orthogonal planning and digital signal processing, the learning-samples were obtained. The learning-samples' inputs were the sound pressure levels of 1/3 octave-bands and the specific fuel consumption of diesel engine, and its outputs were the structural parameters of muffler. Through the Neural Networks Toolbox of MATLAB, the training of an ANN was finished. The practicability of application of ANN in the muffler design was perfectly proved by the practical examples, meanwhile the application of the trained-network shows that ANN has many advantages in this area.