| Small diesel engines play an enormous role in national economy and have a big market. Its noise is bigger than others because its work characteristics. Exhaust muffler is one important parts of the engine, so it is of great significance to study the exhaust muffler of the small single-cylinder diesel engine.In this paper, the noise spectrum of the ZS1105 diesel engine was collected with installing initially muffler and without muffler. According to the target value, the need noise elimination value was calculated at 1/3 spectrum. Based on the center frequency of 1/3 spectrum, the author designed a muffler structure with taper pipe inside.Based on characters of the designed muffler structure, the author designed the meter of orthogonal layout, which includes three factors and three levels, and the L18 (27) orthogonal layout was selected. The 18 kinds of the test muffler were manufactured according to the designed structure. And 1/3 spectrum of the noise elimination value was measured in the test bench.The test results were treated through B-P Nervous Network. The B-P NN includes two implied layers was selected, and the input layer contains three nerve units, the output layer includes one unit, the first implied layer contains 20 nerve units and the second implied layer includes five units. The active function of the input layer and implied layers were all Log-Sigmuid function, and the active function of output layer was Pureline function. Through compile program in the MATLAB code calling the NN toll box. the B-P NN is training by computer.In order to avoid divergence by S-function. the value of the input and output were standardized before training. Through verified by the four test results, the accuracy of BP NN is good, its maximal error was 8.9%.Using the trained BP NN, the designed muffler structure is optimized, the exhaust noise is 99.8dB through testing in the engine test bench, and the power loss was 3.3%. These were better than original muffler. |
PDF Full Text Request