基于GA-BP神经网络的柴油机噪声品质评价方法

Dieselengine noise quality evaluation method based on GA-BP neural network

在柴油机运转中存在有比汽油机更大的噪声与振动,当其强度达到一定程度时,会给环境造成严重的危害。基于对柴油机噪声品质预测的目的,采用(GA-BP)算法建立了一种预测模型。其中的BP算法是一种基于梯度下降原理在局部中寻优的算法,由于人对柴油机噪声的主观评价是个非线性的过程,BP算法可以解决非线性可分问题,所以可以在柴油机噪声品质主观评价中应用。但是BP算法的过程收敛速度慢,很有可能陷入局部极小值。而遗传算法(GA)具有全局寻优的优点。文中通过将二者结合起来.由GA寻找最优的BP神经网络权值与相应节点的阀值,可以有效防止搜索过程收敛于局部最优解。通过仿真结果得出:此方法既能快速收敛,又能大大提高避免陷入局部极小的能力,并且预测精度高,为柴油机噪声主观评价提供了一种新思路。

In the diesel engine running there are greater noise and vibrationthan gasoline engine, when the strength reaches a certain degree, it will do serious harm to the environment. In order to predict the quality of diesel engine noise, this article establishes a kind of evaluation model based on （GA - BP） algorithm. The BP algorithm is a kind of algorithmbased on gradient descent principle in the local optimization.Because the subjective evaluation of the diesel engine noise that people make is a nonlinear process and BP algorithm can solve the nonlinear separableproblem, it can be applied to the subjective evaluationof diesel engine noise quality. But the the process of BP algorithm has slow convergence speed, it is likely to reach local minimum and the genetic algorithm （GA） has the advantages of global optimization. In this paper, we combined the two. Finding the optimal BP neural network weights and threshold of corresponding node through GA can effectively prevent the search process to convergeto local optimal solution. Simulation results show that this method can Rapid convergence and greatly improve the ability of avoidin greaching local minimum, and the prediction precision is high.So it provides a new way of thinking to the diesel engine noise subjective evaluation.