基于BP神经网络的电机噪声在线评价系统

Analysis of an Online Evaluation System of Motor Noises based on BP Neural Network

为有效甄别存在异常噪声的车窗电机,提出一种基于BP神经网络的车窗电机噪声在线评价系统。由于现有客观参量不能够完全适用于车窗电机的噪声评价,根据车窗电机噪声特点提出了一种形如窗函数的加权因子用于修正尖锐度参量,并通过主客观评价实验优化了心理声学客观参量。在此基础上,以有效的客观参量和物理参量为特征值,构建使用附加动量法优化的BP神经网络分类器,并最终建立了噪声在线评价系统。测试结果表明,该系统对车窗电机噪声的分类准确率在90%以上,且与传统BP神经网络分类器相比,具有更高的准确率和更少的耗时,可用于车窗电机噪声的在线评价。

In order to realize the effective detection of the noise of window motors, an online evaluation system for window lift motor noise based on BP neural network is presented. Since the current objective parameters is not completely applicable to the window motor noise evaluation, a weighting factor similar to the window function is proposed for harpness parameter correction according to the characteristic of the window lift motor noise. Through the subjective and objective evaluation experiments, the psychoacoustic objective parameters are optimized. On this basis, with the effective objective parameters and the physical parameters as the characteristic values, the BP neural network classifier with additional momentum method is constructed, and the online noise evaluation system is established. Test results show that the accuracy of the evaluation system for the window lift motor noise classification can be above 90 %. In comparison with thetraditional BP neural network classifier, this online evaluation system has higher accuracy and less time consuming. So, it is feasible for online evaluation of the window lift motor noises.