基于改进PNCC-SVM的滚动轴承故障声纹识别方法

Voiceprint Recognition Method for Rolling Bearing Faults Diagnosis Based on Improved PNCC-SVM

针对滚动轴承声信号分析信噪比低、易受环境噪声干扰的问题,提出一种基于改进功率归一化倒谱系数(Power-Normalized Cepstral Coefficients,PNCC)和支持向量机(Support Vector Machine,SVM)的滚动轴承故障声纹识别方法。首先对轴承声信号进行预处理;然后提取改进的PNCC并将其作为特征向量;进而根据SVM算法建立声纹识别模型对轴承故障类型进行识别,并测试所提方法在叠加噪声后的识别准确率。结果表明,改进PNCC具有识别准确率高的特点,在噪声干扰下相比原始PNCC识别准确率均值提高13.35%,鲁棒性更强。研究结果可为滚动轴承的声信号特征提取和故障识别应用提供参考。

Aiming at the problems of low SNR and being prone to be disturbed by environmental noise in the sound signal analysis of rolling bearings,a voiceprint recognition method for rolling bearing faults diagnosis based on improved power-normalized cepstral coefficients(PNCC)and support vector machine(SVM)is proposed.Firstly,the bearing sound signal is preprocessed,and the improved PNCC is extracted as the feature vector.Then,the voiceprint recognition model is established by SVM algorithm to identify the bearing fault type,and the recognition accuracy of the proposed method after superimposing the noise is tested.The results show that the improved PNCC has the advantage of high recognition accuracy.Compared with the original PNCC,the average recognition accuracy is raised by 13.35%under noise interference,and the robustness is stronger.The research results may provide a reference for the application of sound signal feature extraction and fault identification of rolling bearings.