基于CNN和MFCC的供水管网漏损声信号识别方法

A Method for Identifying Leakage Acoustic Signals in Water Supply Network Based on CNN with MFCC

针对供水管网漏损识别效率低和对人工经验依赖性强等问题,基于卷积神经网络(CNN)和梅尔频率倒谱系数(MFCC)提出了一种供水管网漏损声信号识别方法。对噪声记录仪和水音传感器采集的漏损声信号提取MFCC及其一、二阶差分作为漏损声信号特征,得到了包含漏损特征的特征图像,将其输入到CNN模型,通过超参数优化后最终得到了漏损识别模型。结果表明,使用MFCC与MFCC的一阶差分特征参数组合作为输入特征训练模型时的识别效果最好,其测试集准确率达到95.26%,F1分数达到89.22%,具备优良的漏损识别能力。

Aiming at the problems of low efficiency and strong dependence on artificial experience in leakage identification of water supply network,a leakage acoustic signal identification method of water supply network was proposed based on convolutional neural network(CNN)and Mel frequency cepstral coefficient(MFCC).MFCC and its first-and second-order differences were extracted from the leakage sound signals collected by noise recorders and water sound sensors as leakage sound signal features.Feature images containing leakage features were obtained and input into a CNN model.After hyperparameter optimization,the leakage identification model was finally obtained.The results showed that using the combination of MFCC and its first-order differential feature parameters as input features to train the model yielded the best identification performance,with a test set accuracy of 95.26%and an F1 score of 89.22%,demonstrating excellent leakage identification ability.