基于声音深度学习的发动机失火故障诊断  

Engine misfire fault diagnosis based on sound deep learning

作　　者：李志臣 凌秀军[1] 李鸿秋[1] Li Zhichen;Ling Xiujun;Li Hongqiu(School of Mechanical&Electrical Engineering,Jinling Institute of Technology,Nanjing,211169,China)

机构地区：[1]金陵科技学院机电工程学院,南京市211169

出　　处：《中国农机化学报》2024年第9期134-140,共7页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金面上项目(51775270)。

摘　　要：构建基于深度学习的轻量级卷积神经网络实现发动机声音信号的失火故障检测。运用计算机的麦克风阵列记录发动机不同转速下的正常状态、一缸失火状态、二缸失火状态的声音信号。将声音信号转换成时频图像应用于卷积网络模型的训练、验证和测试。声音时频图像特征提取网络主要由可分离卷积模块构成。特征提取网络连接时频图像分类器,对含有不同个数的特征通道分组卷积模块的网络模型进行训练、验证和测试试验的比较分析。设计的卷积神经网络应用于发动机失火故障检测的准确率达到99.60%。网络的计算量小、检测时间短。基于特征通道分组卷积的深度学习网络能够快速地完成对发动机失火故障声音信号的检测诊断,为发动机失火故障的在线实时检测提供智能决策支持。A lightweight convolutional neural network based on deep learning was constructed to realize misfire fault detection of the engine sound signals.The computer's microphone array was used to record engine sound signals in different states including normal,one‑cylinder misfire and two‑cylinder misfire with different engine speeds.These sound signals were then converted into time‑frequency images.These images were applied to the training,verification and testing of a convolutional neural network model.The sound time‑frequency image feature extraction network is mainly composed of separable convolution modules.The feature extraction network connects the image classifier.Comparative analysis of training,verification,test experiments are conducted on the network model with different numbers of feature channel grouping convolution modules.The designed convolutional neural network has 99.60%accuracy applying in engine misfire fault detection.The calculation amount of the network is small and the detection time is short.A deep learning neural network based on feature channel packet convolution can quickly complete the detection and diagnosis of the sound signal of engine misfire fault.The method provides intelligent decision support for the online real‑time detection of the engine misfire fault.

关 键 词：发动机 失火故障 深度学习 特征通道分组卷积 声音 时频图像 

分 类 号：S219.031[农业科学—农业机械化工程] TP391.4[农业科学—农业工程]