基于神经网络的电机噪声性能在线检测技术研究  

RESEARCH ON ONLINE DETECTION OF MOTOR NOISE CHARACTERISTICS BASED ON NEURAL NETWORK

作　　者：蒋伟康[1] 严莉[1] 

机构地区：[1]上海交通大学振动,冲击,噪声国家重点实验室,上海200030

出　　处：《振动与冲击》2004年第4期51-53,57,共4页Journal of Vibration and Shock

摘　　要：根据车用直流电机振动、噪声的特点 ,在自动生产线上测量电机振动 ,从中提取加速度均方根值、振动能量的波动度、分频段振动能量、频谱中 5 0个最大峰值等特征指标用于电机噪声特性检测。考虑不同型号电机振动噪声特性的差异 ,建立了电机噪声检测的BP网络模型 ,用以训练噪声检测的算法和阈值。将MATLAB语言环境下的神经网络工具箱和LabVIEW虚拟仪器平台相结合 ,在LabVIEW平台上开发了基于振动测量的电机噪声智能检测系统 ,解决了电机制造厂在线检测电机噪声的难题。The vibration and acoustic noises of direct current motors are analyzed. According to the characteristics of acoustic noise and vibration of motors, some measurable parameters such as RMS of acceleration, waviness of vibrational energy, vibrational energy in some definite frequency bands and 50 maximal peaks in spectrum are suggested to be used for acoustic noise detection. A BP neural network model is designed to train valuation procedure and to determine threshold of noise detection, in which the difference of various type motors are taken into account. With the help of combination of Neural Networks Toolboxes provided by MATLAB and Lab VIEW, an online system for detecting acoustic noise of motors is developed based on Lab VIEW, which can be used to pick up the abnormal noise of motors on the manufactory line.

关 键 词：电机噪声 电机振动 声检测 直流电机 车用 在线检测技术 峰值 神经网络工具箱 LABVIEW平台 虚拟仪器 

分 类 号：TU311.3[建筑科学—结构工程] TM343[电气工程—电机]