Research On Audio Fault Diagnosis Method Of Mechatronic System Based On The EMD-ICA

With the advance of science and information technology, the modern industrial equipment gradually develops to the direction of integration, high speed and intelligence. In order to avoid unnecessary loss brought by the mechanical and electrical equipment, It must carry out accurate fault monitoring and diagnosis on mechanical and electrical equipment. Traditional fault diagnosis methods are mostly based on the approach of system internal sensing, but the way can’t get enough of the sensing information. So it has to be supplemented by external sensing. And it is one of the important ways with sound external sensing, therefore, it is particularly important to study fault diagnosis technology based on audio of the mechanical and electrical system.This paper is the research on fault diagnosis of audio signal of mechanical and electrical system. Traditional audio signal processing methods mostly adopt the way of filtering and noise reduction, aiming at the shortcomings of which, the paper uses the independent component analysis method of blind source separation. This method, only based on mixed signal acquired by sensors, can recover source signals generated by various mechanical components, thus can accurately carry through condition monitoring and fault diagnosis for each vibration source.Due to the ICA method has certain requirements for the number of monitoring signals, it puts forward the signal processing method based on EMD-ICA, according to combination the empirical mode decomposition(EMD) with ICA methods. This method not only avoids the modal aliasing phenomenon of the single EMD method, but also solves the problem of the limitations of the single ICA method, so as to realize the sound source separation of single channel monitoring signal. For the separated components, using kurtosis analysis, peak analysis and envelope spectrum analysis, it extracts the fault characteristic information, so as to realize the condition monitoring and fault diagnosis of mechanical and electrical system.According to the study of the above theory, this paper constructs the audio fault diagnosis system based on Labview platform, and uses the system to do the experiment about condition monitoring and fault diagnosis of drills state in the drilling process for normal drill bit and damage drill bit. The experimental results show that, both fault sound signal and normal sound signal, the recognition rate of fault diagnosis is around 95%, and it has reached the satisfactory result of diagnosis; And compared with separately using the EMD method, the recognition rate of system approach for fault diagnosis of the audio increased obviously. The above shows the feasibility of the audio fault diagnosis system based on EMD-ICA method.