Fault Diagnosis Method Of Motor Based On Noise Detection

Since the advent of generators and motors in nineteenth Century and we can easily use the electrical energy. the motor has brought us everfount power,freeing people from the heavy labor. Use of the motor is very extensive, from steel mill to pump water plant and air conditioners, washing machines and to people’s daily wear watches are driven by the motor. With the rapid economic development, more and more appliances go into the general population of families. Like air conditioners, fans and washing machines is becoming everyday household items. Only air conditioners reached110million much in2013. The main driving force of air conditioning parts are the motor. The number of motor will be huge if coupled with washing machines and other household appliances. If the motor failure rate is high, it not only lead to huge economic losses, but also seriously affect people’s daily lives. Therefore, fault detection of the motor is very necessary.Traditionally, the motor fault is detected temperature, vibration and current of the motor. But it is hard to measure a huge number of motors quickly and accurately in the production line by using the traditional methods of detecting motor parameters. Therefore, the cuurent detection methods are generally to detect the noise by human themselves. Due to the sound absorption rate is low, the motor will radiate out sounds when it is running. The normal operation of the motor sounds different to the failure motor. Therefore, detecting characteristics of the sound of the motor is the most direct and easy. And because the motor failure rate is very low in the actual production process in current factory and it is difficult to collect enough samples of various types of failures. The general method of artificial intelligence based on the sample statistics and need multiple types of samples that represent all kinds of failure to train the net. It tend to have the ability to obtain sufficient Pan discriminate function. Therefore, it is difficult to applicant neural network algorithm and traditional pattern recognition under this condition.To address this problem, one class learning algorithms based on support vector machine is proposed. Use one class learning algorithm to build a judgment to detect the sound of motor by collecting enough normal samples of motor. It says that it only need normal samples. The motor is failure if it doesn’t meet all the characteristics of a normal sample after the motor is detected. It allows a small amount of samples in the failure motors that is detected. Therefore, this method is proposed based on different tone detection because it only need normal sounds motor samples.In this paper, use acoustic sensors to sample the motors in the axial position of the motor. Process the signal, extract features from the spectrum of the motor sound signal and build a judge function by using normal sound samples. Finally, use the various types of fault motor to test the discriminant function. Experiences shows that the method can solve this problem nicely.Finally, the paper is a summary of the work carried out in this article, and the future direction of development and research are discussed.