Application Research On Industrial Wireless Sensor Networks For Equipment Fault Diagnosis

To avoid economic loss and safety accident, condition monitoring and fault diagnosis are necessary for mechanical equipment. Compared to the traditional wired way, wireless sensor network eliminates expensive cables and saves the costs. But the early wireless way needs to pass all the raw data, which leads to the large wireless transmission energy loss. In order to reduce the amount of data transmission and lower the energy loss under the premise of ensuring diagnosis effect, on-sensor feature extraction using wavelet transform and on-sensor fault diagnosis using support vector machine are proposed in this paper, and only the diagnosis results will be transmitted to and displayed on the computer.The motor is taken as the research object in this paper and the whole simulation work of motor bearing fault diagnosis is completed on the computer. The program simulated on the computer will be then embedded in JN5139 and realize on-sensor fault diagnosis.Db4 wavelet function is used in simulation to extract the energy-torque feature vectors after a three-layer wavelet packet decomposition of motor normal baseline data, fan end bearing fault data and drive end bearing fault data. The extracted feature vectors are divided into training samples and test samples. The training samples are trained to establish the support vector machine classifier model. RBF kernel function is selected in training process and PSO algorithm is used in the process of cross validation to find the best penalty coefficient and kernel function parameter. Finally,the test samples are tested in the support vector machine classifier model to get the diagnosis results. The simulation results show that the energy-torque feature vectors extracted by wavelet transform can differentiate the motor running state effectively,and support vector machine has high reliability and accuracy in fault diagnosis.In the process of on-sensor implementation, fault diagnosis using support vector machine is conducted on the end node, and the diagnosis results are then transmitted to the computer via the coordinator node and displayed by a serial debugging assistant.Because only the diagnosis results are transmitted, it lowers the energy loss to a certain extent.