Data fusion of wireless sensor network for prognosis and diagnosis of mechanical systems

In order to improve the data collection efficiency, network lifetime and reduce transmission congestion of a Wireless Sensor Network, a data fusion algorithm based on the combination of BP Neural Network (NN) and Wavelet Packet is proposed. The research is based on Industrial Wireless Sensor Networks (IWSNs) application on mechanical diagnosis of wind turbine monitoring. This proposed algorithm uses the cluster protocol, and each cluster head is modeled with a three layers NN. The raw data feature is extracted by the Wavelet Packet and transmitted to the sink node for feature fusion. The fault classification result is used for mechanical diagnosis; the simulation results show that the diagnosis precision can achieve 90%.