Fault diagnosis method of wireless sensor based on convolutional neural network

Abstract

The present invention relates to a wireless sensor fault diagnosis method based on a convolutional neural network. The steps are as follows: 1) Construct a wireless sensor network system model composed of a base station, a mobile car and several rechargeable wireless sensor nodes; 2) Construct a convolutional neural network sequentially connected to the input layer, hidden layer, fully connected layer and output layer at the base station; 3) Use the mobile car to collect the sensory data of all nodes and transmit it to the base station for storage, and convert it into a matrix form; 4 ) Input the perceptual data in the form of matrix into the convolutional neural network for training and self-learning, extract data features through the convolutional layer convolution kernel, compress the data features in the pooling layer, connect the last two layers in the fully connected layer, and output the layer Output the final data classification results; 5) According to the data classification results, the convolutional neural network performs fault diagnosis for each category corresponding to the corresponding sensor fault type, and outputs the node diagnosis status through the output layer.

Description

technical field [0001] The invention relates to the technical field of wireless sensors, in particular to a convolutional neural network-based wireless sensor fault diagnosis method. Background technique [0002] A wireless sensor network is composed of several wireless sensor nodes in the form of self-organization. A wireless sensor node is mainly composed of four parts: sensor module, CPU module, wireless communication module and power module. Among them, the sensor module is mainly used to sense data, and the CPU module The role of the sensor is to process and calculate data, the wireless communication module ensures that the sensor node communicates with other sensor nodes, and the power supply module carries limited energy to provide energy for the sensor node. Due to the small size, easy deployment, and low price of wireless sensor nodes, they are widely used in all aspects of life. With the enhancement of wireless sensor computing power and storage capacity, wireless ...

AI Technical Summary

Problems solved by technology

[0005] (2) Hard failure node: the node cannot communicate, so the sensing data of the node cannot be received;

Although the distributed fault diagnosis method alleviates the problem of excessive energy consumption of network nodes, the algorithm has high requirements on the neighbor nodes. Accurate fault detection for nodes in wireless sensor networks with high rate

Images