Research And Realization Of Data Fusion Algorithms Based On WSN Safety Monitoring Of Enterprise Manufacture

Many safety production accidents are happening in Chinese enterprises for a long time, such as fire, explosion, leakage, etc, which is closely related to the production equipment and environment. Nowadays, researchers in this filed are paying more and more attention to the monitoring of the equipment and the environment in the production process, especially the collection, processing and analysis of the equipment parameters and environment variables, hoping to prevent the accidents or take effective measures immediately after accidents.Based on the Production, Teaching and Research Cooperation Project of Guangdong Province, this paper investigates and analyzes the common cause of fire, explosion and leakage in the production process as well as the domestic and oversea researches and current application of related monitoring technologies. Internet of Things, Wireless Sensor Networks(WSN) and Multi-Sensor Data Fusion are involved in the course of the present research.Firstly, due to WSN’s strength in practical application, this paper analyzes and designs the hardware platform, communication protocol, target function, implementation scheme and monitoring mechanism of the monitoring system in detail according to the production environment and actual need of the enterprise. Secondly, this paper conducts a detailed study on WSN-based Multi-Sensor Data Fusion Technique and adopts D-S Theory of Evidence as the fusion algorithm, which can be applied under weaker conditions than Bayesian theory. And it also employs the concept of membership in the Theory of Cauchy Fuzzy Sets to solve the issue of assigning elementary probability in D-S Theory of Evidence. The decision-level data-fusion algorithm fuses the feature data collected by each sensor in the safety production monitoring system of the enterprise in order to make optimum judgment and decision on the in-situ production environment. Thirdly, in order to lower the correlation of the evidences and make better use of D-S Theory of Evidence for data fusion, this paper adopts the Back Propagation(BP) Neural Network in the feature level of data fusing and trains the membership degree of the raw data with respect to the target state and verifies its effectiveness through simulation experiment and specific numerical calculation. In addition, this paper makes improvement regarding to the shortcomings of the D-S Theory of Evidence and BP Neural Network. Meanwhile, it conducts simulation experiments, then analyzes and compares the results with the help of MATLAB experimental platform. Moreover, it integrates the improved D-S Theory of Evidence with BP Neural Network and sets up the Multi-Sensor Data Fusion model, thus making full use of their advantages to analyze the feature data collected by each sensor in the safety production monitoring system and make better judgment and decision. In the end, this paper summarizes the research work done by the author and puts forward some implications for future researches.