A Self-organization Method For Sensor Networks In Condition Monitoring System

To adapt for the changes of advanced manufacturing mode, various kinds of on-line condition monitoring and fault diagnosis technologies have been developed recently. The E-diagnostic technique which based on the'self-organized flexible reconstruction'idea, has been put out to satisfy the need of rapid reorganization and dynamic configuration of modern manufacturing system. One of the key issues of this technique is the reconfigurable distributed technologies and methods of condition monitoring system, while its bottle-neck of realization is the flexible reorganization technology of the device layer. In the monitoring system, the study of self-organization algorithm of the substratum unit is the basis of the realization of flexible reorganization for the whole system, and its implementation method is helpful to realize self-organization of other units. In this article, with typical manufacturing equipment like NC machine and machining center as objects, based on the theories of information fusion and self-organization, the method of self-organized flexible reconstruction of the multi-sensor system were studied, which led to the realization of flexible reorganization of the device layer sensors.On the basis of E-diagnostic system, the structure and realization principle of the flexible reorganized condition monitoring system were analyzed, and it was pointed out that the key issue for the realization of self-organization of the monitoring system is the flexible reconstruction of the device layer sensors. Established in the sensor network of the device layer, the general method and three forms of information fusion were analyzed, and then the scheme of fusion on the data layer of sensor network was determined. And the primer-fusion of the sensor network was implemented by the SOM neural network.According to final-fusion, the information entropy and interrelating entropy of the sensor network were deeply analyzed, and then the gist for information fusion was proposed. According to the comparison among common information fusion methods, the method of ration fusion with Genetic Algorithm (GA) was determined, and then feasibility of the algorithm was proved by simulation analysis.With experimental study, self-organized optimization for the sensor network was implemented. The result proclaimed the feasibility of the algorithm.The method of multi-sensor self-organization discussed in this article can efficiently supervise the sensor network in the monitoring system, providing the basis for flexible reconfiguration of the device layer. Simulation and experimental results showed that abundant original information was reserved, while the cost on data transfer and storage were saved, reducing the operation time of the whole monitoring system. Self-organization algorithm has favorable characteristics of openness and easy-portability, facilitating its extension in engineering practices.