Study On Distributed Fault Diagnosis System Of Large-scale Complex Electromechanical Equipment

The automation level of modern electromechanical system is increasingly improved, the scale of system is gradually expanded, system structure is more and more complicated, and correlation degree of subsystems is becoming closer and closer. Thus, the possibility of fault occurrence is greatly increased, fault is shown in various forms, and a fault source may cause chain reaction, resulting in bigger fault. These traits make it quite difficult to carry out system fault diagnosis. On the other hand, to make large-scale complex electromechanical equipment run safely and stably, to predict fault in early time, to carry out prediction maintenance management, and to reduce financial loss brought by fault and maintenance become the goal of modern enterprises. Thus, to find proper monitoring and fault diagnosis method becomes an important research direction for science and technology researchers.For Large-scale complex electromechanical equipment, traits like diversity, uncertainty, progressiveness and concurrency of its fault make its fault diagnosis a quite complicated systematic project. Using independent diagnosis systems can not satisfy the diagnosis need of actual equipment, especially the distributed complex electromechanical system composed of many pieces of equipment. Nowadays, with the development of distributed artificial intelligence, system based on multi-agent provides a good way to design and realize large-scale complex diagnosis system, which can develop in the direction of being intelligent and network-based.A multi-agent distributed fault diagnosis problem-solution model FMAS was proposed, the model had the functions of multi-agent information passing, agent cooperation and intelligent diagnosis. In the model, every independent agent which has diagnosis function could take on the task of local diagnosis in parallel, and in the meantime, exchange related information with each other, thus agents could check and correct information and cooperate together to finish overall diagnosis task like human experts. The role and function of each agent in multi-agent structure were explained in detail, deployment method in actual application was discussed, and diagnosis object oriented deployment plan and method were given.In order to share and exchange diagnosis knowledge and related information, multi-agent diagnosis network structure and coordination method were studied aiming at uncertainty in the process of multi-agent negotiation. In multi-agent system, coordination can not only help improve the performance of individual agent and the whole multi-agent system, but also help enhance the problem-solving ability of agents and the flexibility of the system. The contract net protocol was used to carry out the coordination of agents, and improvement was done aiming at the weakness of traditional contract net protocol. An improved dynamic contract net protocol P-CNP was used to carry out coordination in state of multiple tasks being given in parallel. P-CNP was used to ensure the correspondence among different agents, so the timeliness of diagnosis system was guaranteed.Simulation results and experiment analysis could prove the effectiveness of the method proposed.With respect to diagnosis method, multi-agent data fusion was studied aiming at the conflict of multi-agent information, a multi-tier time-space domain D-S evidence theory data fusion diagnosis model (MT-TS-DS) was proposed, and agent technology was used to realize distributed fault diagnosis. The method divided the process of fault diagnosis into two layers:local time domain diagnosis result fusion and global space domain diagnosis result fusion. As for the same diagnosis object, time domain fusion can avoid diagnosis uncertainty caused by being monitored at different time. With space domain fusion, every monitoring point could refer to diagnosis results with other points to enhance the reliability of final diagnosis result. Different feature extraction methods (time domain analysis, wavelet analysis, etc.) and a number of intelligent fault diagnosis methods (BP neural network, RBF neural network and SVM, etc.) could be used at different monitoring points, or same monitoring point at different time, this also avoid diagnosis deviation caused by using a single method. The use of multi-agent data fusion technology made the distributed fault diagnosis system possess the quality of being intelligent, strong and fault-tolerant.The fault diagnosis of large-scale electromechanical equipment of a coal preparation plant was studied. Multi-agent fault diagnosis structure and fault diagnosis methods proposed was used to design and realize electromechanical equipment diagnosis system based on FMAS in the coal preparation plant. Every step of the system design and implementation was described in detail, the system software implementation method was analyzed in detail, sensor arrangement, system topology structure, software function were described, and with a concrete example, how to use this system to carry out distributed intelligent fault diagnosis was explained.