Research On Remote Fault Diagnosis System Of NC Cutting Machine

Along with the continuous improvement of automation in light industry, the applications of NC Cutting machine become more and more popular, especially in the industry of clothing, shoes and hat, bag and box, leather, glass carving, and so on, which has greatly increased the efficiency of production in these companies. However, the general technician could not analyze and determine the causes of the failure of NC Cutting machine accurately. And the company should send professional technician to the scene to give them some guidance, which will certainly affect the efficiency of production, so it is of great value to establish the NC Cutting machine fault diagnosis system.At first, the paper analyzes the present state of development of remote fault diagnosis technology at home and abroad, and introduces the existing diagnostic techniques. The diagnostic framework model is established according to the characteristics of remote fault diagnosis system. After analyzing and comparing C/S (Client/Server) model and B/S (Browser/Server) model in detail, the system network model has been founded.And then, the fuzzy controlling theory and neural network theory are introduced in this paper. Due to the advantages and disadvantages of fuzzy controlling theory and neural network theory, both theories are combined to make up Fuzzy Neural Network. Based on Fuzzy Neural Network, the fault diagnostic technology will reflect the respective advantages of neural network and fuzzy controlling theory. It is a very promising research direction, and becoming more and more widely applied.Finally, the author introduces the structure and running principle of NC Cutting machine, and analyzes the machine's common failure as well as its solutions. According to the characteristics of failure of NC Cutting machine, fuzzification of the input signals, which reflects the fault characters, can reflect the randomness and uncertainty of the faults better, the knowledge acquired by neural network learning algorithm. The improved BP learning algorithm applied in this paper, when the error back-propagation, it is not only modifying the network weight value, but also modifying the compensating parameters in the fuzzy operator, so that algorithm has less chance to be trapped in local minimums, enhanced approximation accuracy of function, and strengthened the network mapping capabilities. The feasibility of Fuzzy Neural Network fault diagnostic was proved through simulation.