Research And Development On On-line Monitoring And Fault Diagnosis System Of Sinter Machine Based On Genetic Neural Networks

With the rapid development of science and technology, the machinery and equipments for modern industrial production are increasingly featured by large size, complexity, high operation speed, automation and high power. Though we have achieved improved production efficiency, we suffer a lot from the increasing equipment maintenance costs and maintenance difficulty. The traditional concept of"Breakdown maintenance"or"Preventive maintenance"has already been outdated and can not meet the needs of enterprises. Mechanical equipments, especially the fault diagnosis technology for important devices,have been attached greater and greater importance. Taken into account of the safety production and maintenance costs, the"preventive maintenance"based on the state monitoring, is becoming more and more popular in the production industry. In particular, with the development of artificial intelligence, pattern recognition, and other theories, the technology of intelligent diagnosis has become a new research field of fault diagnosis.This thesis makes an in-depth study on the intelligent diagnostic technology at home and abroad, especially the BP neural network fault diagnosis technology, which is quite popular recently. This thesis covers the details of neural network theory on its development, models, algorithms, and genetic algorithm theory. However, BP network has its weaknesses. For example, the order of the samples has certain impact on the modeling results. And there may be slow rate of training, training fitter problems, easy convergence on the local minimum point, and the algorithm without convergence. To offset the limitations and weaknesses of BP network, the author put forward the following measures: by using the genetic algorithms to the overall search of BP network's training weights and thresholds, we make an integration of the genetic algorithms and BP algorithm so as to complement each other. This thesis adopts the MATLAB to test the feasibility of faults samples'simulation algorithm. Finally, for smoking sintering machine in Ironworks, the author develops a set base on BP network to work as the monitoring and fault diagnosis system. Besides monitoring the sinter machine, The system can also realize the real-time data acquisition processing and analysis, configuration parameters setting intervals, historical trend analysis, intelligent fault diagnosis, and other functions. By using human-computer exchange interface, this system is friendly, easy to manipulate, and accurate in its forecasts. This thesis introduces the composition of hardware system, which includes sensors selection, acquisition card network, on-site wiring and the realization of software functions, and pays great importance on the introduction of intelligent diagnosis model's offline training and online diagnostic function.