Research On Intelligent Real-time System For Operation Guiding In Sintering Process

The computer control has been a major sign of sintering modernization, it plays a important role in increasing the productivity of sintering plant. obtaining good quality~ reducing energy consumption and lengthen the service life of equipment. At present. there are more than 300 sets of sintering plai~t in our country with their total sintering area beyond 18200m2. However, low automation level is a normal problem and has been a major fact that limit the improving of sinter quality and quantity. The process detection and basic control equipment are equipped in large and middle scale sintering plants in our country up to now. It is the next assignment that investigating the method of total control of sintering process and develop the 損rocess control?system.Sintering process is a dynamic system with long circuit. multivariable and complex mechanism. it is hard to perform control task of total sintering process by using conventional control theory and methods including modem control and classical control. In recent years. artificial intelligence theory such as expert system. fuzzy control and artificial neural network has been induced to sintering process control. which provide an efficient approach to realize the computer control in sintering process.In this paper. the mechanism arid technique of sintering process control are investigated by comprehensive utilization of sintering theory. modern control theory and artificial intelligence. An intelligent guide system for controlling sintering process is developed, which earns theoretical significance and practical value.After an analysis of the characteristics of sintering process. the basic control scheme including long-time control and short-time control is suggested. The complicated control of total sintering process is divided into six subsystems and each of them perform a control task. A main system perform data collection and general operation mode recognition task, which assign and coordinate subsystems meanwhile. This paper emphasis on the main system. the burning through point control subsystem and the subsystem for abnormity diagnosing in sintering process.The four state categories such as productivity, quality of sinter. permeability of sintering process and heat-pattern are used to judge the operation mode of sintering process based on the analysis of sintering mechanism, detected variables for judging each state category are chosen depend on experience of sintering experts and detection situation. The mathematics model for recognizing the operation mode of sintering process is developed and combined with knowledge-based model in order to realize the operation mode recognition of sintering process.The state of burning through point (BTP) is the mainly dependence for operating on-strand sinter plant. so it is a key middle variable of sintering process. The strategies for controlling the BTP are studied deeply in this paper including judging on-line strategy and prediction strategy, which offers the solution of long time delay and hard-detection of BTP. The adaptive prediction model for BTP based on time-series system identification is established?Ill .wa~Jf~and used as a tool in comparing three kinds of heat variable, such as the temperature of normal inflexion, the rising point of waste gas temperature curve and the area of waste gas temperature curve. As the result, the rising point of waste gas temperature curve is chosen as the predictive variable of BTP. The adaptive prediction of burning through point based on artificial neural network (ANN) is investigated~ the back-propagation algorithm of ANN is modified and the self-organizing algorithm is put forward, the universal limit new-information on-line training method is proposed in order to meet the need of real-time training of ANN. The adaptive predictor of BTP based on ANN is designed and makes accurate prediction of BTP. The fuzzy control strategy of BTP is proposed according to experience of experts in sintering field. The universal on-line optimal method on fu...