Soft-Sensoring, Modeling And Control Of Fill Level In Ball Mill Pulverizing System

As one of important parts of pulverizing system, ball mill is used widely in coal-fired power plants. However, most pulverizing system can not be operated economically. One main reason is that fill level in ball mill is hardly measured precisely. Without an effective and precise method of measurement, operators have to make ball mill operated in low fill level and not to consider about the economy problem, in case of the accidents of heaped load and coal powder running over. Therefore, an effective ball mill fill level measurement method is pratically valuable.In order to solve the problems above, soft-sensoring, dynamic process modeling and control of fill level in ball mill of pulverizing system are studied in this paper. The main contents of this dissertation are as follows:1. Summarized current status of instrument and control of fill level in ball mill of pulverizing system. And it is put forward that using of soft-sensoring technology is a inevitable trend of fill level measurement.2. Studied and analyzed dynamic model of ball mill, and summarized function relationships between operating parameters, especially the relationships between fill level and other parameters. These lay a solid foundation of independent variables selections and fill level sample obtaining.3. Introduced theory and project appliance of soft-sensoring. Summarized and compared various methods applied into soft-sensoring technology. And also put forward that non-linear PLS algorithm is more appropriate to be used for fill level soft-sensoring modeling.4. Analyzed and studied pretreating methods of input datas in soft-sensoring model. And wavelet analyse is used to eliminate noises of ball mill bearing vibration signals. Before building up fill level soft-sensoring model, independent variables are selected by prediction error sum of squares. When soft-sensoring model is built up by non-linear PLS algorithm, function chain neutral network is used to correct fill level measurement model.5. Based on the achievement of fill level soft-sensoring, system identification is used to build dynamic models of balance area and non-balance area. Accordingly, models switch strategy and prediction control method based on two models are put forward.Based on data acquisition of field experiments, this paper achieve the measurement of fill level using soft-sensoring technology, and build up corresponding dynamic models and control system. These are of great importance and value to economical operation of pulverizing system.