Research On Pridiction And Control Of Burning Through Point Based On Support Vector Machines In The Sintering Process

In the sintering production, Burning Through Point (BTP) is a very important technical parameter related to sinter quality, quantity and cost, and which is the key point for blast furnace to get high grade technical target. The right position and keeping stabilization of BTP can improve product ratio and use sinter area completely, and maximizes the sinter production rate, and decrease the energy cost. But because of the long-time delays and dynamic complexity of sintering process and the parameters for judge and analyze the BTP can not be measured directly, it is difficult to solve the BTP control problem. Therefore, this problem has been considered as a key difficulty of steel enterprise automation in a very long time.According to the particular sintering production of Laiwu Iron and Steel Corporation, the dissertation pays more attention to the prediction and control of BTP for the improvement of sintering quality. The following aspects are investigated in the dissertation: the BTP prediction model and control method, on-line inference the sintering quality based on cross-section infrared thermal imaging of discharge end, optimization and control in the sintering process. These researches earn theoretical significance and practical value. The main creative works of the dissertation are as follows:(1) In the BTP prediction, a new prediction method based on support vector machines (SVMs) is presented for BTP. The long-time prediction model of BTP is constructed for Laigang No.2 265m2 Sinter machine. The results indicate SVMs outperform the three-layer Backpropagation (BP) neural network in predicting BTP with better generalization performance, and are satisfactory.(2) In the BTP control, the fuzzy control strategy is investigated for BTP. The short-time prediction model is applied to the BTP fuzzy control system. The BTP short-time prediction model is constructed via burn rising temperature(BRT) and burn rising position(BRP). The application results have verified the effectiveness of the control system.A predictive control algorithm based on least squares support vector machines (LS-SVMs) model for the sintering process with strong nonlinearity is presented. The nonlinear offline model is built via LS-SVMs. In the process of system operation, the offline model is linearized at each sampling instant, and the generalized predictive control (GPC) algorithm is employed to the controller design. The GPC controller is designed to adjust strand speed for sintering temperature control. The performance is demonstrated by several simulation results. The results show the effectiveness of the presented algorithm.(3) In the sintering quality, the reasonable feature parameters are obtained by processing the cross-section infrared thermal imaging of discharge end. The multi-class classifier based on SVMs is constructed to on-line inference the sintering quality (undersintering, oversintering and normal).(4) In the sintering process optimization and control, the parameters optimization model is constructed to guide the sintering process control. The optimum parameters are obtained by using matching optimization algorithm based on clustering method. The sintering production quality is improved obviously and the energy consumption is decreased.