Rearsh On Process Control Of Lithopone Rotary Kiln

The main purpose of rotary kiln automation is to stabilize the quality of the product. This paper focuses on several problems and solutions of a lithopone calcination process control, including critical signal preprocessing, multiple time-scale factor analysis of the process, nonstationary analysis of product quality series and process predictive model, considering the multiscale and nonstationary characteristics of the processFirstly, after studying the present state of modeling and control of calcination process of rotary kiln in domestic and foreign countries, reviews of the following four interactive problems are raised: temperature measurement and preprocessing, the division rule of hierarchical control system, nonstationary analysis of process signal and process monitoring modeling.Secondly, effects of the process variables on the product quality and difficulties in process control are analyzed based on the calcination processing mechanism, which provide a foundation of the subsequent research.Temperature is a key variable reflecting the variety of the process, which inevitably includes miscellaneous interference. The rotation of the kiln, the flickering of the flame, the barrier of the smoke and dust, the defective contact of the metal ring, and other unsuspected elements result in a confusing judgement of the process operation and a disadvantage to modeling and control. This paper reveals the internal principles of the temperature series of the lithopone rotary kiln and the barium sulphide rotary kiln respectively after spectrum analysis, according to their different temperature measuring mechanism. Causes of the unexpected perturbation are found attributed to the rotation of the kiln, and appropriate filtering algorithms are adopted to remove disturbances of this type.Considering the multiple variable and multiple time-scale characteristics, some key variables such as hot-end temperature and rotary speed are decomposed by empirical mode decomposition(EMD) to eliminate the coupling effect of variables under different time-scales. Then the series under different scales are analyzed using factor analysis method, whose result together with field study indicates calcining zone temperature and rotary speed should be chosen as the primary variables to establish a control model。Then an exponential function and a hyperbolic function are used for curve fitting respectively. Furthermore, another control model of simplified hyperbolic function unveils the implied purpose of the manual operation. Based on this conclusion and the result of multiscale factor analysis, the process control system is divided into three independent control loop: the hot-end temperature control loop, the calcination control loop and the feed-rate control loop. In the second one, the immeasurability of the product quality is settled with the help of a disclosure of the implicit goal under manual control. A simple, reliable and effective control algorithm of the calcination process is proposed, and it is proved feasible by a suitable statistical method.The calcining zone temperature and the product quality are improved evidently in terms of stabilization after the multiple independent loop control. But there are still some kinds of wide range fluctuations occurring in both of them, whose frequency and amplitude are time-varying with low-frequency and nonstationary features. At the first stage Fourier method is carried out on the product quality series for the purpose of a thorough understanding of the reason and an expected improvement of the control system in use. Considering the ineffectiveness of FFT when applyed to nonstationary and nonlinear series, although it finds the periodicity, HHT method, which is capable of describing the energy distribution in both time domain and frequency domain, is implemented at the second stage. Not only the accurate period is calculated, but also the time spot of the undulatory motion is allocated, which is of benefit to further comprehension of the fluctuation. The analysis uncovers some problems in management and process technique. On the other hand, it assists data filtration in the process monitoring modeling.At last, efforts are put on the process monitoring predictive modeling, to solve the immeasurability of the product quality. The result which comes out from HHT analysis gives a heuristic hint in selecting data for a data driven model. An achromic power predictive model based on least square support vector machines(LS-SVM), which is simple-structured and quick learning within required accuracy, is obtained by historical data. After analyzing the relation between calcination condition and energy alteration of the kiln's mass, the unit energy is introduced as input variable of the model based on gray box modeling principle, which results in a simplified model and a fast learning speed. A simulation result shows the effectiveness of the presented method.