Research On Soft Sensor For Apparent Degree Of Kiln Inlet Calcination In NSP Cement Production

At present, New Suspension Preheater (NSP) cement production technology has become the developing mainstream and most advanced process in cement industry. The precalcining technology is the core technology in NSP cement production. The calciner is a new heat source between suspension preheater and rotary kiln. Fuel combustion, exothermic process, raw decomposition and endothermic process are carried out at the same time, and make the apparent degree of calcination of kiln inlet reach 85% ~95%, so that it reduces the heat load of firing zone of the rotary kiln and greatly improve the production efficiency of the kiln system. So the apparent degree of calcination becomes the important target for controlling the calciner and kiln.Currently, most cement plants adopt off-line detection of detecting the apparent degree of calcination. So the reference value on controlling the calciner and kiln is decreasing. So it is of great significance for studying on on-line measurement of the apparent degree of calcination in NSP cement production. With the rapid development of the production technology, for ensuring the safety and efficient operation of the production equipments, the important process variables, which are closely related to the stability of the system and the product quality, must be detected and controlled. So the soft sensor technology becomes the effective way to solve the problem of process variables real-time detection.This paper relies on the sub-topic of National High-tech Research Development Plan (863): Soft sensor, unit-level optimization control system design, development and engineering applications (NSP cement production process integrated control applications) issues. It researched on soft sensor for apparent degree of calcination in NSP cement production.Based on in-depth analysis of NSP cement production process, the furnace temperature and pressure of the calciner, the outlet temperature and pressure of the calciner, the temperature of the tertiary air and the current of the raw hoister are selected as assistant variables. 3σCriterion is used to remove gross error. And standardization of assistant variables is accomplished. Because the calciner is strong coupling, great delay, nonlinear, multi-factors and time-varying, it is difficult to establish exact mathematical model of calciner decomposition process. So after analyzing and comparing the modeling methods, BP neural network is chosen as the modeling method. The results of the experiment show that the topology and type of the network have some influence on modeling, and the over-fitting phenomenon appears. Then a soft sensor model based on Least Squares Support Vector Machine (LS-SVM) for the apparent degree of calcination is established. The results of the experiments show that the model has favorable learning ability, generalization performance, lower dependent degree of data samples, and satisfactory prediction accuracy.In this paper, the implementation scheme of the soft sensor of the apparent degree of calcination which is based on the OPC technology, ADO technology and the Control Build F system of ABB's Freelance 2000, is presented. And the online LS-SVM algorithm is given. Soft sensor of the apparent degree of calcination which is used in product is developed with Visual C++6.0, and the application effect is good.