Research On Soft Sensing Of The Overflow Particle Size For Grinding Process

Mineral processing is an important part of metallurgical industry. With our country's economic development and becoming the center of the world's processing manufacturing industry, the metallurgical industry is developed rapidly. As a result, the demand of mineral processing's products is increasing. To promote grinding process's development, many kinds of instrumentation and automatic control system have been widely used in this field. Grinding process is considered as an important part of mineral processing. However, the main measure of its products'quality----overflow particle size, cannot be measured on-line. This greatly reduces the quality of grinding process, impact the mineral processing's production efficiency, and become a bottleneck of the mineral processing. Achieving on-line measurement of the grinding process's overflow particle size is practical, which can bring great economic benefits by increasing productivity as well as saving costs.Based on the 1# grinding plant of Jiangxi Copper Corporation this thesis build an overflow particle size soft sensing mode of grinding process, and design the overflow particle size on-line measurement system. The main content including:To overcome the problem that it is hard to online measure the overflow particle size, this thesis studies on soft sensing, using regression analysis method. The instrumental variables are chosen and the synchronization of them is found based on study of the main equipment of grinding process and the actual situation of the production process. As a result, the inputs of the soft sensing model are the instrumental variables at the exact time. Based on the analysis of algorithm principle, this thesis establishes the grinding process model by Partial Least Square (PLS), gauss kernel PLS and algorithm principle kernel PLS. And the final algorithm of establish the overflow particle size soft sensing model of grinding process is PLS, which is determined by the analysis and comparison of the simulation results. Considering that the situation of the production process would have some changes as time passed by, and the factors that omitted when chose the instrumental variables, an available solution of online updating the soft sensing model is presented, which is on the basis of recursive PLS. the simulation results of the online updating model are shown in this thesis.Finally, an implementation plan of soft sensing of grinding process's overflow particle size is designed, which is the integration of the whole thesis. Taking the actual hardware and software structure into account, this thesis detailedly introduces the soft sensing prediction system, the human-computer interaction system and the PLC control system, and meanwhile presents their interrelationship.