Fuzzy Modeling Of Grinding Process Based On Fuzzy Sets Merging And Rule Simplification

The industrial grinding process is complex and there are many control variables in the production line. The Strongly nonlinear dynamic relationship between different variables makes it difficult to describe the grinding process with an accurate mathematical model, so it is hardly to control it with general control theory. Most plants still need operator to control the grinding process by their experience in time. But it is difficult to achieve the expected target because of the limited subjective experience, complex working condition and the variety of the boundary conditions. Modeling the grinding process effectively is helpful to the realization of the automatic production of the grinding process, and refrains from the fault operation by manual operate. It can also improve the utilization of mineral resources and reduce the cost of production.This research is on the background of the complex grinding system in the mineral process. To the problem of the typical grinding process modeling, a fuzzy modeling method based on fuzzy sets merging and rule simplification is proposed in this paper. First, modeling the grinding process by a data-driven method to form an initial fuzzy rule base. Then, aiming at the fuzzy rule extraction process of the initial Takagi-Sugeno model, the proposed method adopts Fuzzy C-Mean (FCM) clustering to partition the fuzzy membership function of every variable. Then, the parameters of a new membership function are calculated in order to represent for different working conditions and to reduce the negative impact of the overfitting phenomenon. As such, using the similarity of each redundancy rule, the fuzzy rule base is further simplified by merging the fuzzy rules with the same premises and the final fuzzy model with better generalization ability is obtained.To verify the validity of the proposed approach, choose several other methods to perform the comparative experiments. A series of comparative experiments are carried out by using classic data and industrial data. The experimental results demonstrate that the proposed method exhibits remarkable generalization ability and the practical application potential. At last, an intelligent control system using the proposed method for grinding process is developed for the industrial manufacture.