Optimal Setting Method For Raw Meal Grinding Process Of Vertical Mill Based On Case-Based Reasoning

Grinding process of raw meal with vertical mill is an important step in the new dry cement production. At the present stage, there is no more accurate model in the grinding process, and raw meal fineness still need offline manual check. At the same time, the key control variables of vertical mill are often set manually by the operator, and cannot be timely adjusted with the changes of working conditions, with the result that it cannot timely make the production process index reach to the expected value. Therefore, modeling for grinding process of raw meal with vertical mill and timely optimize the set values toward the dynamic working conditions, which is of great significance for improving cement raw meal quality, guaranteeing the stable equipment operation, and reducing the unit power consumption for production.In this paper, we carry out the research from two aspects of modeling and optimization setting. This paper first discusses the research status of modeling and optimize of raw meal grinding process, and introduces the research and application status of case-based reasoning and particle swarm optimization algorithm. Second, we determine the subsequent key variables needing modeling and optimization based on analyzing in details the specific technical process and parameter requirements for grinding process. It introduces the structure of wavelet neural network (WNN) and learning algorithm, and by using the data collected from the cement plant, it builds index prediction models of grinding process separately based on WNN and BP network, and compares the performance of these two models. Third, by combining case-based reasoning and particle swarm optimization algorithm, it proposes an intelligent optimizing setting method of key control variables of grinding process. It uses particle swarm optimization algorithm to optimize the original case data and establishes typical case database. When a new working condition occurs, by searching and reusing the cases, it determines the set values for the current working condition, and loads the set values to WNN prediction model to verify if the production index meets the objective expectation. And according to the difference between expected value and predicted value, it corrects the set values by utilizing expert rules, and makes the corrected set values as the final set values for final grinding process variables. Finally, it combines the intelligent setting method with monitoring software, to guide the manual setting using Lab VIEW to design the optimal setting software..Through experiment simulation, it verifies that WNN-based index prediction model, compared with BP neural network model, has a higher rate of convergence and higher generalization accuracy. By using the intelligent optimized setting method that combines case-based reasoning and particle swarm optimization algorithm, it not only can optimize the case data, but also can timely give the optimized set values of variables toward dynamic working conditions, so as to avoid the lagged and subjective adjustments of set values because of the dynamic working conditions, and it also has a role of guiding production operation of cement raw meal, and has a vital significance for guaranteeing the stable equipment operation, enhancing productivity, and reducing power consumption.