Prediction Of Product Quality And Optimization Of Operations In The Continuous Production Process

In steel production,the steel production process is a large-scale industrial process with complex procedures and a combination of multiple control variables.The continuous annealing process,as a key process in the strip production process,plays a decisive role in the quality of the strip.However,due to the limitations of the annealing process itself,the original strip quality inspection methods are no longer able to meet the growing production needs.In addition,in order to further improve production efficiency and production quality,the optimization of operations for different steel products has gradually attracted attention.Therefore,it is necessary to predict and optimize the product quality in the continuous annealing process.In view of these two issues,this article has done the following researches.For the quality prediction of continuous annealing process,it is based on the analysis of the characteristics of continuous annealing data such as high dimensionality and strong coupling.This paper combines the isometric feature mapping dimensionality reduction algorithm with the integration idea,and uses the characteristics of the integration model to fit the dimensionality reduction results of different samples.After choosing support vector machine as the sub-learning machine,this paper proposes an integrated modeling method based on high-dimensional data characteristics.The method fully considers the characteristics of different samples,making the modeling more accurate.In addition,in view of the decrease in model accuracy caused by changes in equipment,environment and strip characteristics in the model,this paper studies the incremental learning based on the dimensionality reduction process on the basis of the incremental support vector learning machine.Use the support vector set and the newly added sample set to reduce the dimensionality,and filter through the distance-based strategy to complete the model update.This method is verified by experiments and is suitable for continuous annealing production changes.On the basis of quality prediction,this paper studies the continuous annealing production process and the optimization of the production process for cold rolling.After that,a multi-objective operation optimization model for the continuous production process was established with the goal of minimizing strip hardness deviation,minimizing unit energy consumption,and maximizing unit capacity.At the same time,in order to enable the algorithm to find the optimal solution earlier,the initial population needs to be of higher quality,and a heuristic based on normal distribution is designed to generate the initial population.Finally,the differential evolution algorithm DE(Differential Evolution)is used to solve the model,which has achieved better results than the original production data,and can play a guiding role in the production process.Because continuous annealing optimization has many operating variables and complex optimization,it is difficult to solve the continuous annealing multi-objective operation optimization model.In order to improve the optimization ability of the optimization algorithm,this paper proposes an improved mutation hybrid multiobjective optimization algorithm.The global search capability of NSGA-Ⅱ(Nondominated sorting genetic algorithm-Ⅱ)is enhanced by introducing a gene-hopping mutation operator,and the improved mutation operator is used to enhance the local refinement search capability of the DE algorithm.After the integration of the two algorithms,the mixed solution set is screened by the greedy algorithm,and chaotic search is performed near the optimal solution,and the optimal solution set is finally obtained.After the test function verification,the improved algorithm has a good optimization efficiency.Finally,the algorithm is applied to the continuous annealing multi-objective optimization model,and the solution result is greatly improved compared with the DE algorithm.This method improves production efficiency and reduces energy consumption while ensuring prediction accuracy.