基于主成分分析模型和BP神经网络模型在钢水“脱氧合金化”配料方案优化中的应用和研究

Application and Research of Principal Component Analysis Model and BP Neural Network Model in Optimization of “Deoxidation Alloying” Batching Scheme for Molten Steel

摘要: 钢铁和一个国家的工业发展息息相关，随着制造2025的到来，对钢铁的要求更是进一步提高，因此提出一种钢水“脱氧合金化”配料方案对工业制造和经济社会的发展具有十分重要的意义。本文综合考虑了钢水配料中C、Mn两种元素的历史收得率与其他影响因素之间的关系，对历史数据进行了预处理，建立了主成分分析模型综合分析影响两种元素收得率的主要因素；然后构建BP神经网络，对C、Mn两种元素的收得率进行预测，并在此基础上利用分数阶对BP网络进行改进，优化预测结果；最后利用动态规划的思想求出了最优成本和配料方案。最终结合本文研究结果制定了配料方案的优化的建议与策略。

Abstract: Steel is closely related to the industrial development of a country. With the advent of Manufacturing 2025, the demand for steel is further improved. Therefore, it is very important to propose a “deoxidizing alloying” solution for molten steel, which is very important for the development of industrial manufacturing and economic society. In this paper, the relationship between the historical yield of C and Mn in molten steel and other influencing factors is considered. The historical data is preprocessed, and the principal component analysis model is established to analyze the impact of the two elements. The main factors are then constructed BP neural network to predict the yield of C and Mn elements, and then improve the prediction results by using fractional order to improve the prediction results. Finally, use the idea of dynamic programming to find out Optimal cost and ingredient planning. Finally, combined with the research results of this paper, the recommendations and strategies for optimizing the ingredients are formulated.