Study on the Influence of Process Parameters in Cold-Rolled Strip Using Statistical and Machine Learning Methods

In the steel industry, the mechanical properties of cold-rolled strips are crucial for product quality and cost. This study aims to optimize the continuous annealing process parameters. Data preprocessing involved identifying and removing outliers using box-and-line plots and addressing missing values with a decision tree-filling method. Descriptive analysis was performed to determine the statistical characteristics of the process parameters, and normality was assessed using P-P plots, Q-Q plots, and histograms, revealing that the data did not follow a normal distribution. The Spearman correlation coefficient was used to examine parameter correlations, identifying those that significantly affect strip hardness based on their significance levels. A Decision Tree machine learning model was then employed to analyze feature importance, identifying six key parameters influencing strip hardness. The top three factors were the temperature of the fast-cooling furnace (29.1%), carbon content (25.1%), and the temperature of the quenching furnace. Notably, the importance of fast-cooling furnace temperature and carbon content was equal, while strip speed was the least influential factor, accounting for only 3.2%. This research provides essential insights and practical references for optimizing the production process and quality control of cold rolled strip steel.