Design and Application of a Large-Scale Multi-Objective Dynamic Steel Allocation Optimization Model Integrating Thought Chains

Abstract

In view of the problems of traditional steel distribution relying on manual experience, poor dynamic adaptability and multi-objective optimization imbalance, a multi-objective dynamic steel distribution optimization model integrating the thinking chain is proposed.The model is based on the scrap steel image data set, expands the data through StyleGAN3 and Improved-Diffusion models, improves robustness by Swin-IR ultra-resolution processing, and adopts Swin-Transformer and SE Attention mechanism optimization YOLOv8 to build a scrap steel grading model;Fusion reinforcement learning and cross-modal feature adaptive fusion technology, establish a multi-objective decision-making system including cost, quality and energy consumption, and combine metallurgical process knowledge map and reaction dynamics model to achieve constraint optimization.Through the dialogue and interactive interface, the optimal steel scheme can be output by entering the target steel type. Experimental verification shows that the control accuracy of the steel composition of the model has been increased by more than 15%, the energy consumption has been reduced by 8%, and the production cost has been reduced by 10%, providing technical support for the intelligent and low-carbon transformation of the steel industry.