在现代钢铁冶金领域，钢水质量精准控制是高端钢材性能提升的核心瓶颈。国内某钢厂产品提升，对钢水的纯净度、成分均匀性的要求不断严苛，传统依赖经验的合金化与吹氩控制模式已难以满足需求。精炼过程中合金成分波动、精炼阶段氩气流量匹配失衡等问题常导致钢水质量不稳定、产品返工、降级等，本工作构建了工业视觉-数据-机理融合吹氩智能控制系统，通过动态协同机制实现工艺参数的一体化调控。工业实验表明：采用冶金机理-反向传播神经网络-改进单纯形机器学习，Q195钢种成分达标率提升至98.1%，合金成本降低5.3元/吨；采用工业视觉引导的自适应吹氩智能控制系统，吨钢氩气消耗降低9.9%，夹杂物等级在CT3.5以上的占比由35%降低至16%，年综合效益约228.32万元。

In the field of modern steel metallurgy, the precise control of molten steel quality constitutes a critical bottleneck constraining the performance enhancement of high-end steel products. As a domestic steel plant endeavors to upgrade its product quality, the increasingly stringent requirements for molten steel cleanliness and compositional homogeneity have rendered traditional, experience-reliant control modes for alloying and argon blowing inadequate to meet practical demands. Confronted with challenges such as alloy composition fluctuations and argon flow mismatches during the refining process—issues that frequently result in unstable molten steel quality, product rework, and grade downgrading—this study develops an intelligent argon blowing control system integrated with industrial vision, data analytics, and metallurgical mechanisms. Leveraging a dynamic coordination mechanism, the system achieves the integrated regulation of key process parameters. Industrial experiments show that: by employing a machine learning approach that combines metallurgical mechanisms, back propagation (BP) neural networks, and the improved simplex method, the compositional compliance rate for Q195 steel increases to 98.1%, while alloy costs were reduced by 5.3 CNY/ton; The industrial vision-guided adaptive intelligent argon blowing control system reduces argon consumption per ton of steel by 9.9%, decreases the proportion of inclusions with a grade of CT3.5 or higher from 35% to 16%, and yields an annual comprehensive benefit of approximately 2.2832 million CNY.