基于DNN的中厚板组织性能逆向优化系统及应用

产品性能是热轧钢材生产的重要指标。生产工艺参数的调整和新产品的研发都需要较长的调试周期，容易造成产品性能的不稳定、研发成本过高等问题。为解决上述问题，进一步优化工艺，缩短研发周期，基于深度神经网络和规则期望算法，建立了中厚板组织性能逆向优化模型，对神经网络框架进行了选型以及超参数调参。基于某钢厂中厚板生产线在线生产数据，使用深度神经网络模型对最终产品性能进行了测试及应用，预测值与实测值的吻合度较高。

Medium Section Mill Reformation with Roller Bearing

Product performance is an important indicator of hot rolled steel production.The adjustment of production process parameters and the development of new products require long debugging cycles,which are likely to cause problems such as unstable product performance and high R&D costs.In order to solve the above problems,further optimize the process and shorten the development cycle,based on the deep neural network and the rule expectation model,a reverse optimization model of the plate structure performance was established,the neural network framework was optimized,and the algorithm was selected.Based on the online production data of a steel plate production line in a factory,the deep neural network model was tested, and the final product properties were predicted.The predicted values were good agreement with the measured values.