基于改进的RBF神经网络的热轧板冲击功预报模型

Prediction Model for Impact Energy of Hot Rolled Plate Based on Improved RBF

为了对Q345B中厚板的冲击吸收功进行预报并指导生产,采用RBF网络预测冲击吸收功与化学成分、工艺参数、屈服强度及抗拉强度等因素间的关系。针对RBF神经网络的不足,采用改进的遗传算法优化RBF网络结构和权值参数,并建立了基于改进遗传算法的RBF神经冲击功预测模型。实验结果表明,用这种改进的遗传算法优化后的神经网络算法是有效的、可行的。

In order to predict the impact energy of Q345B hot rolled plate, a RBF neural network was used to predict relation between impact energy and the factors of composition, process parameters and strength of the plate. The genetic algorithm was used to optimize the weights and structure of RBF neural network aiming at its shortcomings. The prediction model for impact energy of hot rolled plate based on improved RBF was developed. The experimental result shows that the neural network algorithm optimized by this improved genetic algorithm is effective and feasible.