基于有限元的三辊行星轧制力预测及分析

为预测三辊行星轧制力，建立了三维热力耦合有限元模型，用均匀设计法安排有限元模拟样本，采用基于LM算法的BP神经网络学习有限元轧制力结果，确定了轧制参数与轧制力的映射关系，并根据训练后的神经网络分析了摩擦系数、轧辊偏转角和轧辊转速对轧制力影响．预测结果表明：摩擦系数和轧辊偏转角对轧制力影响是多方面的，在高轧辊转速时，较大的摩擦系数有利于降低轧制力．

Rolling force prediction and analysis of three-roll planetary rolling process based on FEM

To predict the rolling force of three-roll planetary rolling process, 3 D thermo-mechanical coupled finite element model was built up. Uniform design method was used to arrange training samples of finite element simulations, and the simulation results were trained in BP neural network based on LM algorithm. The mapping relationship between process parameters and rolling force was established to predict rolling force at realtime. The effects of friction coefficient, offset angle and rotational speed of rollers on rolling force were analyzed using the trained neural network. The predicted results show that the effects of friction coefficient and offset angle on rolling force are multiple, and larger friction coefficient is useful to reduce rolling force when the rotational speed of the rollers is high.