机器学习辅助高强铝合金挤压力和制品出模口温度快速预测与调控

Machine learning assisted prediction and regulation of extrusion force and exit temperature of high-strength aluminum alloy

挤压力和挤压制品出模口温度是影响铝合金挤压成本和制品质量的关键参数.本文基于有限元数值模拟数据,采用人工神经网络建模,研究了不同挤压工艺参数(坯料温度、工模具温度、挤压速度)下高强铝合金挤压力和制品出模口温度的变化规律,实现了挤压力-挤压行程曲线和制品出模口温度-挤压行程曲线的准确预测,预测误差分别为2.84%和1.30%.采用支持向量回归算法和输入变量筛选,实现了挤压过程峰值挤压力、制品出模口最高温度和最低温度的准确预测,预测误差分别为0.82%、1.18%和1.88%.基于上述预测模型,可在给定制品出模口目标温度时,采用遗传算法快速确定合适的挤压工艺参数,或者在坯料和工模具温度发生波动时快速确定合适的挤压速度.

Extrusion force and exit temperature are the key parameters that affect the ex-trusion cost and product quality of aluminum alloys.Based on the finite element numerical simulation data and artificial neural network modeling,this paper studied the evolution rule of extrusion force and exit temperature of high-strength aluminum alloy under different extru-sion process parameters(billet temperature,die temperature,extrusion speed),and realized the accurate prediction of extrusion force-displacement curves and exit temperature-displacement curves.The prediction errors were 2.84%and 1.30%,respectively.The sup-port vector regression algorithm and input variable screening were used to accurately predict the peak extrusion force and the maximum and minimum exit temperature of the extrusion process,and the prediction errors were 0.82%,1.18%and 1.88%,respectively.Based on the prediction model,the genetic algorithm could be used to quickly determine the appro-priate extrusion process parameters when the target exit temperature was given,or the appro-priate extrusion speed could be quickly determined when the temperature of the billet and the die fluctuated.