论文部分内容阅读
为了充分发挥钢中析出相的晶内形核细晶作用,以便优化热模拟工艺,对合金钢热模拟试验组织中的析出相进行了分析及预测。研究了合金钢热模拟试验工艺参数,包括化学成分、变形温度、变形量以及保温时间等与析出相粒径、形态与分布之间关系,并建立了相应的映射关系,在此基础上建立了合金钢析出相分析及预测模型。应用L-M算法对该神经网络模型的权值进行了优化,从而克服了神经网络训练速度慢、容易陷入极小局域和全局搜索能力弱等缺点,提高了神经网络的预测精度。通过实例验证表明,改进后的神经网络对合金钢析出相粒径的预测精度达到93%以上,对合金钢析出相形态的预测精度达到90%以上。
In order to give full play to the intragranular nucleation and grain refinement of the precipitated phase in the steel so as to optimize the thermal simulation process, the precipitated phases in the hot simulation test of the alloy steel were analyzed and predicted. The parameters of thermal simulation experiment of alloy steel, including the chemical composition, deformation temperature, deformation amount, holding time and so on, and the relationship between particle size, morphology and distribution of the precipitated phase, and the corresponding mapping relationship were established. Based on this, Analysis of Alloy Steel Precipitates and Prediction Model. The L-M algorithm is used to optimize the weight of the neural network model, which overcomes the shortcomings of the neural network training speed is slow, easy to fall into a very small local area and the global search capability is weak, and improve the prediction accuracy of neural network. The experimental results show that the prediction accuracy of the improved neural network for the precipitated phase diameter of the alloyed steel reaches more than 93% and the prediction accuracy of the precipitated phase of the alloyed steel reaches more than 90%.