论文部分内容阅读
采用三层反向传播(BP)人工神经网络(ANN)技术建立钢淬透性预测预报的数学模型,基于Visual Basic编程语言建立了中碳结构钢淬透性的BP-ANN预测预报系统。将其对45、35Mn2、48CrMnMo、30CrMnNiMo钢的预报结果与近年来发展的改进型非线性数学模型所得结果以及对应的试验数据进行了比较研究,发现BP-ANN预测预报系统可以根据钢的化学成分较准确地预测钢的端淬试验各个距离点处的硬度值,但是各点处预报结果之间相不关联,且难以与钢的端淬显微组织建立有机联系;而改进型非线性数学模型可以较好地预报整条端淬曲线,辅以梅尼尔(Maynier)模型还可对各点处对应的显微组织进行预测预报。将两种模型结合使用来对钢材的淬透性进行预测预报,则有望进一步提高预测预报结果的可靠性。
The mathematic model of steel hardenability prediction was established by using BP artificial neural network (ANN) technology. Based on Visual Basic programming language, a BP-ANN prediction and forecasting system of medium carbon steel hardenability was established. The results of the prediction of 45,35Mn2,48CrMnMo, 30CrMnNiMo steel and the improved nonlinear mathematical model developed in recent years are compared with the corresponding experimental data. It is found that the BP-ANN prediction system can be based on the chemical composition of steel The hardness values at different distance points of steel quenching test are more accurately predicted, but the results at each point are not correlated with each other, and it is difficult to establish the organic connection with the quenched microstructure of steel. However, the improved nonlinear mathematical model Which can predict the whole end-quenching curve better. The Maynier model can also forecast the corresponding microstructure at each point. Combining the two models to predict and predict the hardenability of steel is expected to further improve the reliability of the forecast results.