关键词: 人工神经元网络, BP算法, 遗传算法, 轧制力预测, 数学模型

Abstract: In the process of real production, the calculation results by conventional models have relatively big errors in the prediction of rolling force. It is therefore suggested to combine the BP network with modified GA algorithm to develop a prediction model, i.e., predicting the rolling force to be applied to medium thick plates by means of the guidable idea about search from BP network and the capability of global search from genetic algorithm. Moreover, the corresponding program and interface were given according to the prediction model. Based on the in-situ data taken from Handan and Puyang plate rolling mills, some optimal data was selected to predict the off-line rolling force. The results showed that the predicted accuracy resulting from the model developed is higher than that from conventional models, i.e., the relative error is within 4%, thus meeting the actual requirement for plate rolling.