人工神经网络在带钢力学性能预测中的应用

Application of Artificial Neural Network Forecasting on Mechanical Performance of Strip

本文运用神经网络原理建立了带钢力学性能预测的神经网络模型,并用该模型对某钢厂带钢力学性能进行了预测,还对预测结果和真实数据进行了比较,结果表明:相对误差低于1%,而且收敛速度快,泛化能力好。预测模型可减少或取消实际生产中的破坏性检测试验,从而提高经济效益。另外,文中还总结了提高神经网络性能的方法。

A modal for strip mechanical performance forecasting was established, based on artificial neural network, Using this modal, a test of forecasting on strip mechanical performance was done. After that, the real data was compared with the data came out from the forecasting modal, the result showed that the modal have an error rate no more than 1%, a rapid constringency speed, and a good popularization ability. Ruinous tests can be reduced or completely cut down if the modal was introduced into and economic benefit can be improved also. In addition, the method of improving neural network performance was proposed in this paper.