基于神经网络的核电蒸汽发生器环焊缝局部热处理过程U型管凹痕预防技术

Tube Ding Prevention Technology Based on Neural Network During The Local Heat Treatment Procedure for Nuclear Steam Generator

在核电蒸汽发生器局部热处理过程中,U型管凹痕问题是一个非常重要的质量问题,本文利用支撑板角位移数据建立神经网络模型实时预测支撑板在热处理过程中的变化趋势,并通过热处理过程实验数据进行验证,分析结果证明了其有效性。通过对支撑板角位移的预测能够辅助现场操作人员实现防止U型管凹痕出现的目标。

Tube ding is an important quality problem during the local heat treatment procedure for nuclear steam generator. In this paper, a data-driven modeling method based on neural network is applied to predict the deformation trend of the support plate. The verification of this technology is proved by using the real data of a local heat treatment procedure. The angular displacement prediction of the support plate can be used to help the operator avoid tube ding.