基于声信号识别的焊后残余应力处理质量检测方法

焊后残余应力处理过程的非线性程度高且参数耦合性强，导致处理质量不稳定。而现有检测方法仅做抽样检测，存在检测精度低、周期长等问题，且无法进行实时在线检测。为此，提出一种新的基于声信号识别的焊后残余应力处理质量在线检测方法。该方法先实时采集焊后残余应力处理过程中的声信号并提取其特征，然后构建基于多权值神经网络的焊后残余应力处理质量检测模型，以实现在线识别。实验结果表明，相比于传统检测方法，所提出方法可实现焊后残余应力处理质量的在线检测，可为焊后处理过程中的参数优化和质量控制提供参考。

Detection method of post-weld residual stress treatment quality based on acoustic signal recognition

The post-weld residual stress treatment process has high nonlinearity and strong parameter coupling, which leads to unstable treatment quality. However, the existing detection method only performs sampling detection, which has the problems of low detection accuracy, longperiod, and can not carry out real-time online detection. Therefore, a new online detection method of post-weld residual stress treatment quality based on the acoustic signal recognition was proposed. In this method, the acoustic signal in the post-weld residual stress treatment process was collected in real time and its features were extracted, and then a post-weld residual stress treatment quality detection model based on the multi-weight neural network was constructed to realize online recognition. The experimental results showed that, compared with traditional detection methods, the proposed method could realize the online detection of post-weld residual stress treatment quality, which could provide reference for parameter optimization and quality control in the post-weld treatment process.