基于声发射的激光冲击强化表面硬度智能监测

Intelligent Monitoring of Surface Hardness Based on Acoustic Emission in Laser Shock Peening

为了实现激光冲击强化过程中材料表面显微硬度的实时评价,提出了一种结合声发射技术和机器学习技术的用于7075铝合金板材表面/次表面硬化率的在线监测方法。首先,通过离线硬度检测构建了表征材料表面硬化的综合指标——亚表面硬化率;其次,利用模态声发射理论实现了基于反对称A0模态的梅尔倒谱时频图特征提取;然后,构建了融合多个感受野和注意力机制的神经网络质量评估模型;最后,通过激光冲击强化的实测数据对所提出方法的有效性和可行性进行了验证。实验结果表明,提取的时频图特征中具有更丰富的细节信息,相比于传统的神经网络,所提出模型的平均准确率最高达到了97.41%。

In order to achieve real-time evaluation of the surface microhardness of materials during laser shock peening,an online monitoring method for the surface hardness of 7075 aluminum alloy combining acoustic emission technology and machine learning technology is proposed.Firstly,a comprehensive metric to characterize the surface hardening of material,i.e.the sub-surface hardening rate,is constructed through offline hardness testing;secondly,the anti-symmetric A0 mode-based Mel cepstrum time-frequency map feature extraction is implemented using modal acoustic emission theory;then,a neural network quality assessment model incorporating multiple sensory fields and attention mechanisms is established;finally,the validity and feasibility of the proposed method are verified by the measured data of laser shock peening.The experimental results show that the proposed time-frequency map features are richer in detail information,and the proposed model achieves the highest average accuracy of 97.41%compared with the traditional neural network.