超声信号的复合编码激励及与焊接缺陷相互作用机制研究

To guarantee the quality and service reliability of welded structure, it is critical to improve measurement precision and identification accuracy of ultrasonic testing for welding defects. And it has attracted more and more attention. Coded excitation is an effective method to improve measurement precision in ultrasonic testing. But the interaction mechanism between ultrasound field and welding defects is not clear, and it is lack of theoretical basis of coded excitation signal design and evaluation. So it is difficult to further improve to apply coded excitation in ultrasonic testing. Ultrasonic measurement model and testing system for coded excitation ultrasonic testing will be established in this project. Ultrasound transducers response and ultrasound field distribution characteristics of coded excitation will be also studied. Furthermore interaction mechanism between coded excitation ultrasound field and welding defects will be proposed, and a more theoretical method will be given to design and evaluation coded excitation signal. In previous research, bandwidth and central frequency of ultrasound transducer limits time resolution improvement of ultrasonic testing results. To overcome the problem, frequency, phase and amplitude combined modulation signal will be proposed as excitation signal in this project. A reasonable parameters selection will be given to acquire results with high time resolution and signal-to-noise ratio. Measurement precision and identification accuracy of welding defects will be improved through analyzing coded excitation ultrasound field distribution characteristics and interaction mechanism with welding defects.

提高焊接缺陷的超声检测与识别的精度对于保证焊接结构的质量和服役可靠性至关重要，也一直为人们所关注。采用编码激励的超声检测技术是提高检测精度的有效手段，但是由于对该激励方式产生的声场与焊接缺陷的相互作用机制不明，影响这种技术的进一步提高，同时对于编码激励信号设计与评价方法也缺乏合理的依据。本项目拟建立相应的超声测量模型和编码激励的超声检测系统，分析超声换能器对激励信号的响应，研究编码信号激励所产生的声场分布特征，进而明确编码激励的声场与焊接缺陷的相互作用机制，设计出更为合理的激励信号并完成信号的评价。申请者拟采用频率、相位、幅度复合的编码激励方法，克服以往超声激励中换能器频带宽度与中心频率限制检测结果时间分辨率的问题，通过分析不同激励参数下超声信号的传播特征及声场分布，给出合理的激励参数，获得高时间分辨率和信噪比的检测结果，进而提高焊接缺陷的测量精度与识别的准确性。

焊接结构在生产和服役过程中由于工艺偏差、腐蚀、疲劳载荷等原因易产生焊接缺陷，影响其服役可靠性。精确测量缺陷的位置和尺寸等信息是对焊接结构进行可靠性评价和剩余寿命预测的前提，因此具有高时间分辨率和信噪比的编码激励超声检测方法在焊缝检测中有很好的应用前景。本项目采用频率、相位、幅度复合的数字编码方式设计了复合编码激励信号，给出了合理的激励参数，获得了高时间分辨率和信噪比的检测结果。建立了编码激励及阵列超声的测量模型与检测系统，阐明了编码激励声场的分布特征，揭示了超声声场与焊接缺陷的相互作用机制，给出了预测超声回波信号的方法。依据所建立的超声测量模型，开发了“超声智能化检测软件”，针对焊缝进行检测模拟仿真，辅助检测工作者制定检测工艺，提高超声检测的效率与准确性。基于卷积神经网络的深度学习算法及YOLOv3网络结构模型，对碳钢对接焊缝超声检测图像进行训练，建立了超声检测图像自动识别模型，使其具备检测图像中焊缝缺陷自动识别和定位能力。提出了一种多模态缺陷识别方法，构建了包含三个分支的卷积神经网络分别对焊接熔池图片、焊接过程声音和焊接电流电压进行处理，在网络中增加了空间注意力模块和通道注意力模块，以关注焊接熔池图片的细节部分，进而提高焊接缺陷识别的准确性。