激光直接金属沉积成形过程的声发射在线监测方法

On line monitoring is an important method to ameliorate the quality of workpieces in laser direct metal deposition and improve the automation and intelligent level of manufacturing processes. This project will comprehensively use the acoustic emission Non-destructive detection, signal processing, machine learning and other theoretical methods to conduct in-depth researches on online monitoring of powder feeding malfunctions and micro cracks in workpieces during laser direct metal deposition. The main contents include: (1) Analyze the mechanisms of acoustic emission sources and acoustic emission signals which are generated by laser direct metal deposition, as well as find out the mechanisms of acoustic emission which are produced by rub-impacts between powder flows and powder supply pipelines. (2) Develop an AE-based monitoring platform. Propose the reliable monitoring solution. (3) Improve the theories and methods of acoustic emission signal processing. On the basis of the analysis of acoustic emission characteristics of powder feeding.malfunctions, online identification methods of powder feeding malfunction characteristics based on the discrete probability distribution similarity of acoustic emission waveform are proposed. (4) Explore the feature recognition methods of instantaneous, weak and bursting acoustic emission signals. On the basis of the analysis of acoustic emission characteristics of micro cracks in workpieces, a micro crack identification algorithm based on the instantaneous similarity of the acoustic emission waveform and KL divergence is presented. This.project is conducive to the developments of acoustic emission signal processing theories during manufacturing procedures. It also equips a theoretical basis for monitoring the process of laser direct metal deposition and provides technical supports for improving the intelligence and automation level of laser direct metal deposition.

在线监测是改善激光直接金属沉积成形工件质量，提升加工自动化、智能化水平的重要手段。本项目拟综合运用声发射无损检测、信号处理、机器学习等理论方法，针对成形过程中的送粉故障与成形工件微裂纹缺陷的在线监测展开深入研究。主要内容包括：（1）分析激光直接金属沉积成形的声发射源与声发射信号产生机理，厘清载粉气流与输粉管道碰摩声发射的发生机制；（2）构建金属构件3D打印的声发射检测方案及检测系统；（3）完善声发射信号处理理论与方法，在分析送粉故障形成特征基础上，提出基于声发射波形离散概率密度分布相似度的送粉故障特征在线识别算法；（4）探索瞬时、微弱、突发声发射信号的特征识别方法，在分析微裂纹声发射特征基础上，提出基于KL散度的声发射波形瞬时相似度的微裂纹识别算法。本项目研究可丰富和发展加工过程中声发射信号处理理论，为激光直接金属沉积成形过程监测提供理论依据，为提高加工过程智能化、自动化提供技术支撑。

本研究以激光直接金属沉积成形缺陷和运行状态监测为研究对象，重点研究了采用声发射对激光直接金属沉积成形典型缺陷和设备运行状态进行有效监测和对缺陷发生的到达时间的有效识别。具体来讲，本研究聚焦激光直接金属沉积成形送粉状态监测、裂纹识别、疲劳裂纹监测等三类问题，提出了基于声发射理论的送粉状态监测方法和构件冷裂纹在线监测方法，构建了系列基于声发射原始波形相似度异常监测和缺陷到达时间识别方法。..其中，本研究探索了基于声发射的激光直接金属沉积成形送粉状态监测方法。围绕激光直接金属沉积成形加工送粉机构的送粉稳定性问题，探讨了采用声发射进行送粉稳定性监测的可行性。在试验数据的基础上，提出了基于声发射原始波形相似度的送粉状态监测方法和送粉异常到达时间识别方法。..本研究探索了基于直方图相似度的激光直接金属沉积成形冷裂纹识别及信号处理方法。围绕激光直接金属沉积成形结构的冷裂纹问题，结合裂纹声发射信号极为微弱、信噪比极低等特征，分别提出了基于巴氏距离的声发射到达时间识别方法、基于AIC和直方图相似度的到达时间识别方法、基于改进巴氏距离的激光直接金属沉积成形冷裂纹监测方法和机遇RCNN和AIC的激光直接金属沉积成形冷裂纹到达时间识别方法，从不同角度对激光直接金属沉积成形裂纹的声发射到达时间识别进行了探索。..本研究探索了基于声发射的激光直接金属沉积成形疲劳裂纹监测方法。围绕激光直接金属沉积成形结构的疲劳裂纹生成及扩展问题，探索了采用声发射技术进行疲劳裂纹识别的技术方案，获取激光直接金属沉积成形结构疲劳裂纹生成到断裂的声发射数据，提出了基于直方图相似度的疲劳裂纹识别方法，对疲劳裂纹的生成、扩展和断裂过程进行了有效识别。..结果表明，本研究能够将其从高度复杂的噪声信号背景中提取关键特征，并构建了基于声发射波形相似性的瞬时、微弱、突发声发射特征识别方法，为声发射无损检测理论和技术提供了基础性的理论支撑。