基于视听觉多特征解析的VPPAW穿孔熔池动态行为机理与熔透定量评价研究

Characterization of keyhole pool behavior and on-line weld penetration detection has been one of the challenging and difficult tasks in aerospace intelligent manufacture field. Based on perception properties of the welder’s visual-auditory system, the applied project proposes a novel visual-auditory multimodal sensing system to measure the keyhole pool behavior, in order to form a new way to predict and evaluate the weld penetration status. Main contents including as follows: realizing quantification analysis of the keyhole pool behavior based on the visual-auditory character analysis, and revealing the behavior evolution mechanism. Analyzing the influence rules between the weld formation and keyhole pool behavior under different welding conditions (welding parameters, gap, misalignment and heat dissipation, etc.), and illuminating the formation/transformation mechanisms of weld penetration status. Establishing a multimodal information fusion model based on the integration of CNN and fuzzy-integral methods, and verifying/optimizing the proposed model through the forecast verification of in-sample and out-sample. In the end, forming a universal system for quantitatively evaluating the VPPAW weld penetration, in order to provide the valuable scientific method and new technological approach for on-line supervision and optimization of welding process.

铝合金变极性等离子弧焊接VPPAW的穿孔熔池行为表征及焊接熔透在线检测一直是宇航智能制造领域极具挑战性的共性难题之一。借鉴焊工视听觉感知特性，本项目建立了视听觉多模态传感同步实时测量穿孔熔池行为的新方法，探索形成一种铝合金VPPAW焊接熔透的实时预测及定量评价新途径。主要内容包括：开展基于视听觉多特征解析的穿孔熔池行为综合量化分析，揭示穿孔熔池建立、稳定或塌陷/烧穿的演变行为机理；分析不同焊接条件（工艺参数、间隙、错边及工件散热）下穿孔熔池动态行为对焊缝成形的影响规律，阐明焊接熔透状态的形成/相互转变机制；构建基于深度学习网络CNN和模糊积分集成的“特征-决策”逐层级熔透预测模型，并通过样本内/外预测验证及多重优化，最终形成具有高普适性的铝合金VPPAW焊接熔透在线定量评价体系，为焊接质量在线监控及工艺优化提供有价值的科学方法和新技术途径。

铝合金变极性等离子弧焊（VPPAW）由于具有能量密度高、 电弧穿透力强、焊后无气孔缺陷及焊接变形小等优势，已成为宇航制造工业首选的焊接方式。然而，等离子弧复杂独特的物理特性，以及装配间隙、错边和工件散热不可避免的波动难以保证穿孔熔池稳定性和全熔透均匀焊缝，并容易形成未熔透、咬边及烧穿等焊接缺陷。项目组充分利用焊接过程的穿孔熔池视觉图像、电弧声音及电信号等多模态传感信息，采用理论分析-原位传感-知识建模的研究手段，探索提出了基于视听觉感知的穿孔熔池行为准确表征及焊接熔透在线预测新方法，揭示了孔熔池建立、稳定或塌陷/烧穿的演变行为机理，构建了基于多信息融合和深度学习算法集成的熔透预测模型，并申报了国家发明专利。发表SCI收录论文发表了国际SCI收录论文12篇，中文核心2篇；参加焊接智能化国内外会议10次，口头汇报5次。培养硕士生6名，项目成果丰富并发展了焊接智能制造理论，为发展新一代高可靠、高精度和强适应能力的智能化焊接制造系统提供了有价值的理论方法和技术实现路径。