Hybrid Process Prognosis for Metal Ultrasonic Welding - Pro²MUSS

在“金属超声焊接混合工艺预测”项目- Pro²MUSS中，亚琛工业大学焊接与连接技术研究所（ISF）和亚琛工业大学机械动力学与机器人研究所（IGMR）共同开发了一种新的参数化模型，用于接头形成，以便更好地理解超声波金属焊接的过程。超声波金属焊接特别适合于连接电子技术元件，由于电子系统的日益复杂，它日益成为工业关注的焦点。尽管在工业上应用广泛，但在金属超声焊接中仍可能出现工艺波动。这些波动往往无法解释，因为缺乏关于焊接过程中工具与连接部件之间复杂相互作用的科学可靠知识，因此，在大多数实证研究中几乎没有考虑到这些波动。在联合研究项目（AiF IGF项目20.161）和ISF基础研究（DFG项目编号395129909 / GZ RE 2755/52-1）的框架内，已经表明，可以从整个机械系统（包括焊接工具和连接部件）的外部可测量振动行为中获得连接区域内发生的热力过程的信息。整个系统的振动特性与节理的形成有关。两个项目合作伙伴使用这个基础来创建一个参数化的联合地层模型。为此，项目合作伙伴创建了两个子模型。通过对工艺过程的统计确定性建模，可以根据工艺边界条件（如几何形状、焊接参数和材料性能）确定工艺过程的关键数字，如焊接时间、性能过程，特别是工艺阶段的特征。因此，该子模型的目的是根据上述参数预测一个典型的过程。这个模型可以理解为过程的宏观视图。第二，基于机器学习（ML）的建模旨在根据工艺参数的测量提供关于单个焊缝实际实现的接头质量的声明。该模型代表微观节理质量。通过连接两个模型，将关节形成和过程阶段的单个机制分配给机器学习的组件，并尽可能以参数化形式重写。整体模型从参数到工艺顺序，再到可达到或实际达到的接头质量，描述焊接过程。

In the project "Hybrid Process Prognosis for Metal Ultrasonic Welding" - Pro²MUSS, the Institute of Welding and Joining Technology at RWTH Aachen University (ISF) and the Institute of Mechanism Theory, Machine Dynamics and Robotics at RWTH Aachen University (IGMR) are jointly developing a new parameterized model for joint formation for a better process understanding in ultrasonic metal welding.Ultrasonic metal welding is particularly suitable for joining electro technical components and is increasingly coming into industrial focus due to the increasing complexity of electronic systems. Despite its widespread use in industry, process fluctuations can occur in metal-ultrasonic welding. These fluctuations often cannot be explained, since there is a lack of scientifically sound knowledge about the complex interactions between tools and joining parts during the welding process and, as a result, they are hardly taken into account in the mostly empirical research. Within the framework of joint research projects (AiF IGF project 20.161) and basic investigations by the ISF (DFG project number 395129909 / GZ RE 2755/52-1), it has already been shown that information on the thermomechanical processes occurring within the joining zone can be obtained from the externally measurable vibration behaviour of the overall mechanical system, consisting of welding tools and joining parts. The vibration behaviour of the overall system correlates with the joint formation.The two project partners use this basis to create a parameterized model of joint formation. For this purpose, the project partners create two submodels. With a statistical-deterministic modelling of the process course, essential key figures of the process, such as welding duration, performance course and in particular the characteristics of the process phases are to be determined on the basis of process boundary conditions such as geometry, welding parameters and material properties. The aim of this sub-model is thus to predict a typical course of the process on the basis of the parameters mentioned. This model can be understood as a macroscopic view of the process. A second, machine learning (ML) based modelling is intended to provide a statement on the actually achieved joint quality of the individual weld based on measurements of process parameters. The modelling represents the microscopic joint quality. By linking both models, individual mechanisms of joint formation and process phases are to be assigned to components of the ML and, as far as possible, rewritten in a parameterized form. The overall model depicts the welding process from the parameters through the process sequence to the achievable or actually achieved joint quality.