An official website of the United States government
Abstract This review paper examines the application and challenges of machine learning (ML) in intelligent welding processes within the automotive industry, focusing on resistance spot welding (RSW) and laser welding. RSW is predominant in body-in-white assembly, while laser welding is critical for electric vehicle battery packs due to its precision and compatibility with dissimilar materials. The paper categorizes ML applications into three key areas: sensing, in-process decision-making, and post-process optimization. It reviews supervised learning models for defect detection and weld quality prediction, unsupervised learning for feature extraction and data clustering, and emerging generalizable ML approaches like transfer learning and federated learning that enhance adaptability across different manufacturing conditions. Additionally, the paper highlights the limitations of current ML models, particularly regarding generalizability when moving from lab environments to real-world production, and discusses the importance of adaptive learning techniques to address dynamically changing conditions. Case studies like virtual sensing, defect detection in RSW, and optimization in laser welding illustrate practical applications. The paper concludes by identifying future research directions to improve ML adaptability and robustness in high-variability manufacturing environments, aiming to bridge the gap between experimental ML models and real-world implementation in automotive welding.
more »
« less
Virtual Sensing by Dense Encoder for Process Signals in Resistance Spot Welding
Abstract Resistance Spot Welding (RSW) is one of the largest automated manufacturing processes in industry, consequently making it also one of the most researched. While this ubiquity has led to advancements in the consistency of this process, RSW is innately uncertain due to the high degree of interplaying mechanics that occur during the process. Additionally, to ensure the quality of a completed weld empirically, expensive analysis tools are required to inspect the result. One solution to removing this monetary and temporal cost is in-line process monitoring. During the weld, various signals can be measured and evaluated to predict the weld quality in real-time. The most common signal to measure is the Dynamic Resistance (DR) due to its ease of sensor implementation and richness of information. Other common signals are the electrode force and displacement. These give a more inclusive look into the overall process, especially the mechanical aspects, but these are typically limited to lab settings due to the increased cost of deploying them at scale. One solution to realize the insight of these other process signals on the factory floor is to utilize Machine Learning techniques to create virtual sensors that convert extant sensing data to other domains. This would allow for more robust and interpretable signal processing without incurring additional costs or downtime.
more »
« less
- Award ID(s):
- 2237242
- PAR ID:
- 10583993
- Publisher / Repository:
- American Society of Mechanical Engineers
- Date Published:
- ISBN:
- 978-0-7918-8788-2
- Format(s):
- Medium: X
- Location:
- Seattle, Washington, USA
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
This study aims to extract critical scenes/continents in the weld pool region during gas metal arc welding (GMAW). The scenes considered include the wire, arc, and weld pool, while other secondary ones such as oxides are temporarily excluded. They are critical to understanding, analyzing, monitoring and controlling the welding process, in particular the critical correlation how the welding parameter, arc and weld pool are dynamically correlated. Unfortunately, such fundamental correlation has not been studied and lack of effective ways to simultaneously monitor/extract these scenes is responsible. With the development of optoelectronic devices, weld pool regions can be better imaged. However, because of the nature of the scenes in particular the arc which is formed by ionized gas without a clear boundary and highly dynamic, detecting them using computer vision is challenging. Deep learning is an effective method, but model training usually needs a large number of labels. As manually labeling is expensive, we propose an approach to address this challenge that can train a model from a small, inaccurately labeled dataset. This approach is designed, per the characteristics of the scenes and their dynamics All-in-One Network (AOD-Net) was deployed first for defogging, and then the YOLOX network was utilized to identify regions of interest to reduce the impact of molten metal splashes on image sharpness. Subsequently, we developed a timed segmentation network incorporating the Long Short-Term Memory (LSTM) mechanism into U-Net, which can be used to extract more accurate information about the weld pool by combining the temporal and spatial information in the continuous process of welding at a low cost because our scene of interest is in a continuous and dynamic evolutionary process. After defogging and removing the effects of molten metal spatter, we can obtain information on the dynamics of the weld pool and the weld arc at the same time. Experimental results verified that the trained network could extract the critical boundaries accurately under various welding conditions despite the highly dynamic changes and fuzziness of the views.more » « less
-
Abstract UNS N06693 is a Ni-base alloy that provides metal dusting corrosion resistance in steam generator pipes with operating temperatures above 500°C. A crack failure occurred in a 6.5mm thick similar weld pipe joint, located at both fusion zone and heat affected zone, after about 10 years in service and 2 months after weld repair in adjacent weld, which warranted an investigation into possible root causes of failure. This study investigates the potential failure mechanisms that may arise during service (such as stress relaxation cracking, stress corrosion cracking, ductility dip cracking, and creep failure) for UNS N06693 in order to understand the observed cracking behavior. In this year, preliminary fractography, metallurgical characterization, thermodynamic and kinetic CALHAD simulations, and investigation into potential contributing factors (e.g., weld procedure specifications (WPS) and post weld heat treatment (PWHT)) to failure have been completed. The fracture surfaces indicate brittle, intergranular failure, such that no shear lips were observed, and radial lines (crack propagation) were primarily observed in weld fusion zone. Metallurgical characterization near the fracture surface is conducted to reveal the contributing factors to failure, such as intermetallic phases (e.g., Cr-rich α-phase) and distribution of carbide particles (e.g., intergranular chromium carbides), that may contribute to reduced cracking and sensitization resistance. Blocky, intergranular Cr-rich precipitates, either Cr-rich α-phase or Cr-rich M23C6., are observed behind secondary cracks. Based on the initial findings, contributing factors for failure considered are increase in tensile residual stresses due to nearby repair field weld and grain boundary embrittlement due to coarse, blocky Cr-rich phase that likely developed during initial PWHT and within the 10-year service window. In the following year, a more in-depth metallurgical characterization, discussion on contributing causes and possible mitigation strategies for improving microstructural stability and performance-based weldability (e.g., weld procedure and PWHT design), and conclusions with root cause analysis will be provided.more » « less
-
Abstract Sub-surface voids and material heterogeneities resulting from the friction stir welding (FSW) process often necessitate post-weld inspection to ensure the quality of weld obtained from this solid-state welding process. In this context, in-process void detection techniques can potentially help in optimizing the process conditions and thereby reduce expensive and time-consuming post-process inspection of welds. Current in-process void detection techniques rely on approaches that try to directly correlate the part-scale welding quality to void formation, without a fundamental understanding of the underlying mechanics and materials physics that modulate void evolution. In this work, we demonstrate an effective in-process numerical technique that uses process force signals to detect volumetric void formation and connect the variations in the force signals to interactions between the tool probe and the underlying material voids. Our approach relies on a high-fidelity finite element analysis simulation of the FSW process and on correlation of numerically obtained process force signals with the corresponding void structures. This correlation is obtained in the phase-space relating in-plane reaction forces on the tool to the tool rotation angle. We focus on the interactions of the tool geometry and tool motion with the surrounding material undergoing plastic deformation and deduce novel insights into various correlations of tool motion and void formation. Through this approach, we can identify tool-related process conditions that can be optimized to minimize void formation and demonstrate a potential in situ force-based void monitoring method that links to the underlying plastic flow and void structures during the FSW process.more » « less
-
During machining, kinetic energy is imparted to a workpiece to remove material. The integrity of the machined surface, which depends on the energy transfer, affects the quality and performance of the product, therefore needs to be quantified. Prior studies have indicated the potential of using machining power, or the power consumption at the tool-chip interface, as a process signature for predicting machined surface integrity. However, direct measurement of machining power is constrained by the availability of special equipment and the associated cost. To address this gap, this paper presents a machine learning-based method for machining power prediction through multi-sensor fusion and sequence-to-sequence translation from acoustic and vibration signals, which represent portions of the in-situ kinetic energy dissipation, to the machining power signal as a process signature. Specifically, a neural network architecture is developed to separately translate the acoustic and vibration signals to corresponding machining power signals. The two predicted power signals are subsequently fused to arrive at a unified power signal prediction. To check for spurious decision logic, the sensor fusion model is interpreted using integrated gradients to reveal temporal regions of the input data which have the most influence on the machining power prediction accuracy of the fusion model. Systematic cutting experiments performed on a lathe using 1018 steel have shown that the developed sensor fusion method for process signature prediction can successfully map machine acoustics to power consumption with 5.6% error, tool vibration to power consumption with 8.2% error, and acoustics and vibration, jointly, to power with 2.5% error. Model parameter interpretation reveals that the vibration signal is more influential on the machining power prediction result than the acoustic signal, but that overall model accuracy is diminished if only the vibration signal is used.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1115/ISFA2024-140071
