Machine Intelligence in Welding and Additive Manufacturing
A special issue of Journal of Manufacturing and Materials Processing (ISSN 2504-4494).
Deadline for manuscript submissions: closed (31 May 2023) | Viewed by 10081
Interests: smart manufacturing; machine learning; human-robot collaboration
Interests: programming for industrial robots; wire arc additive manufacturing (WAAM); industrial automation
Interests: applied machine vision; control systems; manufacturing processes; robotic welding
Special Issues, Collections and Topics in MDPI journals
Arc welding and wire arc additive manufacturing (WAAM) use arcs to melt/process workpieces. While arc welding joins separate workpieces together, WAAM joins the added materials to the workpieces to grow it. All such processes share a common mechanism: arc melts/heats/processes the workpiece forming a weld pool whose dimension/shape/cooling decides the outputs of concern. This is an extremely complex process involving complex heat/mass transport and metallurgical reactions with many variables to determine the outputs of concern.
Producing desirable outputs requires the ability to predict and control process dynamics, upon fundamental knowledge about how process parameters/operations affect the dynamics. Experimental approaches to acquire such knowledge first conduct experiments to measure carefully chosen phenomena, which are not only measurable but most importantly believed to be capable of fully determining the outputs. Because of the complexity of the underlying physics, it is not straightforward how the measured phenomena, although sufficient and redundant, determine the outputs and what plays fundamental roles in the measured phenomena.
Conventional approaches propose/define features or characteristic variables to represent the measured phenomena to determine the outputs. Features are extracted/computed from the measured phenomena and correlated to the outputs through data fitting using chosen model structures that can be linear or nonlinear, parametric or non-parametric, and can be as simple as linear regression models or as complex as multi-layer artificial neural networks. Despite the variation in the model form and complexity, they share a common basis: the features are manually selected, and changing to new features to improve the modeling is a discontinued process requiring human intervention. The resultant modeling process is thus labor-intensive, not automated, and the best results are not easily achievable.
Deep learning (DL) networks are capable of automatically generating different “features” so that proposing features and evaluating the degree of success and effectiveness are automated toward the best results. To provide the capability to generate different features, such networks are typically very complex in structure and require a large amount of data to train. Recent advancements in deep learning have resulted in not only capable structures but also effective training approaches. This opens opportunities to operate arc welding and WAAM processes for higher quality and productivity in innovative ways. To share the experience in taking advantage of this historical opportunity to advance the arc welding and WAAM process, this Special Issue calls for papers that innovatively use the deep learning approach to better solve existing challenges in arc welding and WAAM and to lead to more effective arc welding and WAAM processes. Relevant topics include but are not limited to:
- DL for feature extraction and representation of welding state evolution
- DL for modeling the process-quality relationship toward real-time quality prediction
- Data-driven optimization and learning techniques for adaptive process control
- Robotic welding perception, modeling, optimization, and control
- DL for analyzing sophisticated operations by skilled human welders in complex welding
- Transfer learning for improving the generalization of DL models in prediction and control
- Big data processing, learning, automated network architecture search and refining for online, real-time decision making
Dr. Peng Wang
Prof. Dr. Zengxi Pan
Prof. Dr. YuMing Zhang
Manuscript Submission Information
Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.
Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Journal of Manufacturing and Materials Processing is an international peer-reviewed open access semimonthly journal published by MDPI.
Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 1800 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.