J.E. Jones, V.L. Rhoades, M.D. Mann, T. Surrufka
EnergynTech, Inc.,
United States
Keywords: welding, optimization, modeling, quality, monitoring, control
Summary:
Weld quality monitoring is extremely difficult. The several controlled/uncontrolled, parameters of welding have complex interactions, not amenable to traditional mathematical models. Metallurgically, the computer modeling of melting and solidification is difficult and complex. This is exacerbated by the high speeds of welding and the highly complex electric arc. A new Artificial Neural Network system (P/NA3) was developed which uses Machine Learning to capture the intricate relationships between the parameters (voltage, current, wire feed speed, travel speed, etc.), and the very complex weld cross-section shape. The AI based model is used to optimize the process; and, subsequently, to monitor the welding. Millions of feet of weld are produced every day, to produce products/structures for which the weld is critical to the operation or safety. ArcSentry(TM) uses the AI based model to monitor weld quality in real-time during. It can identify defects and their precise location internal to the weld, so repair/rework can be performed. The currently used method is post-welding visual inspection, but that approach cannot “see” internal defects which could cause failures of the weld during product operation. A modified version of ArcSentry can also be used to control welding process by identifying the beginning of the formation of a defect, and using control feedback to insure a high quality product. The software systems to characterize the complex cross-section shape of a weld, and to model the welding process, were developed by EnergynTech, Inc. and are in the ArcSentry product for use in all industries that produce welded products. This application of AI combined with Machine Learning, to produce the ArcSentry System is unique in the welding industry and significantly better than other weld monitoring products. ArcSentry is being used in advanced manufacturing processes, some of them invented by EnergynTech, which apply conventional welding, or the new EnergynTech Hybrid Induction Welding processes. Both for large structure fabrication as well as small parts mass produced on, for example, an automobile industry assembly line, the ArcSentry system identifies defects, which could – in service – result in a failure. Weld failures can cause significant monetary losses, and/or result in injury or loss of life. Real-time weld quality monitoring and process control will, in almost all applications, reduce the likelihood of a weld failure. AI and Machine Learning has made the ArcSentry system the most capable weld monitoring and control system available. It has been applied to the fabrication of Navy ships, wind towers, safety critical components of automobiles and trucks. ArcSentry was developed for the automobile industry to monitor safety critical welds. The technology is now being used for large structures including shipbuilding and the new EnergynTech UltraTall wind tower. EnergynTech, Inc. is also beginning to apply ArcSentry to other applications, including: 1) Monitoring and Control of manufacturing processes other than welding; 2) Control of robots, including portable and Mobile robots; and 3) Control of the erection of the UltraTall™ wind tower.