MHI, Algomatic AI video tool turns welders' tacit skills into data

MHI and Algomatic Dynamics developed the AI video-analysis solution as a way to address a specific workforce problem: many experienced site welders are expected to retire within the next five years, making it difficult not only to secure replacements but also to pass on their knowledge and skills to younger workers. The developers positioned TIG welding, a base technology supporting products from energy plants to rockets, as one field where veteran know-how must be transferred more effectively to the next generation. Their basic approach is simple: record both a skilled engineer and a less-experienced worker performing TIG welding, then use AI to analyze the differences between them.

The system extracts key evaluation points from the uploaded videos and uses individual analysis modules to visualize and verbalize the differences, focusing on things like how long the torch tip stays in one position, the torch tip's path, and the brightness distribution. The team describes these differences as a kind of "bodily knowledge" that is well suited to drawing out tacit knowledge. When a trainee's video is uploaded, the AI agent compares it with the veteran's characteristics, evaluates the skill level, and presents improvement points, which lets both the worker and the instructor grasp specific, quantitative gaps.

MHI expects the system to shorten TIG welding training periods and contribute to getting welders productive sooner and improving productivity, and it is considering applying the same video-analysis AI to other tasks that rely heavily on veterans' bodily knowledge. The developers also caution against over-expecting from AI, noting that video alone cannot capture everything and the AI cannot supply all the answers. How much the training period actually shrinks may hinge on how well those limits are handled in practice, and on whether the same approach transfers to tasks beyond TIG welding.