Why High-Mix Shipbuilding Needs a Different Kind of Welding Robot

Industrial welding robots took hold where the same part is made thousands of times: car bodies, exhaust systems, seat frames. Tooling and part datums are highly repeatable, so days of path tuning are easy to justify when the same program runs at high volume. Programming is a one-time cost spread over enormous volume.

A shipyard works the other way around. A hull is assembled from blocks, blocks from panels and sub-assemblies, and each of those from stiffeners, frames, brackets, collars and plate. Some features repeat many times on one vessel; many never repeat. A yard may build a handful of hulls to one design, and sister ships still pick up changes. High-mix, low-volume work breaks most of the assumptions behind a conventional robot cell.

The bottleneck is before the arc starts

Teach-pendant programming is still common in industrial robot deployments. A programmer jogs the arm to each point on a seam, records positions and torch angles, sets parameters, then tests and corrects. That is acceptable for an automotive bracket. On short-run, high-mix work, manual teaching can eat most of the labor the robot was meant to save.

That programming burden pushes shipyard robots toward the most repetitive work, where setup can be spread across many similar welds. The robot welds quickly enough. What slows everything down is programming, fixturing and checking that the program matches the part actually sitting on the floor.

Think in weld families, not parts

A more useful unit than the part is the weld family: joints that share a joint type, material, thickness range, position and procedure. Fillet welds joining longitudinal stiffeners to plate are one family. Collar plates around a stiffener passing through a web frame are another. Every panel may be different, but the families recur.

A robot built for high-mix work starts there. It generates programs for each new part from data the yard already has, then applies the yard's qualified welding procedure to each family of joints.

From CAD to path

Modern ship-design systems hold most of the structural geometry before steel is cut, although weld intent and production attributes are not always explicit enough for automatic path generation. Offline programming uses that model to identify joints, assign a procedure to each, plan torch angles and approach directions, check reach and collisions in simulation, and only then send the program to the cell.

Offline programming is not new, but it often still needs an expert for each new part. For high-mix work, going from model to path has to become routine, so that a new panel does not turn into a new programming project.

The model is not the part

Real steel never matches the model exactly. Plates distort from heat, fit-up gaps vary, and parts sit a few millimeters from where the model puts them. A path generated purely from nominal CAD can miss the seam when fit-up, distortion or datum error exceeds the welding process window.

That is why seam finding and seam tracking matter so much here. Touch sensing and laser or vision sensing locate a joint before welding. Laser seam tracking or through-arc sensing then corrects the path during welding, depending on the process and joint. CAD-generated paths plus sensor correction is what makes automated welding practical when parts vary.

Reach in tight geometry

Ship structures are physically awkward. Stiffeners and frames form deep, narrow cells, and the torch has to reach around obstacles while holding its angle to the joint. Large panels usually require robot travel on a rail or gantry, and some cells add positioners. Those external axes may move between welds or be coordinated continuously during welding, which makes this a motion-control problem as much as a robot problem.

What the robot needs

A welding robot for high-mix shipbuilding needs capabilities an automotive cell never did:

• Programs from data. New parts come from CAD and the yard's procedure library, not the pendant.
• Sensing by default. Every weld assumes the part may have moved.
• Coordinated external axes. Rails, gantries and positioners are part of one motion system.
• Reach in constrained geometry. Extra degrees of freedom help hold torch angle in tight spaces.
• A record of every weld. Traceability for quality systems and inspection.

Catfish Vanguard designed ARC-R and the seven-axis W7 around that list. Before anything is quoted, we measure a yard's real weld families with the yard.

Industrial robotic welding is mature. Applying it economically to high-mix ship structures is still hard, because programming effort, sensing and motion coordination all have to absorb part variation. Once programming stops costing more than the automation saves, high-mix yards can put robots on far more of their work.