· 3 min read
Designing a welded frame for distortion control
Weld metal shrinks as it cools. Everything a frame does afterwards follows from that, and most of it is decided at the drawing stage rather than in the shop.
A weld is a small casting that solidifies while clamped to something colder and stiffer than itself. As it cools it wants to contract, the surrounding steel will not let it, and the difference is stored as residual stress and paid out as movement. Every distortion problem in a welded frame is a version of that sentence.
The movement comes in three directions and it is worth naming them, because the cures are different. Transverse shrinkage pulls the joint across the seam. Longitudinal shrinkage shortens the member along the seam and, if the seam is off the neutral axis, bows it. Angular distortion closes the joint about the root, because a fillet or a V-groove deposits more metal near the top than the bottom and the top therefore shrinks more.
The designer controls more than the welder
By the time a frame reaches the shop, most of its distortion behaviour is already fixed. These are the choices that matter, roughly in order of effect.
- Weld size. This is the single most effective lever and the one most often set carelessly. Fillet metal volume goes with the square of the leg, so a 6 mm fillet lays down about two and a quarter times the metal of a 4 mm one, shrinks proportionally, costs proportionally and heats the part proportionally. A fillet sized for the load rather than for the drawing habit is the cheapest distortion control there is.
- Symmetry about the neutral axis. Two seams balanced either side of the axis pull against each other; one seam on one side bows the member. Where a joint cannot be made symmetric, expect the bow and either pre-set for it or machine it out afterwards.
- Continuous versus intermittent. A continuous fillet on a long, lightly loaded seam puts in heat and shrinkage the joint never needed. Intermittent fillets, where the loading and the corrosion environment allow them, cut both.
- Section stiffness. A closed box resists the shrinkage it contains far better than an open channel of the same mass. Closing a section early in the build sequence often costs nothing and removes a whole class of movement.
- Where the functional faces are. A face that has to be flat and a face that has to be near a weld are in tension with each other. Say which faces are functional and put the seams somewhere else.
What the shop controls
The fabricator owns the sequence, and a good sequence does a lot: balancing welds either side of the axis as the build goes up, back-step and skip welding on long seams, welding from the middle outwards, and fixturing that restrains without over-restraining, because a joint clamped rigidly does not distort but does store the stress instead, and it will move when the clamps come off.
Robotic welding helps here in a way that is not obvious: the value is not speed but repeatability. A consistent heat input in a consistent order means the distortion is the same on every piece, and a distortion that is the same every time can be pre-set for, which a variable one cannot.
When to stop fighting it and machine
For a functional face on a machine frame there is a point where sequence and fixturing stop being the economical answer and the honest one is to weld the frame, let it settle, and machine the face afterwards. That is the ordinary route for the frames that leave this hall: weld, stress relieve if the drawing calls for it, then bore and face on the horizontal mill in one setting so the mounting faces and the bores share a datum.
The most useful thing a drawing can do is separate the two kinds of dimension. Mark the two or three that are functional and give them real values; leave the rest to a general tolerance class. A frame with every dimension tightened is not more accurate, it is only more expensive, and it hides which dimensions actually mattered.