A tube with four bends can be made in more than one order, and the orders do not give the same part. Each bend consumes a length of straight material and each one rotates the plane of what follows. Bend the wrong end first and the straight you need to hold for the next bend is no longer there, or the last leg comes out short with no way to recover.
This is why a bending sequence is written down before the bender is set. It is not paperwork for its own sake. It is the only way to know, before any material is committed, that every leg has enough straight left to grip and that the rotations add up to the shape on the drawing.
The most forgiving approach is to start at a bend near the middle of the run and work outward in both directions. Material can then be fed from either end, and an error at one end does not accumulate along the whole part. It also keeps the longest unbent straight available as a handle for as long as possible.
The cost is handling. A part bent from the middle has to be turned around on the machine, sometimes more than once, and every extra handling step is a chance to lose the rotational reference. On a long assembly with a helper and a steady, that is a fair trade. On a short complex part it may not be.
The alternative is to start at the end with the longest straight leg, use that leg as the reference for the first bend, and work along the part in order. This suits parts where one leg is clearly dominant, such as a system that starts at a flange and runs to a muffler inlet, because the critical fit is at the start.
The risk is accumulation. Each bend is positioned from the previous one, so any error in a bend angle or a bend location carries forward and is added to the next. On a part with several bends that error can end up at the far end, where there is nothing left to adjust.
A bend does not begin and end at a point. It has a radius, and the material consumed by the bend is measured along the centreline, which means a layout that treats bends as sharp corners will come up short every time. The straight between two bends is shorter than the distance between their tangent points on the drawing suggests.
There is also a minimum. If two bends are placed too close together, the second one starts before the first has released the material, and the result is a distorted section rather than two clean bends. That minimum depends on the radius and the wall, and it is the first thing to check when a drawing shows a tight zigzag.
In a single plane, a bend is an angle. Out of plane, it is an angle plus a rotation, and the rotation is where most multi-bend parts go wrong. A degree of error in rotation at the second bend becomes a visible offset at the end of a long leg, and no amount of care with the bend angles will bring it back.
The practical control is a reference that travels with the part: a marked line, a flat that was left on the tube, or a snug that stays clamped until the last bend. Checking the rotation between bends, rather than at the end, is what keeps the error small enough to correct.
Some of this is visible in the finished exhaust pipe work, where a run has to clear a subframe and still land on a flange in one plane.
Every bend springs back when the load is released, and the amount depends on the material, the wall and the radius. Thin-wall titanium springs back differently from stainless of the same size, and a bend that was correct on the machine can be visibly open once it is out. The usual remedy is to overbend by an amount established by trial on the same material.
The second pass, where a bend is re-hit to bring it in, is sometimes necessary and always a compromise. Re-bending work-hardens the section and can ovalise it further, and on a thin wall it may be the point where the material starts to fail. Better to establish the springback allowance on an offcut than to correct on the part.
Fabricated systems are often built from straight sections and bought bends rather than from one continuous tube, and that changes the sequence entirely. Each joint is now a place where the geometry can be corrected, and also a place where it can move during welding.
The advantage is that a misfit can be trimmed at the joint instead of scrapped. The disadvantage is that every joint adds heat and shrinkage to a part whose shape is defined by its bends. On cut and welded sections the sequence therefore has to cover both the bending order and the welding order, and the two are not independent.
A trial bend on an offcut of the same material and wall is the cheapest information in the shop. It gives the springback, it shows how the section behaves at that radius, and it confirms the machine setup before a real part is committed. It should be kept, marked with the setup, and used as the reference for the next run.
Records matter for the same reason on repeat work. The sequence, the rotation references and the springback allowance together are the process. Written down, they let a second operator make the same part; held in one person's memory, they retire with that person.
Bending sequence is decided at the drawing stage, because it depends on lengths, radii and clearances that cannot be invented at the machine.
None of this is guesswork once the sequence is written. The part is either possible in the order proposed or it is not, and that is worth knowing before the material is cut. The bent pipe assemblies on this site are made this way, the fabricated components pages cover the brackets that hold a run in place, and stainless welding covers what happens at the joints. Where a system is copied from a vehicle, the customization process starts by gauging the original, and the mounting hardware that fixes the finished shape sits under other parts.
