Reverse engineering is what happens when the part you need no longer exists but the vehicle or machine still does. The manufacturer has stopped supplying it, the aftermarket never covered it, and the only reference is the part still bolted in place, or a single spare on a shelf. The task is to turn that physical object back into a specification precise enough to make new ones from.
It is a normal piece of custom work rather than a specialism. Negotiation comes first and the project starts after that. Where the customer has a drawing containing dimensions, materials and special requirements, the mould is made from that drawing. Where no drawing exists, the sample route exists precisely for this situation.
The first job is deciding what the reference actually is. A part that has been in service has been through heat cycles, vibration and corrosion. The original shape may now be distorted, and clearances that were once correct may have opened or closed. Before anything is measured, the reference has to be assessed and its condition understood.
Where more than one example of the part exists, comparing them is the cheapest verification available. Features that agree across several worn examples are probably original. Features that vary are probably wear, damage or an earlier repair, and they should not be copied just because they are present on the part in front of you.
Not everything that differs from the original is wear. Many parts are modified in service: a hole opened out, a bracket bent, a mounting face machined to clear something. A replacement that copies those modifications will not fit a standard vehicle. Separating deliberate change from deterioration needs the customer's knowledge, so it is worth asking what has been done to the car as well as to the part.
Damage is usually easier to identify. Cracks, bent sections, fretting marks and gouged faces all have characteristic appearances. The engineering judgement is whether to reproduce the part as designed or to design out the reason it failed. Both are legitimate, but they are different products, and a customer should know which one they are ordering.
Where a part is too worn to measure confidently, the intent has to be reconstructed. That means working from how the part fits its neighbours, what its mating surfaces require and what it has to do. Measuring the space it occupies rather than only the part itself is often the more reliable route, especially for a distorted or partly missing original.
Once the geometry is understood, it is written down. That is the same discipline as the sample route, where a sample goes to an engineer who gauges the dimensions and produces a drawing, and the mould is then made from it. The drawing is what makes the replacement repeatable rather than a one-off guess.
A discontinued part is rarely alone. It usually sits between two things that are still in place and still dimensionally controlled: a manifold face, a mounting boss, a hose connection, a bracket. Those interfaces are the fixed points, and the replacement should be built from their requirements outward. The middle of the part can be reinterpreted; the ends cannot.
That is where drawings of other parts and physical checks work together. Where a mating feature is critical, confirming it against the real vehicle or the surviving counterpart is worth more than any assumption, and it catches the errors a desk-based reconstruction will not.
There is a strong argument for copying a discontinued part exactly: it fits, it works, and nobody has to re-engineer anything around it. There is an equally strong argument for improving a weak design while you have the chance, particularly where the original failed for a predictable reason. That decision belongs to the customer, informed by what the engineer found.
What should not happen is an unplanned mixture of the two, where some features are copied and others are quietly improved without anyone recording it. Every change from the reference should be deliberate, written on the drawing and understood, so the finished part is the one that was actually agreed.
