I wanted to develop a process for reproducing vintage castings using modern technology and the most effective tools available. It is getting harder and harder to find some vintage automotive parts, and there are fewer American manufacturers interested in producing them. The exhaust manifold I chose for this project was from a vintage BMW 2002. It was the high performance design used exclusively on the fuel injected Tii model, and had an excellent representation as a very durable and high flowing exhaust alternative to fabricated race headers. Exhaust cast iron exhaust headers such as this one are always in demand for auto enthusiasts and restorers.
The part was painted with a low gloss white primer to improve the surface charactaristics for the laser scanner. There are other, less permanant surface finishes, such as chalk and soluble paints, but the primer was cheap and effective. This manifold must be bead blasted to remove the paint, which I had planned on doing anyway.
I then clamped the manifold into a heavy machining vice which was mounted to a rotary table. The vice could hold the manifold in the awkward vertical orientation preferred by the laser scanner, and provide enough mass to keep the whole thing from tipping over. The rotary table allowed me to rotate the entire manifold smoothly around the vertical axis, so that I could take scans from different sides in order to reproduce the entire surface of the part in 3D. This was helpful, as it made assembling the individual surfaces in the computer much easier, because I could first rotate around a known axis
Once the manifold was positioned correctly, I calibrated the laser scanner, and began scanning the first surface. I used a hand held red laser for this procedure, and slowly passed over the part. The computer recorded the position of the laser relative to the calibration cards in the background, and recorded the 3D surface profile. The scanner can only pick up one visible surface at a time, so I needed to reposition the part and take multiple scans to reproduce all of the part surfaces.
The initial scans were still a bit rough. I have not yet built a proper fixture to house the laser, camera, or the calibration cards, so everything is still sensitive to vibration and ambient lighting. That said, the results were usable, and what data was present, after a bit of cleanup was very accurate and true to the part.
The scan was first imported into the open source surfacing program Blender for some cleanup. Blender is designed specifically for 3D surfaces, renderings and animation. I was able to quickly trim away extraneous data points, and smooth out some of the surface anomalies to make the part a little more CAD friendly. I then ran a routine in Blender to reduce the total number number of data points, which reduced the surface detail, but it made the part much more manageable for my CAD software.
A typical high volume production casting in iron can vary by 3mm from the theoretical perfect surface, so there really isn’t any point in scanning with any more detail than that. Even inspecting my own casting, There were a number features that were clearly out of tolerance; drilled holes not centered on their bosses, flat surfaces that weren’t quite flat, and funky parting lines. The molds associated with this part were designed and built by hand, years before CNC. My reproduction would be close, but I have the benefit of solid modeling software to design a part with excellent symmetry, even wall thickness, and even parting lines, and I intended to use it. I also had the option to add or remove features from the casting, and fix design flaws that existed in the original part.
I positioned the simplified part in my CAD program using the mounting bolt centers as a key. I then went to work rebuilding a new solid using the scanned surface as a reference guide. I took hand measurements of critical features to improve the accuracy of the reproduction, but still relied on the 3D surface data to position tricky features and complicated surfaces. The new part in development is shown in green.
The rough casting, with fillets and draft applied. There is some minor cleanup needed, but the part looks fantastic, and it is almost ready for tooling. I added material to the casting which will later be machined off to ensure a high grade surface (such as the mounting surface to the engine, and the threaded bosses. At this point I simply need to scale the part to account for shrinkage in the mold, and I can begin carving my tools on the CNC.