Toolroom Software


Volkswagen, Uitenhage, South Africa

The toolroom software was started at Volkswagen in South Africa in 1988. The project lasted two years. The software ran on Hewlett Packard 835 graphics mini-computer and HP 68030 based workstation. The CASE tool was Cadre’s Teamwork—the same price as a Passat, which was not particularly suited to this project, but it kept management amused and added to our collective CV’s. The actual software was written in C and also ran on PCs connected to CNC milling machines for “drip feeding” over an Ethernet network. The project might not seem that novel some twenty four years later, but we had nothing to go by and it was certainly cutting edge equipment.

When we visited the parent plant in Germany, they did not believe we had actually succeeded, as their network was not based on Unix but on DEC VAX/ PDPs written by a Swiss software house who specialised in air-traffic control. There we were, two people from the bottom of the dark continent who managed to keep records of meetings, play along with the CASE tool and write the whole system from scratch. The breakdown in effort was roughly in proportion to the thickness of the documentation; meetings were three times as thick as the source code listings, and the verbose CASE tool spewed out pages and pages of Entity relationship diagrams.

Away from the computer section, the Toolroom people built a new building, purchased digitizers, milling machines, CATIA for NC path generation, try-out presses and the network of Unix machines and PCs. It was a project that would save the country over $200 million in foreign exchange by producing the tooling locally rather than in Japan or Europe.

There were several people who managed to get things done on the same day in a place as big as Volkswagen. Louis van Rensburg was such and individual and it was a pleasure to be seconded from IT to the toolroom under Louis’ care. I have not doubt that it would have been a different experience if the software had not worked or been delivered in time, however, after being the first to place such a big order on Fidia in Italy, shortly to be duplicated by Toyota, the visit to the hosting companies is how you imagine rockstars live. But we skipped the woman and cocain. We were dined along the beach at Genoa in the best seafood restuarant I have ever been to and I have mostly lived on the coast! We visited toolrooms in Genoa, Turin, Milan, Lake Como and a small village were Fidia lived. Absolutely fantastic.

At the end of the project, we were expected to be guardians of the Unix boxes, but as young developers, that was similar to imprisonment... A few years later the toolroom was closed with the machines sent to Brazil.

Before leaving Volkswagen on this page, it would be unfair not to mention Peter Smith, who was a pioneer for introducing Unix into a plant heavily committed to batch processing on IBM mainframes. That was my first experience of IBM’s calling up the army to discredit any non-mainframer until they had to backtrack with the AIX box introduction. Sounds like science fiction, I know... Well done, Peter, and thanks for a great project.

Delta Toolroom

Delta Motor Corporation was the name given to General Motor’s plant when they disinvested to comply with sanctions. They also made extensive changes to their local capability, which allowed me to put in a Sun SPARCstation as a server for NC files. The second toolroom installation had different milling machines, but most of the concepts were similar. Toolpath verification was added at each milling machine on a local PC.

Toyota Toolroom

Toyota Tool and Die was similar to Volkswagen for their supplier of CNC machines. They also ran CATIA on IBM AIX boxes. After the initial inquiry, IBM tried their best to keep us out, however, the offer from our side was simple; pay our flight, we will bring up a Sun workstation and a PC and interface between the IBM AIX boxes and the CNC controllers within a day. If it did not work, we would repay the flight and walk away. The price for the complete job with three days of customisation, five PCs for “drip feeding”, and Sun SPARCstation and all source code was about $30,000. It was one of the more pleasant jobs. The airport in Durban was close to Toyota Tool and Die, so by lunchtime we were machining parts. Two hours later we had a cheque for five PCs which were picked up at the local dealer, installed by that evening and all up and running before IBM even visited. The code could also run on an IBM AIX box, which I did as part of the job as there were a few hours before the flight. It basically involved changing a Makefile and some scripts for directory differences. As that was the last big toolroom to automate in South Africa, I simply left the source code on all the boxes, including the IBM as it would have little value to anyone locally. The local computer support department was extremely capable, so after the initial installation, that was the last we heard from the plant.

Component Suppliers

After the three motor manufacturers had completed their computerisation in the toolrooms, there was not much else that could be done. The software was rock solid and the Unix network was a fortunate choice (PC-NFS for the PCs and Sun's NFS between the IBM CAD systems and the Sun NC code server. For smaller toolrooms, Linux on a PC was a viable option, particularly once Motif ran under Linux or BSD/Unix on PCs. Another very well run toolroom, Smith’s Manufacturing, had a Sun SPARC CAD system for spark erosion, detailed plastic injection parts and a knowledgable computer department. One day was needed to port the system to their SPARCs and network a couple of PCs for local tool path verification plus “drip-feeding”.

Floor Pan tool path

The floor pan of a vehicle. Not that impressive on the plot, but a sight that you would not forget if you saw the machine that chewed through the casting with anything from a 63mm fluted cutter to 11mm and 6mm ball nose cutters for finishing off curves. In all, each vehicle probably had two hundred and fifty or more metal pressing parts, with an average of three and a half operations per part (plus a male and female die), so it easily added up to two gigabytes of NC data. The Sun server ran for a year between reboots. During the December shutdown, I would clean out the disk and shutdown the box.

DEC tool path screen

The DEC Alpha workstation was bought due to the good fortunes of toolroom software. I had ported the software to a standard Linux distribution shortly after buying the DEC workstation, but the Linux on a PC was no match in 1995. The other ports were to Sun SPARCs. I always kept a running system with a complete port of the software to swap out in the event of a failure at Delta or one of the other component suppliers.

Take a quick look at the screen shot; the paths could be scaled, rotated, or multiple paths could be plotted in different colours over the same screen. This was to identify any surface plunging or improperly closed digitizing files. It also allowed the machinist to verify that the casting was properly positioned on the milling machines, as there is no way anyone can visualise a couple of hundred thousand lines of NC code without graphics. A large casting took some time to produce at a foundry and cost thousands of dollars. Welding up a major mistake was not feasible, so we would like to think that the software paid for itself many times over.

Surface Inversion

This was a fairly risky project which none of the motor vehicle companies wanted to try out. They had bought extremely expensive software from either Fidia or Cliks in Japan for surface inversion. Basically, what was required was pushing a mesh over all the NC points, offsetting for material thickness tangentially to the surface and generating the NC code of the opposite tool. I looked into using a network of Transputers with standard C code, but there were no takers.