A typical problem for a car manufacturer: A small and cheap part in production causes big and expensive problems in the vehicle. It is suspected that it is due to the scattering of the coating thickness. The experienced engineer creates an Excel template and starts to measure the contour with a caliper gauge. He measures a few parts with 64 measuring points each and needs 16 minutes and 31 seconds for one part. A digital display is a valuable aid. A first evaluation shows that not all measuring points are necessarily needed and the effort is reduced. The employee, who then has to carry out a larger series of measurements, finds that it is quite tedious and takes a lot of time to constantly change the caliper against a pen or keyboard just to note the measured value. And two employees are already carrying out the measurement. One announces the values, the second one records them. To avoid any errors, the values are repeated. A solid process is established.
However, a complete measurement costs twice as much as the part itself. It is therefore impossible to carry out the measurement for every part. We have made the effort once and replayed the scenario. A typical measurement result could look like this:
The very coarsely resolved data is also recorded incorrectly despite all the effort. Depending on who is measuring, the measured values differ.
Oh, if only you were a bit of a maker, like our student Phil for example…
… then you would take a Raspberry Pi, put a PANDA I TIMESWIPE board on it and connect two laser distance meters. In addition connect a known motor control, write a small Python script (or download Phil’s version here), quickly print a case and recording (download here) and already, about 2 days later…
The automated laser station measurement takes 2.7 seconds, whereas the manual measurement takes 16 minutes and 31 seconds.
This means that our TIMESWIPE measurement is 370 times faster, 1000 times higher resolution and 100 times more accurate, i.e. together 37,000,000 times better than by hand. In fact, it is carried out so fast that you can actually measure any part. With a few more commands in Python, the first day of measurement with Machine Learning, the first analyses are carried out and a model is trained which tells whether the quality is OK. Instead of a specification sheet, the manufacturer simply gets the device as a template to automate this measurement in-house for quality control. Find the repository here https://github.com/Daffeldoff/TwinLaser.
Would that really be too much to ask in 2019, in the land of engineers? Or is everyone afraid of the wrath of the caliper mafia?