A visit to Mitsui's roadheader factory in Omuta, Japan was a nice way to round out the end of the year. It was a highly successful visit in which Geodata KODA and Mitsui performed our acceptance test of the electrical, mechanical and sensors systems we've been planning for months.
Calibrations - A Tale of Two Cities
Our main objective was to commission the PLC streaming of sensor data, and then calibrate the machine, ready for deployment in Australia. In retrospect, it doesn't really surprise me that much, but both GEODATA and KODA had two relatively different approaches to calibrating roadheaders. There was a good overlap - there's simply only one way to do certain thing - and then again, there were some interesting differences.
Given that KODA is historically grounded more in Mitsui systems we took a bit more of a lead with the processing of the calibration. I've had a few ideas I've been looking to try out for years, and the results were fantastic. After a machine-based point cloud of all the observations was created we used our coding system to extract key features from the data. The semi-automated workflow managed to find a nice balance between 'doing it all for you' and 'using the experience of the user'.
This exercise also highlighted that all too often the calibration requirements rest heavily on only one (maybe two) individuals in an organisation. We're working towards improving this by: 1) creating a standardised workflow for all roadheaders, 2) creating software to manage the processing and quality checking of the data, 3) and, making the process quicker. We think the software we ended up with at the end of this visit will allow me to process a calibration in about 15 minutes with only a few mouse clicks, and without burning too much grey matter - Let's see.
KODA looking forward to seeing how Mitsui's integrated sensor systems perform. They're not a new concept, but there are some skeptics. The main problem we're talking about here is one related to 'mean-time-to-fail' vs. 'mean-time-to-repair'. In-built sensor systems are far better protected because they are properly engineered, installed in the factory, and generally installed in far more protected locations - like inside cylinders. As such, these systems are known for their reliability and longevity. They also maintain good stability in both scale and offset.
The BUT we need to talk about is: But, if/when the sensors do fail, the time taken to repair the system is lengthy. This isn't specifically a Mitsui or Sandvik problem; just an observation.
Either way, a calm confidence fell over the Mitsui engineer when I quizzed him about these sensors. Hopefully the build lives up to the expectations the project has for it.
One minor point worth noting is the slight refinement we needed to make to the angular values being provided by the Mitsui. For the vertical movement we didn't need to do anything, but for the horizontal movement a 4% scale factor and a small offset were required. These were easily entered into the GEODATA software and the accuracy started checking in at better than 10mm levels.
No big deal, but it reinforces the importance of a survey-based check for any of these systems. Often vertical angles are easy for non-survey staff to calibrate in because digital spirit levels can get you really close without too much effort. Without "horizontal gravity" you're relying upon other measurement approaches, or you take the gamble that the mapping of cylinder lengths and leverage dimensions to horizontal angles is correct. We suspect that's where the 4% discrepancy came from. Easy to imagine. Easy to fix.
Merry Christmas and a Happy New Year. For KODA, you know you've had an interesting year when you're living in one country at the start of the year, and a different one at the end!
Like, share and stay tuned.