Load Cell Calibration
Designing the Rig
To mimic the application on the robot, I designed a stand that would suspend the load cell in the air, and could be bolted to the force gauge.
Using the drawings of the 3-axis load cell, as well as the bolt pattern of the force gauge, I made an L shaped piece that would place the load cell in the path of the applied load, secure it to the machine's base, and allow space for wiring.
This piece was initially designed to be one piece, but I changed it to 2 pieces for assembly and cost of production
At Contoro Robotics, our end effector utilized a load cell to determine the weigh being held by the robot. I was given the task of calibrating these load cells and streamlining the process.
Manufacturing the Rig
Due to the lead time of fabricating the pieces through a manufacturer, we decided to machine the 2 pieces in house.
First we ordered aluminum stock in sizes that were close to the dimensions of the designed parts. Then I used an end mill to machine the stock to the correct size, and to smooth out the faces. To make the holes, I used an edge finder to ensure the placement, and drilled the through and counterbores to the correct depths.
Data Collection
To collect the data, the first step was to secure the rig to the force gauge. Using the correct bolts, I was able to fasten the two pieces to each other. Then, bolts were placed to hold the load cell to the rig assembly. This all was then bolted to the base of the force gauge.
The load cell was then wired through a LabJack which was connected to a laptop. This wiring was done through a large resistor and the I/O signals were matched based on the axes
Calibration
Using the LabJack software, I was able to read out voltages as force was applied in all three axes. I then wrote a script in Python that was able to read the LabJack outputs and create an excel sheet with a set of 5 voltages for each. These voltages were then averaged and I used linear regression to find the correlation between voltage (V) and Weight (N). This correlation was used by our software team to more accurately assess how much weight our robot was holding.
