How I Have Failed to Implement my First Robot in Time
Posted on Nov 07, 2017 7:00 AM. 4 min read time
I recently implemented my first robot on our production line. We made a lot of mistakes but at the end of the day everything was working out. I now want to use my experience, errors and successes and give them to people willing to implement their first robot flawlessly.
One of the main reasons why I couldn’t implement the robot on the scheduled time is that I didn’t have any standard method.
Our main goal was to calibrate our FT 300 Sensor by means of an automated process. In fact, using a robot instead of a manual calibration process would allow us to generate a more precise calibration. And so we had the idea; we pretty much knew what hardware we would use, but the way to do it was uncertain. That is where it all started to go wrong.
First, the engineering department did a proof of concept that was convenient enough to start the project (we were lucky enough to have a robot around). So we designed a cell that was running a UR10 which was moving different parts of various weights. We then designed a cool extrusion table that could hold the robot. We put everything in CAD files, set up a budget and showed that to the R&D department.
When they saw the initial design they told us we needed to have a table that was more rigid, and that the parts that needed to be carried should be fixed in a more reliable manner.
From now on, take note that words in bold are total waste to the process!
So here we are again, back to the drawing board. Designing a second version of the cell with everything bulkier and more precise. We went back to the R&D and they told us it was good enough, that we could buy everything we needed.
We builded our setup on a Vention extrusion table like this one
All the hardware came little by little. We assembled and tested everything as it came through. We programmed the robot keeping in mind all the different parts it would carry and the precise way to fix them. We then realized the calibration process would take an hour whereas the previous one was 5 minutes long. Albeit more precise, the lengthy calibration would prevent us from outputting more FT 300 Sensor. On top of that the part fixation was too tight and was causing too much friction, which in turn would trigger the robot safety features. We were all really depressed as we saw the integration phase of our project getting longer and longer. We were not even thinking about the operating phase.
What were we going to do?
With all the problems we were facing and the time constraint, we needed to find a solution. We went back to the drawing board, but this time from scratch. What we realized by doing more testing is that what we didn’t need a more precise fixation method for this fixture could be handled by a robot. We also found a way to shave some time at the calibration step by adding a validation process that could save us almost 15 minutes. Without going into deeper details, what was wrong from the beginning is that we did not spend enough time designing the robotic cell. We started with an idea, but went straight into the wall. It cost us money (parts and time), resources and energy. At the end of the day, the robot was not running and our ROI was tougher to achieve.
So now that I’ve told you my little story and that I’ve read Samuel’s book, I can tell you a couple of tips to accelerate your implementation process from Design to Operate.
Build a dedicated team
Build a team that can bring a project from A to Z. You will need a group of persons who make the design brainstorm (so you don’t start in the wrong direction), someone who will design the cell with all the known criteria, have the final goal in mind and keep safety features in the loop. In addition to your team, you will need a project manager and most likely someone who will program the robot. It can all be the same person depending on the size of your shop. At the end of the day you need to have a team that is aware of the global aspect of the robotic cell, and that will be able to output the minimum viable cell (Lean Robotics) in the targeted time.
Have a bulletproof design
In my opinion, the design is like the base of a pyramid. If you neglect your design, the base of the pyramid will be smaller than the levels above it, and everything will eventually collapse. In fact, building a solid design with the right team and the right planning is the key to success. The base of your pyramid will be built strong and will allow you to go into the integration phase more easily and with less waste. If you test everything and make sure that the robot can achieve the task, you basically just have to program the robot to do it and all the uncertainties should already be solved.
Use technology to help you
We used different methods to debug our robot. First, we tried filming it to see and analyse what was going wrong. We couldn’t identify the problem at that time but I know that some companies use that method to analyse their robot.
We also installed Insights on our robot. It wasn’t indicating what was wrong but at least we knew when it was down and we could go and see what went wrong. Insights hand in hand with the robot log helped us to find the different section of the program that were critical, and solve them one after the other. Read more about: How Insights Helped me Debug my robot.
I hope that this is good enough a hint for you. Implementing your first robot is a tough job. You need to consider a lot of factors and everything has to come together perfectly. I totally encourage you to read Samuel’s book if you are a first timer or even if you want to refine your implementation process.