Hey Guys,
I am a ME grad student designing/constructing/controlling/analyzing a linear delta robot for my masters thesis. I have got to the final stages of my work, which is the testing. The problem that I am having is that my Z plane does not seem to be perfectly flat. What I mean by that is when I tell my robot to move from one XY position to another XY position (with no change in Z) the end effector changes slightly in the Z plane. I have attached a dial indicator to my end effector and after calibrating the bed, the Z plane seems to be roughly .5mm higher in the very center of my work space (When I move towards the edges of my work space the end effector moves down a total of .5mm).
I
suspect that this is because of a small inconsistency in my measured dimensions. I designed the robot in SolidWorks and pulled most of the dimensions from that (verifying them as best I could with a crappy measuring tape afterwards and they looked correct) so I thought that they would be pretty accurate. I was hoping you guys would have suggestions for how I could try to pinpoint the issue. Here are two links to my setup:
[
imgur.com]
[
imgur.com]
Since those pictures were taken I have purchased a 12"x12"x3mm piece of borosilicate glass to act as a flat surface to calibrate on (instead of a piece of oak plywood hehe) and added springs under each screwdown point on the bed so that I can level it. I leveled it by moving the end effector (with the dial indicator) to the three corners and adjusting the bed until the Z height at all three corners was identical down to ~.01mm. I did not use the kinematics used by most of you guys, but a more symbolic solution that my advisor (a former NASA roboticist) developed. I have not looked into the differences between his method and the common method heavily but maybe some of you delta aficionados would be interested in it. Here is a link: [
www.ohio.edu]
I am not using any of the common firmware, I have developed my own motion algorithms and software for my thesis because I am using 5th order polynomial trajectory generation so most of the troubleshooting tips that have to do with the firmware will not be applicable. My motor control is pretty spot on, I am not missing any steps and I have a repeatability of 100% effectively. I have verified this through testing the robot along with optical quadrature encoders attached to each motor (vaguely shown in the images above).
So does anyone have any tips for me to try, or any suggestions?
Convex or concave print area can be corrected by changing the arm radius or arm length definitions to match reality.