Non Linear Axis Control via Data Lookup Table
While it is possible to control a hydraulic scissor lift using Navigator today, the method employed relies very heavily on the PID loop to compensate for predictable changes in the geometry of the lift and other measurable system variances.
The common feedback from end users of hydraulic lifts include;
*the transition off the helper cylinder is too bumpy for the artist
*poor targeting or excessive 'hunting' at the end of a move
*overshooting the target altogether
*poor speed control under varying conditions
One possible solution to these issues is to scale the non-linear output to a data lookup table.
See the attached data table image.
The lift speed demand is placed in the vertical axis and the current lift position is placed in the horizontal axis.
Points are carefully selected based on the characteristics of the lift. For example, multiple position points are selected around the helper cylinder transition to provide the tuner the ability to smooth out excessive bumps. The axis process streams the position and speed demand to the program.
The incoming speed demand and position is parsed out and 4 relavent data points are selected by the program. These 4 points go through a mathematical function called bilinear interpolation and a corrected output is created and sent to the proportional valve.
A test was performed on a rental "Justin HD" scissor lift and valve stack. A CX8090 was programmed usning TwinCAT to handle the lookup table and control the propportional valve.
Please find the attached motion graphs. The first graph represents a typical Navigator axis tuned using the axis output spline. The spline can only provide a technically correct output for a very small portion of the lift's travel. Throughout the rest of the move, the PID is shifting the output.
The next attached graph is the same lift run using the lookup table. It could be further smoothed out by increasing the number of points on the table, but much better control was achieved. It is true that both graphs show the commanded and physical speed lines basically lining up but look at the position error traces. The lookup table operated lift offers PE (position error) that is about 1/5 or less (and that hovering around zero) the error of the spline controlled axis.
The proposed lookup table could be accessed either next to the existing spline in the axis properties or it could exist as an IO function.
-
Jared,
This is a great idea!
-Alex
3 -
This is really cool! Especially love the detail around the helper cylinder transition I've received many a comment on that while onsite.
2 -
Great questions, Mike. I imagine there could be a starter file that has data already populated and the tuner would simply adjust things as needed? I think we could have a clever interface that allows the end user to quickly make adjustments to the table.
2 -
Wow, this would be such a great addition! Good work Jared.
1 -
Super well thought out, Jared. I imagine we could have some helper functions in Nav that allow us to do things like adjust the number of reference points to allow for tighter precision especially on lifts that artists/performers have to ride on. Love this idea, hopefully the Nav team can figure out a way to implement this!
1 -
Great questions, Mike. I imagine there could be a starter file that has data already populated and the tuner would simply adjust things as needed? I think we could have a clever interface that allows the end user to quickly make adjustments to the table.
Being able to adjust the table easily onsite would be fantastic. The weight of props change, turntables get added to lifts and other outside factors can change a lifts behavior after being tuned in the shop.
1 -
Very cool. A couple questions -
1. Would the lookup table have to get generated for each deployment of the lift to account for things like hose length, load and travel?
2. How long does it take to generate the data for the look up table?
0 -
Brilliant idea, Jared!! I'd love to use something like this on a project loading out in the next couple of weeks with a bunch of hydraulics involved. Are there updates on when we could see this feature / options for beta testing?
0 -
Sounds like you want one of these splines per position range you define and the axis to interpolate nicely between them.
0 -
Karl Erickson precisely! I've tested it with real lifts using the data above and it improves control tremendously in my opinion.
0
Please sign in to leave a comment.
Comments
10 comments