The existing RMC Controller is pretty old and it is tied to a TI PLC. So, the customer is wanting to upgrade the entire control system to a modern platform. I will see that I have used these cards in the past and they are great. This one still performs well but a new control system is desired by the customer.
OK, the RMC is still good though. It probably talks to the TI over profibus. One could simply replace the Profibus DP card with an Ethernet card though, but that but buying new HYD02s is OK too.
If I were to use MAM instructions, the one axis wouldn't know if the other one slowed down.
I always hear that . So what happens if you gear the axes and the slave slows down? How does the master know it?
I agree that they would have the same profile and should run the same but what about if one hydraulic cylinder/valve combination responded differently.
That is a tuning problem. If they are tuned wrong they will not track correctly no matter what method you use. You must be prepared to tune each axis so that when you take a plot of the motion the target position is overlaid by the actual position. You should get a graph from the RMC and see what I mean.
This is why I was looking at MAG to eliminate some of that. Let me know your thoughts because you are definitely the motion guru.
Simply put, you should NEVER gear axes when you control the all the axes because the the slave axes will use imperfect master actual position instead of the target positions. You gear to external encoders that aren't controlled by the motion controller only.
I never really had much luck with auto-tuning the servo drives.. They seemed to never respond properly or were always tuned too hot.
That is because most auto tuning programs are there SIMPLY FOR MARKETING REASONS. This allows them to say I can auto tune too. Most servo auto tuning use only 3 positions and times to compute the system model. Any errors is sampling results in a poor model. That RMC you are replacing gathers up to 1024 samples and tries many combinations of of models to find which one statistically fits the 1024 the best. In other words it tries many models in search for the best one.
Too many auto tuning programs assume the world is like a text book.
Does the auto tuning work better for the hydraulic cylinders?
Not always. In either the statistical or 3 point auto tuning cases the model must match the physical system. Poorly designed hydraulic systems do not auto tune well but that goes for electrical-mechanical systems too.
Actually auto tuning hydraulic is harder than auto tuning servo motors. We know that.
I have seen where the auto tuning does so well that even I am amazed. There are other cases where the auto tuning only gets close. This can be due to non-linear servo valves etc. Manual tuning is then required to finish the tuning.
I guess I could manually tune the position command vs actual as well as velocity command vs actual.
Yes, I am home now but I can post some plots.
I always get confused with the scaling of the transducer or encoder counts in the AB. Do you know of a easy way to explain this? I am talking about in the axis setup with the conversion constant and I don't really understand windup etc..
Thanks,
Both the Rockwell and the Delta controllers have scaling wizards. It should be easy just fill in the dialog fields. The big issue is whether or not to use the gradient marked on the Balluff rod or to actually measure the position at two points at either end of travel. At some point one needs to get out the dial indicator.
Integrator windup shouldn't be a problem with either the HYD or RMC. If the integrator winds up then the tuning of the feed forwards is way off. If the feed forwards are tuned up perfectly there is no error and the integrator doesn't change from 0 even while moving. We can plot the integrator's contribution to the total control output. If we can see the integrator's control term vary from 0 we know the feed forwards are not tuned well at that point. Then you must look to see whether the system is accelerating/decelerating or moving at a constant speed to see which feed forward gain needs to be adjusted.
Integrator windup should NEVER be a problem on a well designed motion controller. Usually there is an poorly designed system or tuning issues that are the real problem.