RMC100 analogs & decel distance

This is not exactly a plc problem but I know several people here use the RMC100 control.

I have the RMC loaded up with 7 axis so I set up the remaining two analog inputs as a single differential pressure axis. I was able to scale and zero both of them and they work great. Now I need to see when one of them is greater than a set point. I can't do this with the Poll "?" command because it has to reference the position of the axis which is made up of both of the analogs not just one. Is there another way to do this in the step program or do I need to do it in the PLC?

Second question. I am moving a heavy load and I am having trouble getting it to slow down as it comes into position. I need more deceleration so I can start slowing down quicker to prevent overdriving the valve. How do I calculate the decel distance?

CharlesM said:

Now I need to see when one of them is greater than a set point. I can't do this with the Poll "?" command because it has to reference the position of the axis which is made up of both of the analogs not just one. Is there another way to do this in the step program or do I need to do it in the PLC?

Click to expand...

This can be done in the step table but it isn't pretty. Use the add command that adds a constant of 0 so it effectively copies the Actual Pressure/Force A to a register in the step table. You can then use a subtract command to set flags and use a > or < link type. Now do the same for Pressure Force B.
Note that every channel can be executing tasks so you should have some left over so the pressures can be compared in parallel.

CharlesM said:

Second question. I am moving a heavy load and I am having trouble getting it to slow down as it comes into position. I need more deceleration so I can start slowing down quicker to prevent overdriving the valve. How do I calculate the decel distance?

Click to expand...

It depends on the acceleration/deceleration mode. The motion controller can have its accelerations and decelerations expressed 4 ways.

Mode 0 as a ramp rate in position units per second per second
decel dist = velocity^2/2*accel_rate

Mode 1 as a ramp rate in 1000s or position units per second per second. This is the most common mode. 1 G is 386.4 so a ramp of 100 is a little more than a quarter G. You may need to reduce the decel to 20

decel dist = velocity^2/2000*accel_rate

Normally in the US and CA we calibrate the system so each positin units is .001 inches. Mode 1 then expresses acceleration in inches per second per second since there are a 1000 position units per inch.

Mode 2 as a ramp distance in position units
This is easy as the decel distance is spicified in position units.

Mode 3 as a ramp time in milliseconds.
decel distance = velocity*time/2000

set the s-curve bit.
Did you use the auto tuning?
POST A PLOT. A PLOT IS WORTH 1000+ WORDS! and everyone can learn.
POST your .bd1 file. I may be able to see something by looking at it.
Acceleration feed forwards are your friend.

Curious, why not call tech support? There is someone with a beeper 24/7 although I am prety good at this too.
There is a RMC100 on the internet and we can both access it a the same time and talk on the phone or use MSN messenger.

Here is a plot on my ram I know it doesn’t look very good but I have a few things to deal with. The cycle starts at 1 with the ram moving down with a quick move. At 2 I slow down and bring in my high tonnage cylinders. At 3 I set my pressure limit and continue moving down. At 4 I reach pressure/position and start the pressure ramp down in pressure mode. At 5 I take the tonnage cylinders out when at pressure. After 5 I retract the ram and at 6 I stop the ram quickly and get the overdrive.

My up and down moves with the ram are done with two 6" jack cylinders. When I bring in the tonnage cylinders I turn on two 15" single ended cylinders. By single ended I mean they don’t have a rod side and only push down. I get way behind at 2 when I bring the tonnage cylinders in. I think I can get this better with tuning. The only trouble with that is when its tuned for the tonnage cylinders then the gains are to high without. So I am considering changing my gain settings on the fly and having two gain settings. One for the jacks and one for the high tonnage cylinders. I also think my valve size is not correct. I think it is to large for the jacks but to small for the jacks and tonnage cylinders.

I wouldn't worry about the valve just yet

You are right about having to change the gains on-the-fly. See the 0xD5 to 0xDB commands.
From about 14 seconds to 20 seconds there is a lag that can be corrected by increasing the velocity feed forward. It is the velocity feed forward gains that will help the most while trying to follow position.

When the feedforwards are tuned right you should see the actual position on top of the target position. The integrator term will look like a flat like across the screen. Right now you are relying too much on the integrator gain instead of the feed forwards.

Your extend feed forward should be about 1200. Your retract feed forward should be about 1845.

What material are you compressing? When I see the pressure increase at 15.233 seconds has the press closed on the material? I would stay in position mode until the pressure build up a bit more. It takes a lot of oil to pressurize the oil in the main rams. Keep pressing in position mode until the pressure builds up to at least 500 psi then enable the monitor pressure bit.

Can you actually hold a pressure set point at 2600 PSI?

See mine. This is a CA Lawton press in Green Bay, WI.
ftp://ftp.deltacompsys.com/public/jpg/LAWTON.JPG
This system had a 8 inch jack ram and two 20" main rams similar to yours. During the first 10 seconds the system moved on the jack ram only. It was tracking the position within a couple thousands of an inch. At about 10 seconds the motion controller turns off the fill valve so now the 60 GPM valve is filling all 3 cylinders. The speed MUST be slowed down because the valve cannot move all three cylinders that fast. We also changed the gains on-thy-fly at that point. From 10 to 13 seconds the rams are approaching the material ( silicon rubber ). At 13 seconds the pressure starts to build. I left the press in posiiton mode at this point until the pressure increased to about 1500 PSI. There is no point in trying to control pressure before this because the oil in the main rams needs to compress. ( you need to do the same ). When the pressure reached 1500 PSI I told the pressure to increase slowly to 2000 PSI over about 2 seconds. This last part was repeated up to 7 times. We used a loop for that. All this was done in the controller.

Your system is trying to control the pressure WAY TOO EARLY WHEN IT IS IMPOSSIBLE TO DO!!!! Let the pressure build then switch to pressure mode. USE the pressure feed forwards too!

If you don't have RMCWIN, you can import CharlesM's plot file into Excel and view the data in great detail.

This is a PM press using steel powder. I am using a load cell for the pressure input so I am on the part when I see pressure. I am starting to compress when I set the pressure limit to 2600. I set it lower on the quick move down just as a little extra safety.

I can hold 2600 if i give it time to get into position. My Pressure overshoot should go away when I get the gains correct and get the position correct.

I see what you mean on the analog inputs. I should be able to do the math without any trouble.

Thanks for the plot its good to know that I am going in the right direction.

Peter-

In your JPG, what causes the increase in the target speed at about 14 seconds? I understand the drop that matches it up with the actual speed but not the increase just before that.

Also, that's an interesting ringing right when you switch to pressure control. Is that a response to something physical or is it the control filter responding to an abrupt change in drive requirement?

Keith

Some one is paying attention, excellent questions.

kamenges said:

Peter-

In your JPG, what causes the increase in the target speed at about 14 seconds? I understand the drop that matches it up with the actual speed but not the increase just before that.

Click to expand...

There is a discontinuity in the target position and therefore in the target speed. This is important to understand. When in closed loop position mode the target and actual positions are different because the target position come from the target generator and the actual positions come from the feedback. If tuned right the actual follows the target very closely as it does as the press is closing. The 'jump' in target velocity is caused by the 'jump' in target position. When in not in closed loop position mode, like when in open loop or pressure control, the target position is set to the actual position and the target speed is set to the actual speed. This is necessary so the target will be in the right place when going back to closed loop position mode. The speeds on the graph are computed from the positions so if the position takes a jump the speed will too. We could make the graphing routine smarter and hide or filter this out, but the target position really does jump to the actual position when leaving close loop position mode and entering open loop ( manual ) or pressure/force control. It isn't a problem once you understand why and when it happens. One can control position or pressure but you can't control both.

kamenges said:

Also, that's an interesting ringing right when you switch to pressure control. Is that a response to something physical or is it the control filter responding to an abrupt change in drive requirement?

Keith

Click to expand...

Yes, the ringing is caused by the sudden change in the pressure ramp rate. Normally we would automatically compute a nice pressure profile to the set point. However, in this case there were actually two sets of cylinders that had to be cooredinated to make a long press out of two halves. As soon as either half got above 1500 PSI we told both presses to increase pressure to 2000 PSI over 1.5 seconds. This way both halves would get to the set point at the same time. We could have made the rate after reaching 1500 PSI better match the rate before and then you wouldn't have seen the sudden drop in the control signal. The control signal after reaching 1500 PSI is basically due to the pressure rate feed forward. You can see there is little error.
As far as I know we are the only people that go to such extremes in controlling the ramping during pressure or force control. Notice also that once the pressure set point is reached the valve only needs to dither a bit to maintain pressure. During decompression the feed forwards play a big part again. You can see there is a little more error at the low pressures. Does anyone know why? This question is actually better suited for hydraulic guys but I bet that none of them would get the right answer unless they worked with presses of some type.

Peter-

I can see where the abrupt downward move in the target speed graph would come from setting the target position equal to the actual position. But just before that the target speed moves up at a slower rate. Granted, I'm just looking at a JPEG, but I can't see anything in the target position graph that would account for that. What am I missing.

Since no one else took a stab at your question, I will. I'm guessing that CharlesM's press configuration is fairly typical in that the main rams are driven only on the cap end. As the pressure in the rams decreases there is less pressure available to drive the oil out of the cylinders so the pressure will decrease more slowly. Unless you have the inverse of this modeled and are developing a feed forward to account for this you would expect the pressures to deviate some as the control loop needs to do more to control the pressure. This is just a guess, of course.

Keith

Excellent

kamenges said:

Peter-

I can see where the abrupt downward move in the target speed graph would come from setting the target position equal to the actual position. But just before that the target speed moves up at a slower rate. Granted, I'm just looking at a JPEG, but I can't see anything in the target position graph that would account for that. What am I missing.

Click to expand...

Resolution or the right tools to view the plot in more detail.
ftp://ftp.deltacompsys.com/public/rmcwin/lawton.PLT
ftp://ftp.deltacompsys.com/public/rmcwin/rmcwinse.exe
BTW, you can access our simulators on the net if you install the rmcwin. I will PM you the IP address.

Hopefully you can import the .plt file. or display it using rmcwin. The file is a self extracting file. Also, we do some filtering on the speeds. A MDT gets about .001" resolution so the best speed resolution is about 1 inch per second unless we filter the velocities. The .plt file about will allow you to see the raw data and tell me the exact time you are talking about.

kamenges said:

Since no one else took a stab at your question, I will. I'm guessing that CharlesM's press configuration is fairly typical in that the main rams are driven only on the cap end. As the pressure in the rams decreases there is less pressure available to drive the oil out of the cylinders so the pressure will decrease more slowly. Unless you have the inverse of this modeled and are developing a feed forward to account for this you would expect the pressures to deviate some as the control loop needs to do more to control the pressure. This is just a guess, of course.

Keith

Click to expand...

Excellent. I wish this was obvious to more people.

The right tools do help

Having the plot file and RMCWin certainly make things easier to look at. I'm still just as confused but at least I can see clearly.

The thing that gets me starts between 14.1016 seconds and 14.1206 seconds. The target speed, which had been just twiddling between 117 and 122 for the longest time, suddenly jumps up to 194, then to 204 at 14.1406 seconds. It continues to rise until it tops out at 261 at 14.2969 seconds then it locks to the actual speed at 14.3164 seconds.

The two things that get me are :
1) I can't see a reason why the target speed should increase at that point. You had mentioned that the target speed is generated from the target position. From what I can see the target position doesn't do anything in that time span that would explain the speed change.

2) The target position locks to the actual position at 14.2188 seconds. I would expect the target speed to lock to the actual speed at the same point. But it doesn't lock on until 14.3164 seconds. Why the time difference?

Keith

kamenges said:

The thing that gets me starts between 14.1016 seconds and 14.1206 seconds. The target speed, which had been just twiddling between 117 and 122 for the longest time, suddenly jumps up to 194, then to 204 at 14.1406 seconds. It continues to rise until it tops out at 261 at 14.2969 seconds then it locks to the actual speed at 14.3164 seconds.

The two things that get me are :
1) I can't see a reason why the target speed should increase at that point. You had mentioned that the target speed is generated from the target position. From what I can see the target position doesn't do anything in that time span that would explain the speed change.

2) The target position locks to the actual position at 14.2188 seconds. I would expect the target speed to lock to the actual speed at the same point. But it doesn't lock on until 14.3164 seconds. Why the time difference?

Keith

Click to expand...

I told you there is smoothing on the velocity. We use data points from 10 points before to 10 points after to filter the velocity. This smoothing causes the velocity to rise 10 points BEFORE the target position gets set to the actual position.

Try ploting the velocity using the position data using Excel. You will see the need for smoothing. This is our fix for the problem that many have tried to solve on this forum. It just looks odd in the case of a step change in position.

BTW, the difference between smoothing and filtering is that smoothing uses data from the future.

Were you able to access the controller on the net? It is running now moving back and forth on axis 0 and executing an electric cam on axis 1.

Old post new questions

Peter on your Lawton plot at 14.2 there is a change in the pressure ramp speed. I don't see any change in the pressure setpoint or the event step. What is this and how are you doing it?

Also I am using pressure limit in my plot and the target pressure is a strait line from when the pressure is set. Your target and actual are on top of each other. I don't under stand how you are doing this.

I left the press in posiiton mode at this point until the pressure increased to about 1500 PSI. There is no point in trying to control pressure before this because the oil in the main rams needs to compress. ( you need to do the same ).

Click to expand...

I have not had much trouble with this until now. The press I am working on now is smaller and faster. I am seeing a dip in my position when I set the pressure limit. I think this is a product of the pressure setpoint and the speed that I am moving. So my plan is to follow your suggestion and set pressure after I start building pressure. My question is, will I have the same problem? The only real difference will be I will be closer to the end of the move and I will be seeing a little bit of pressure. My hope is that some of your RMC "magic" will see how quick the pressure is building and make a smooth move out of it. Where now it does not have the pressure reference to work with.

Am I heading in the right direction?

What about the deceleration distance? I didn't see the question.

CharlesM said:

Peter on your Lawton plot at 14.2 there is a change in the pressure ramp speed. I don't see any change in the pressure setpoint or the event step. What is this and how are you doing it?

Click to expand...

It has been awhile. I am pretty sure we used the step table. There were two sets of rams. The first ram that reached 1500 PSI would trigger the step table so both sets would enter pressure mode.

Also I am using pressure limit in my plot and the target pressure is a strait line from when the pressure is set. Your target and actual are on top of each other. I don't under stand how you are doing this.

You have the pressure limit bit set in the mode field of the pressure/force PID. In pressure/force control (P/F) you must set the P/F where you enter P/F mode. That is pressure set point A. Presure Set Point B is where the P/F PID drops out and lets the position PID take over. That is often set to 0 or -1000 so this doesn't happen by accident.

I have not had much trouble with this until now. The press I am working on now is smaller and faster. I am seeing a dip in my position when I set the pressure limit.

Click to expand...

Are you pressing toward 0. I don't know what you mean by a dip in position.

I think this is a product of the pressure setpoint and the speed that I am moving. So my plan is to follow your suggestion and set pressure after I start building pressure.

Click to expand...

Normally you need to do that on big presses. You must first pressurise the oil in the main pressure cylinder otherwise the material push it back. This shouldn't be a problem in pressure limit mode because the press will just keep extending until it either reaches position or reach the pressure limit what ever comes first.

My question is, will I have the same problem? The only real difference will be I will be closer to the end of the move and I will be seeing a little bit of pressure. My hope is that some of your RMC "magic" will see how quick the pressure is building and make a smooth move out of it. Where now it does not have the pressure reference to work with.

Am I heading in the right direction?

Click to expand...

I am not real clear on what the problem is but the RMC does monitor the rate at which the pressure/force is rising. The derivative gain in the P/F PID is critical for controling rapid changes in pressure.

Small presses are usually more difficult to control because of the speeds. Material that doesn't give much is also harder to control. The trick is matching the kinetic energy of the press to that need to form the work piece. If the press closes to quickly there will be pressure spikes. If the press closes too slowly the press will look like it is stalling as pressure build up. The work piece needs to have a certain work done on it ( force x distance ). You try to match the FxD=(1/2)*m*v^2.
The RMC can add or remove a certain amount of energy to make this happen. Make sure the servo valves are not sized too big so the RMC can add or remove energy in fine increments.

Again, a graph would be very helpful.

Click to expand...

Click to expand...

In ram1 I am setting the pressure limit before I get pressure. You can see the dip in position I am talking about.

In ram10 I have moved the pressure limit till I get 50 on pressure. I don't have anything to press on besides steel. I think tomorrow I should be able to get something that will give a little so the pressure is more of a curve and not a spike.

It has been awhile. I am pretty sure we used the step table. There were two sets of rams. The first ram that reached 1500 PSI would trigger the step table so both sets would enter pressure mode.

Click to expand...

I think I get it. You have each one running a step program and you see the pressure change as a result of the other step program.

The graphs say a lot.

You should try ramping the pressure instead of making step changes. In the first plot you can see that when you reduce the pressure set point in one step at, 7.8125 s, it reduces the control output by a step too. You must be able to hear or even feel this.

The second plot looks better but again you are changing the pressure limit in 1 step. You can see that your system doesn't like that. It looks like you don't really want the pressure/force to your initial limit.

Pressure or force is changed by adding or removing oil from the cylinder. The pressure can't be changed at an infinte rate because oil can't be added or removed at an infinite rate.

I would start to ramp down the pressure limit towards your desired pressure limit just before the press closes or I would not even enable the pressure limit until just before the press closes..