A Quandry of Rexroth, Modicon Quantum & Hydraulics

Came into a problem the other day that has me puzzled..maybe you guys can help.
I've got a customer that has several molding machines made by a now-defunct
OEM. Two of the pieces run AB PLC 5's and run reliably and without problems.
The third is one of those "optimized" racehorses that runs a Modicon Quantum.
These machines are basically hydraulically powered and have six operational steps:
1) Prepared sand is blown into a chamber.
2) A Ram squeezes the sand to a operator-set pressure (typical 850 psi).
3) Plate (pattern) opposing the ram raises out of the way.
4) Ram pushes finished mold onto a string of molds.
5) Ram Retracts.
6) Plate opposing ram moves back into position for another mold.
How this is accomplished in the controls scheme is through a Rexroth pump (5,000 psi max) and controller that is straight PID. The Quantum tracks where the machine is in the operational sequence and ramps the Rexroth controller via analog signals (-10 to +10 VDC). A -10 signal "tilts" the pump 100 % one direction, where a +10 signal "tilts" the pump the other direction. Tilting is accomplished through a auxiliary servo with LVDT feedback to the Rexroth.
The main pump does not operate anywhere near its rated capacity except in operation 2 above. Apparently, in that operation, mechanical limits are reached because a "dump" valve opens up to remove the ram pressure for 2 seconds after the 850 psi is reached (this time frame is also selected by the operator) and a flow control is implemented to keep total system pressure from bleeding prior to the next operation.
Two problems here:
1) The pressure in operation 2 is always overshot and
2) The mechanicals move way too fast in operations 3 & 6.
Changing the pump commands from the Quantum seem to help somewhat, but it seems pressure is building up in the system. I feel like we're not dumping enough pressure in operation 2, and maybe the flow control needs opened a little. What is a good practice for tuning the Rexroth controller when it apparently worked fine while the mechanicals were new, but now seems to falter? Do I accel/decel faster? Do I set the LVDT gain down/up? Do I offset the output?
I hope I've given enough information here... any suggestions would be appreciated.
Jake
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Jake wrote:

I'm not familiar with your particulars, so all my comments are general:
First some questions:
One: You've encouraged the customer to think of a mechanical rebuild? They have good reasons not to? If the system is getting worn enough so that it's no longer in tune, how long is a new tuning job going to hold (I know, I'm asking a customer to be sensible, silly me).
Two: Was the "optimization" done by the factory? Was that why they went out of business? Was it tuned to within an inch of its life on day one, leaving it with too little margin for the life of the machine?
I ask because some of the possible causes of the overshoot in step 2 only make sense if they screwed up the tuning on day 1. The differential gain could too low, but this would have been a "since day 1" sort of thing. If the pump itself is worn, particularly if it's gotten leaky, it's effective gain could be lower. This could reduce the stability at the low-frequency end because of the loss of proportional gain and cause nasty overshoot. Again if the pump is worn it may be topping out at a lower flow than the system was designed for and the integrator anti-windup limiting may be too generous.
If the overshoot in step 2 cranks up the overall system pressure (it sounds like it), and if the pressure dump phase is open, rather than closed loop (lazy but plausible), then the excess speed may just be due to excess pressure, and the executive control software calculating the dump time from the commanded pressure for step 2. Alternately, the dump valve could be clogged or otherwise aged so that it isn't flowing as much as it should.
--

Tim Wescott
Wescott Design Services
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

now-defunct
problems.
operational
offset
I agree with what Tom said. In addition I think the original design was rather crude. I would consider the following improvements: 1. Pumps are slow, too slow for controlling pressure accurately. Consider using a servo valve or servo quality proportional valve to control the pressures AND motion but not at the same time. However it can limit both position and pressure at the same time with the right hydraulic motion controller. Pressure relief valves should only be used for safety, not for control. 2. Systems that use only pressure relief valves are not doing pressure control. Unless you can control the differential force across the piston you don't know how much force you are actually applying to the sand around the mold. This requires pressure sensors on each end of the cylinder and the ability to scale the pressures by the surface areas of the piston to calculate differential force. 3. An accumulator will store energy while the pump is running between the cycles. This will help save energy and keep the pressures constant. This helps to keep the controller gains constant. 4. A PID only control is crude as it can't provide any output without have error first. The over shoot you see is due to two things. First the integrator will wind up while the pressure is less than the pressure set point and unwind only after it overshoots the set point. Second, is that the mass of the piston, rod and other stuff that is moving has a kenetic energy which requires the integral of force*distance to stop. If the cylinder rod extends too fast ,the kenetic energy can only be reduced by more force or more distance. Since the sand will only compress so far this means that more force ( pressure * area ) is required. It is best to reduce the speed just before reaching the point where the material starts compressing to avoid this problem.
Modern hydraulic controllers can gather and graph the position and pressures at 1 millisecond intervals. This makes it much easier to tell what adjustments are required. Without this data we can only guess.
Now you need to find our how serious your customer is.
Peter Nachtwey
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
I'm gonna throw a couple of things I ran into at a similar process at a customer's site. It probably won't be worth a damn to you but on the off chance, here goes.
First, are there any signal conditioning blocks between the pieces, like PLC analog to Rexroth or LVDT to PLC? If so, check them for latency. I ran into a particular isolator that took 150 ms to respond to a 5v step input to output. Easy to see with a scope, and plenty to cause nasty problems.
Has Johnny maintenance set the dither to an excessive value or has he played with the dither symmetry enough to cause problems?
If you take the input off the rexroth card and operate it with some power supply or calibrator, does it operate ok...ie is 0 dead stop and +0.5 a little fowrard and -0.5 a little back? I bet the program assumes a linear relationship on the output actuation vs. voltage.
Is the analog output card accurate? Is 0 truly 0, etc.?
In other words, bust the loop and test the pieces individually.
Good luck.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Thanks for the replies so far, guys. I was out at the site today and here's a little more to chew on:
The mechanicals all have recent rebuilds after passing the 1,000,000 mold mark. This includes piston seals, guide sleeves, bushings, etc.
After meeting the maintenance people I cannot rule out buggering.
There is an accumulator installed that has apparently gone "out of whack" occasionally with the ram piston since day 1. I suspect the dump valve is causing this.
Because the pattern thickness changes from job to job, piston extension varies as much as 5 inches over a 48 inch overall stroke.
Servo pressure drives the pump's main swash plate, and it is fired from the Rexroth PWM.
Intermittent noise is present from the LVDT... a good place to start. Can anyone recommend a good isolator for +-10vdc that may be picking up the servo firing noise?
Pump commands are integrated into the system via operational sequence and ram position/velocity via Baliff transducers. The transmitters provide the Quantum both numbers and they're integrated into the programming.
Programming documentation is sparse and what there is... is represented in German only.
Here's an interesting bit... I looked at the numbers from the LVDT and where the pump thought "center" was. I got a -0.385 dc signal while the pump was happily at center (zero pressure output at either left or right port). When I tried to bring it back to "center" (0 vdc as by the LVDT) via Rexroth offset commands, the pump began laboring and producing up to 350 psi on the opposing port. Is this a case of somebody messing with the LVDT's centering? I think the thing really needs to begin with a -zero- feedback so it knows where to go from there (as designed).
Another observation: The Baliff gain had been monkeyed with, too. I did a calibration on it (suspected bottoming of the stroke at one end or the other) and found the span about .575 inches short. I corrected that and we'll see what difference it makes.
Here's the synopsis: I think "Johnny Maintenance" has tinkered with the Rexroth's settings (which are not documented) and there are about 10 menu levels to work from. At a glance, I noticed Output offset was running +4.7 % and Output was 90 %. All of the input factors are zero, which is hard for me to believe in a tuned machine.
I'll go the isolation route for the LVDT (first step) and I'd like any additional ideas as to tuning the thing so (perhaps) this damn dump valve isn't needed. I agree with your replies that it's a crude fix that imbalances the system.
Thanks for the advice.
Jake
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.