Please excuse me if I make silly assumptions for I have little experience with hydraulics. I would very much appreciate your input for the
I wish to lower a drill of known weight into the ground (please see the attached representation). As it goes down, some of the weight of the drill will rest on the material as it drills.
The idea is to monitor and control this amount of weight resting on the material. Following some experimentation and after reviewing your support materials,
I would appreciate any clarification on the following:
You have mentioned before that it takes a flow of oil to move the system where it doesn’t if you controlling pressure or force, making simultaneous force
and motion control tricky. I can see that the force on
the cylinder can be approximated accurately from the rod and cap pressures if the valve is not stroked. But after some experimentation I see that as soon as the valve strokes one way or the other,
the difference between cap and rod pressure varies wildly as one port receives the full supply pressure while the other looses most pressure as it discharges
to the tank. I assume that this is the difficulty that you refer to, although I don’t know how much is due to deficiencies in my test rig. Taking this into account,
would it be more advantageous in my application to just monitor the pressure in the cap?
This would somewhat reduce the large variations in force measurement. If the drill is lowered at a slow and almost constant speed, the pressure in the cap should give
some indication of the load on the cylinder although the port is open and oil is escaping. It becomes very hard to control though, for if the port is opened/closed fast
the pressure will still vary wildly.
You have also stated:
Here I understand that the actual force on the cylinder cannot be measured accurately if it is not acting against the supply pressure and/or if it is small relative to the supply pressure generated force,
which is generally high. This is perhaps my main concern, as I am merely holding the weight of the drill and letting it down gently, so my supply pressure is not acting against the weight being held.
What are your thoughts and recommendations in this regard?
Drill.pdf (7.67 KB)
VERY IMPORTANT: Never try to measure the force of a cylinder by measuring only the cap end. That’s king of like measuring your wieght with a floor scale but not taking into account the fact that you are trying to do a chin-up so the scale’s reading of 30 pounds is false.
It appears that you are confusing the control of force with the measurement of force. Force can always be measured accurately. There is a Bernoulli effect that may affect the measurement a bit if the oil is flowing quickly, but typically the pressure transducers are installed in the cylinder itself where the flow is fairly low, and there will be no effects.
You mentioned that when the valve opens, you think it is difficult to measure because the one side is at system pressure and the other side is close to tank pressure. This is not true because there is a lot of pressure drop across the valve. The valve is not just simply a large on/off opening. It is a very small orifice that oil is squeezed through. Therefore, it’s takes time for the pressures to change on either side of the piston. There will be significant pressure on either side. Your pressure measurement will work very well. Even if there wasn’t significant pressure on either side, the measurement would work well beacuse that’s what pressure transducers do well, is measure pressure accurately.
Controlling force while moving is very doable. If you sometimes need to control force while the axis is stopped or moving very slowly, and at other times when the axis is moving quickly, then the tuning gains will need to be different, because the system behaves much differently in these two cases. Tuning force can be a bit tricky for someone not aware of this fact.
You referred to this quote:
This does not mean that force can’t be maesured (see explantion above). It simply means that you can’t exert 10 pounds of force on nothing. If I tell you to use your hand to exert 10 pounds on nothing, and I place a load cell between your hand and nothing, the load cell will read zero, no matter how fast you move your hand. Now, if I tell you to exert 10 pounds on a chair, it is possible, because the chair has mass, which will resist motion, because Newton said F= m x A. If I tell you to exert 10 pounds on the ground, it is possible, because the ground has mass (lots of it).
Hello Jacob, and thanks for your prompt reply.
I have wild variations of force measurement in my setup when moving which, after reading your reply, I must attribute to a device or setup problem.
I have a 2.5 inch cylinder (1 inch rod) with a supply pressure of 500 PSI moving a load of approx 120 pounds up and down at less than .5 inches/second. The force reading will vary between approx. ±800 pounds.
I will be checking the RMC setup/configuration and the physical devices, but meantime if you think of anything that might direct me to the problem please let me know.
How much do you think the system will be affected due to the valve being 6 feet away from the cylinder through flexible hoses? Although far from ideal, I cannot see that this would make such a great difference.
For slow motion, the 6 ft of hose may not be a major problem. It can also depend on the rest of the system. If you have a lot of static friction, the hose could be a problem.
A plot of the motion including the position, force, and Control Output would be of great help for figuring out the problem.
Here are a couple of plots. If you could take a look at them and let me know what you think I would be very grateful.
The pressure transducers seem to be fine, although receiving a lot of noise and loosing a bit of resolution as they are rated for 0-3625 PSI and I am supplying <500.
I will be reviewing the rest of the system now to see if I recognize the problem.
Raising 120pounds 10inchesAthalfinchsecond.rmcplots (4.55 MB)
Lowering 120pounds 10inchesAthalfinchsecond.rmcplots (4.87 MB)
The force has a lot of noise on it, and there appears to be a significant noise component around 60 Hz. Other than the noise, the force is behaving reasonably well. I would check the wiring on the system. Make sure the cable is twisted-pair, shielded. Make sure the commons are wired correctly, according to the wiring diagrams in the RMC Startup Guide. Also, make sure that the power and common wiring to the transducers are not shared by other noisy devices.
If there is still a lot of noise after checking the wiring, you can apply a filter to the force. In the Axis Parameters, on the All tab, expand the feedback (may be secondary feedback) section, expand Filtering/Modeling, and set the Force Filter value. Start with a high number, perhap 100, then decrease it to apply more filtering. It is a low-pass filer, and the Force Filter value defines the 3 dB cut-off frequency. Filtering too heavily will affect the force control.
But disregarding the noise, should the load reading not stay constant? These motions were done holding a constant weight of about 120 lbs, yet it reads:
Lowering: 100lbf during motion and 575lbf afterwards.
Raising: 610lbf before motion, 1100lbf during motion and 475lbf afterwards.
I would expect that the force would be constant during motion, which it is. I would expect it to be constant when it is not moving, which it mostly is, except for the settling time, but notice that the settling time corresponds closely to the position slowly coming to a stop, so I would say that it is behaving very normally. A hydraulic system requires a fair amount of force just to squeeze the oil out the valve, which is why you are not seeing just 120 pounds. This force increases with velocity, which is exactly what the plot is showing. All in all, it looks OK. Clean up the force with a bit of filtering (or wiring), and it should look real good.
The system seems very sensitive to me. Could you give me some advice on the tuning based on these plots please?
Tuning_Force_Oscillations_5.rmcplots (1.96 MB)
Tuning_Force_Oscillations_4.rmcplots (1.88 MB)
Tuning_Force_Ramp_Up_3.rmcplots (1.97 MB)
Yes, the system does appear to very sensitive. The plots show that not very much Cotnrol output is required for the motion, so the valve is apparently much larger than it needs to be for this motion.
I noticed that the Output Filter is set to 1, which is extremely heavy filtering, and could be contributing to the control problems. The Output Filter value is the filtering cut-off frequency. You should start at a high value and then come down. To begin with, you should leave it disabled (value of 0).
How much feedback filter have you applied?
Once the filters are correct, the control may be easier.
The valve needs approx. 0.5V to stroke enough to allow motion. Even at 1 V the cylinder moves at less than 2 in/sec. Would the position and condition of the valve not be more relevant than the size? (It is old and 6 ft of flexible hose away).
In these plots the total travel range was about 3 inches, just trying to hold, lift or let down a 100lbs load.
I gradually reduced the output filter all the way to its minimum value to try and apply some differential gain. Note that the last plot was under a 0.01filter, but the slightest value of diff gain will make the oscillations worse.
In fact, the slightest increase of any gain will pull the system out of equilibrium. All the plots were under a 0.8 feedback filter.
The valve size, age, and position will all play a factor. I would like to see some plots without any output filter so I can better tell how the systme responds. Do you have the part number of the valve? I would like to look up the datasheet to see if has an overlapped spool, etc.
It does appear that there is an overlapped spool or hysteresis or something that is making control difficult.
The valve is a HR Textron 27G50F-5A02. I was speaking to Donald yesterday, and he actually knows the setup we have here and he did say it was a difficult valve to work with
Tuning_Force_Oscillations_Unfiltered_5.rmcplots (312 KB)
Tuning_Force_Oscillations_Unfiltered.rmcplots (2.96 MB)
Have you applied any feedback filtering yet (Axis Parameters, All tab, Secondary Feedback, Filtering/Modeling, Force Filter)? This will probbaly help quite a bit for tuning. The feedback is way too noisy right now. Start with 100 any go down if necessary until it looks somewhat reasonable.
The previous plots I posted had feedback filtered down to 0.8. The attached plot is filtered down to 3 and 10 (force and force rate).
Tuning_Force_Oscillations_6.rmcplots (501 KB)
Filtering that heavily causes a large phase delay in the feedback signal which leads to instability. Can you set the filter to 100, 50, and 30 and show me what the plot looks like?
I will post them within 10 min.
By the way, thank you for you continued support!
I guess I have to say I feel sorry for you having to try to make this work. There appears to be so much noise, non-linearity and delay in the system that it will be very difficult.
Just a couple more questions if you dont mind:
Do you think that the large valve is the main cause for the non-linearity?
Were do you identify the delay in the system looking at the plots, and what can be causing it?