I think this is a very tricky problem. In the graph, the black line represents force, the yellow line represents the target force, the red line represents displacement, and the blue line represents velocity. When the force enters the holding phase, the specimen being loaded softens due to heating, causing the force to suddenly decrease until the specimen cools and hardens, due to the change in specimen stiffness. How can this variation be compensated for? I originally considered compensating it into the control output by multiplying the actual velocity by a coefficient, but as can be seen from the graph, the velocity signal is too noisy and definitely cannot be used. Are there any other methods?
The valve flow rate is approximately 1000 L/min, with a working pressure of 21 MPa. The hydraulic cylinder has an inner diameter of 1200 mm and a rod diameter of 900 mm. Even a slight increase in the proportional coefficient causes oscillation.
This is indeed a challenging system. In general, when a system changes so drastically, a feed forward is of the best help. You already went down this road with the velocity, which given its noisiness relative to the displacement range in the pertinent area, didn’t work.
I see the position resolution is 5 micrometers. What sensor do you have? Is it programmable to 1 micrometer? This could perhaps improve the velocity noise.
Another theoretical option is a feed forward based off of temperature, but I imagine it’s impossible to get a temperature sensor accurate enough and fast enough to help, and whether the system behaves very consistently relative to temperature is something I do not know at all.
It seems that the system’s tendency to oscillate is a serious impediment. I am a bit surprised that there is so much oscillation given the wide cylinder bore, which usually means a very high natural frequency and low tendency to oscillate. Is the valve mounted far from the cylinder? This could cause a tendency to oscillate, especially if the valve is connected via hose.
Thank you for your reply. I am using a grating ruler with ABZ signal, with a resolution of 5 micrometers, it cannot programmable to 1 micrometer. There is no temperature sensor in the system, Only two pressure transducer can used. The oscillation in the early stage of the force curve is likely caused by an improper switch to force control using the Hold Current Force command. I should add an integral preload as a transition buffer. For the force fluctuations afterward, I plan to try using the Actual Force Rate as a compensation component to see if it can improve the current situation.
I agree that changing the method of entering force should reduce the oscillations at the beginning. The Hold Current Force will certainly cause problems. Because the Actual Force is rising smoothly at the point the Force Control is entered, I recommend using one of the Enter Pressure/Force commands. The Enter Pressure/Force (Auto) command does the smoothest transition the most easily, but should only be used if the entering situation is known and fairly consistent. The Enter Pressure/Force (Rate) can be very smooth if you match the Actual Rate at the time of entering, but it’s difficult to get the Rate and Accel Rate command parameters set correctly - the Accel Rate generally needs to be very high. The Enter Pressure/Force (Time) ensures the Target Force reaches the desired force in a given amount of time, but the transtion may be a tad sharp.
One thing not covered yet is the speed of your pressure transducers. If they are slow your system will suffer greatly. They should be rated for a total rise time of around 1 msec.
Also, I don’t think you answered Jacob’s question about how far the valve is from the cylinder? Distance here is a killer for any system, let alone a difficult one.
Lastly, you mentioned additional proportional gain causes oscillating. just to be clear, are you talking about force proportional gain or position proportional gain?
The response time of the pressure transducer is 1 millisecond. The signal is very noisy, making it difficult to increase the proportional gain of the force closed-loop. Therefore, I sacrificed response speed and added 20 Hz input filtering.
The valve is approximately 3 meters away from the cylinder.
Regarding the proportional gain mentioned by Jacob, it specifically refers to the proportional gain of the force closed-loop.
This does look better now, but I suspect we still have the same issues with the control during the softening of the test specimen. Filtering the velocity will not have any effect because the axis is in force control and is not using the velocity during that time.
For the force, I see you are using the Output Filter, which is very good. Are you also using feedback filtering? Feedback filtering is challenging because it adds phase delay, which cause problems with the control, so we normally recommend using the Output Filter heavily, like you are doing, and then only if necessary, maybe a little force feedback filter, such as 50 to 100 Hz cut-off or larger.
Is the valve connected to the cylinder via hose or hard pipe?
A 20 Hz feedback filter is being used. Previously, a 50 Hz feedback filter was tried, but the system oscillated violently, so the filter was reduced to 20 Hz. The servo valve is connected to the cylinder using a hard pipe with two elbow joints.
A 20 Hz filter provides more filtering than a 50 Hz filter and also causes a larger phase delay than a 50 Hz filter. 20 Hz may be causing enough delay that it slows the response.
Which RMC are you using? The RMC200 has various feedback filter options that can filter heavily with less phase delay.
I don’t see things moving that fast on the force side. If I had some material I wasn’t worried about damaging I’d push things a bit. 5Hz force output filter, 10Hz force feedback filter and then see if I could increase the force proportional gain to try to keep the force on the product more consistent.
