Force control circuit

[kcj]

Several years ago there was an article in Machine Design about 'Cobra Technology' valve in concept/development. I can’t find anything currently on the web. The concept was 4 (I think) PWM valves in a wheatstone bridge arrangement. There was a small damping volume on each port outlet to reduce the pulsation ripple. By timing of the valves, each port opens to pressure or tank and varies the outlet flow or pressure.

Advantage, could work as a flow or pressure control device (with proper feedbacks) and could provide independent control of each port pressure. With a normal servovalve, the port pressures/flows are determined by hard machining of the spool cut.

Disadvantage (probably fatal flaw), at that time was lack of sheer controller horsepower to make the fast switching motions required.

I have been recently searching for the Cobra valve info, and will contact the UM professor on his valve concept.


My application is an existing force control circuit. Module is about 800 lbm, and has hydromechanical natural frequency of about 2 to 3 Hz. Cylinder is 1.5B 1R.

Currently we use two methods:
1. Nozzle/flapper differential pressure control valve that operates at low pressures (300-500 psi) and works very well on force control, but does not have enough port pressure to lift the cylinder off the work surface. A separate valve is required for lift.

2. Flow servovalve (100+ Hz) with pressure transducers on the outlet ports (load cell is not an option) to make it into a force circuit. For scaling and summing pressures for unequal piston areas, some modules use individual digital cards, and some use Peter’s digital controllers. (Both are 1 to 3 ms update time I am told). This version has full pressure/force to lift the cylinder up when necessary. It has barely adequate force control, although I think the valve is too large. However, the actual values of each port pressure are quite high.

The unequal piston areas means the closed side absolute port pressure is applied to an unbalanced rod area tending to push the rod out.

Example: With 250 psi on the closed end, and 150 psi on the rod end, there is 300 lbf pushing. With 625 psi on closed end, and 825 psi on rod end, there is 300 lbf pushing. Both are theoretically the same force, but the port pressures in the second scenario create much higher rod and piston seal frictions, thus the force control is not as uniform.

The appeal to me for the PWM style valve is that, at least in my dreams, I could have port pressures low or high as desired. Low pressures for force control and minimizing seal friction (probably holding one port at steady pressure and only varying the other port), but high pressure on the port for raising the module. There is also a rare scenario where high pressure in both ports is desirable.

My questions are:
1. Anyone heard of the ‘Cobra’ technology? Was it a typical great idea that didn’t pan out?

2. The servovalve/transducer circuit is not very steady. It seems to me the valve and controller are more than adequate (although valve seems to large), and the pressure loop is the fastest response. Granted the hydromech nat freq is low, but I think that the control algorithms are an issue. I suspect the control loops are not truly PID, and also are not symmetrical whether over pressure or under pressure. Any ideas?

I will also post separately a similar situation on propel system pressure control issues. Both scenarios are similar, I think controls related, and I think also relate to my lack of understanding of how digital controllers differ from the ancient analog closed loop amplifier cards.


kcj

 

[swn1]

Digital control loops are inherently "not PID". The theory behind discrete-time (sampled) and continuous-time controls are *analogous* but not identical. If you are applying controls you should be aware of the impact of delay on stability in analog controls. One of the differences is that delay is inherent in the "compensator" stage of a digital closed-loop control.

Example of delay: temperature controler on a hot oil loop, with the sensor at the "sweet spot" in the process tens of feet downstream from the heater. The "engineers" who did this application didn't understand why their "self tuning" PID temperature controller couldn't handle it.

If the 2-3 Hz refers to the application system, it is likely that the valve has some faster modes. A 3ms loop with some inherent filtering delay is pushing the envelope pretty hard, especially if the underlying algorithm is a simple PID emulation.

Also, you should be aware that the whole body of classical control theory (continuous and discrete) is based on the assumption that all the elements in the system are linear. (technically, their input/output characteristics have to obey the superposition priciple.) Even if the valve's response curve is close enough to linear, there are serious nonlinearities at the full open and full closed positions!

I don't know hydraulics all that well but I've seen a lot of projects founder for lack of a digital-aware controls engineer.

The "bridge" style valve idea sounds like someone tried to directly translate the very effective idea of the switching mode power supply from electric to hydraulic power.

Unfortunately, the dynamics and economics are very different. For example, inexpensive devices will switch electricity on and off in microseconds with power gains in the hundreds instantaneously and millions on average.

Also, the typical electical application of this idea depends on TWO kinds of energy storage devices, capacitive and inductive. Capacitors are analogous to receivers (potential energy storage) while inductors store kinetic energy(*). The best analogy to an inductor would be a positive displacement pump/motor coupled to a flywheel. Unless the load has a lot of inertia you would have serious trouble making it (the switching mode concept) work well in a hydraulic application.

* OK, inductors store energy in the form of a magnetic field but the electrons are flowing constantly. It is an exercise in semantics to say whether it is potential or kinetic energy technically.

 

[kcj]

*****Back on the board. Tks for reply.

Digital control loops are inherently "not PID". The theory behind discrete-time (sampled) and continuous-time controls are *analogous* but not identical. If you are applying controls you should be aware of the impact of delay on stability in analog controls. One of the differences is that delay is inherent in the "compensator" stage of a digital closed-loop control.
*****I’m passing this to our controls people. Can you elaborate on the compensator stage? In hydr world, that’s pump control, I assume totally different meaning.

Example of delay: temperature controler on a hot oil loop, with the sensor at the "sweet spot" in the process tens of feet downstream from the heater.
*****That is clear. Delay would relate to flow rate of the system compared to piping and volume and distance, and could be very slow response.

If the 2-3 Hz refers to the application system, it is likely that the valve has some faster modes. A 3ms loop with some inherent filtering delay is pushing the envelope pretty hard, especially if the underlying algorithm is a simple PID emulation.
*****2-3 Hz is the natural frequency at which the mechanical and oil and hoses will oscillate. I.e. must be excited and controlled slower than that speed. The valve is 100+ Hz for 90d phase lag, so it should not be limiting.
*****I thoroughly understand Jack Johnson’s formulas on servovalve speeds and pressure drops, but when he gets into relating nat freq to maximum gains, into the electrical world, I am lost.
*****How does the required update time of the plc ralate to the desired response of the mechanical system? In analog days, I’ve heard rules of thumb, 3 to 5 times, for resolution and nat frequency for each control element. Valve response vs. feedback resolution vs mechanical system nat freq vs. required system accuracy.
*****I naively assumed/told that 3 ms update = 300 hz controller = much faster than 100 hz valve and 3 Hz mechanical. I suspect not nearly that simple as 1000/3 ms = 333 Hz?


Also, you should be aware that the whole body of classical control theory (continuous and discrete) is based on the assumption that all the elements in the system are linear. (technically, their input/output characteristics have to obey the superposition priciple.) Even if the valve's response curve is close enough to linear, there are serious nonlinearities at the full open and full closed positions!
*****Agreed, especially pressure rise with valve signal about center. Will be quite non linear across the spool motion, so the focus is on that region right off center. The existing valve is 10 gpm/1000psid. My calcs indicate it should be closer to 2.5 gpm/1000psid. I am looking also at a two stage gain, very slow near center then increasing at far ends of spool. The non linearity at that transition point would dirve controls crazy, but with proper choice, we won’t operate there. Force control mode operates very close to center, and fast raise operates almost at full lspool stroke, so I think the non linearity at 50% area won’t matter.


I don't know hydraulics all that well but I've seen a lot of projects founder for lack of a digital-aware controls engineer.
*****I know hydraulics very well, and I loudly echo your comment. The lack of awareness of both worlds has sunk many projects. I have two eng’g degress, but both mechanical. Trying to learn the controls world, or better said, how to talk intelligently to the true experts in the controls world who do not understand hydraulics.

*****I need to mull and discuss further. Tks so far. kcj

 

[swn1]

In the textbook analysis of a closed loop control system you have a process, an actuator, a sensor and a compensator. Ie, the controller.

The frequency response of the valve may or may not be a problem, but here's the rub. When you use a discrete-time signal processing system you are bound by Nyquist's theorem, the one that says you have to sample at *least* twice as fast as the highest frequency component in the signal.

300Hz sampling is NOT "much faster" than a 100Hz process component. It is just over 1.5 times faster than the Nyquist limit, barely enough margin for an aggressively designed anti-aliasing filter. With margins that narrow the phase shift introduced by the filter will have to be taken into account too.

We don't know that this is the problem, of course. But it very well could be.

PID controllers suffer from basic structural limitation in that they can only compensate for one level of integration or differentiation in the process. If you want to control speed by positioning a throttle, the system response will include an integral. If you want to control position by positioning a throttle, the system will have a double integral and your PID controller won't do very well. Likewise, if your speed control output command controls throttle positioning motor speed rather than throttle position.

Anyway, I don't know why you are controlling pressure, largely because the details you gave are meaningless to me. I suspect the load and fluid dynamics are pretty complicated if you detailed it all out. Are all those effects negligable? I don't know.

What other programming options do you have for the controller? Have you tried hooking up an analog controller? Have you tried controlling it manually? Can you characterize the instability in more detail?

When the electronic controller looks out at the world it sees a system of differential equations. If you can translate your sysyem into a block diagram with inputs and outputs of each block related by differential equations, the sparkies can take it from there.

If the equations aren't linear and first order, it may be difficult, though