Hi folks, We're working on the d

[Mitchell21]

Hi folks,

We're working on the design of a system that uses four hydraulic cylinders for lifting a payload. The cylinders stand on the ground and the payload is not otherwise guided. Because of the distance between the cylinders, they're not hydraulicaly coupled - each is driven by its own small, fixed displacement HPU. The hydraulic system's max operating pressure is around 200 bar (~3000 psi) and the flow rates are quite low at 2.5 liter/min (~0.7 gpm). The payload is not evenly balanced. There could be a pressure difference of 20% between the four cylinders while operating.

Ideally we want each cylinder to extend and retract at the same rate (and as fast as possible) so the payload rises evenly. The default solution is to put a position measurement system in the cylinder and 'close the loop', but before committing to that, we're trying to think outside the square.

We have a bit of tolerance to play with. From the size and flexibility of the payload, we estimate that we can allow a difference of up to +/- 25 mm (1") over the full cylinder stroke which is 1.75 meters (~70"). That's about +/- 1.5%. Any accumulated error would be reset at the top of the stroke and when the payload comes back to the ground.

One option we thought about was to leave the hydraulic system 'open loop'and use pressure compensated flow control valves tuned to give matching flow rates. The info we've found suggests the accuracy may not be good enough (e.g 3 to 5%). We're also wondering if the valves could go out of adjustment relative to one another. Any comments on this option?

Alternatively, would it be possible to "close the loop" on the flow-rate to and from the cylinders' piston-side, using a flow-rate sensor? This wouldn't detect leakage over the piston seal (which should be negligible anyway) but it would make the HPU a nice, self-contained 'smart' package. The four HPU's could talk to each other, and match their speeds to the slowest cylinder. Feasible? I found the VS sensors from vse-flow.com which might be suitable. Any others sensors that people can recommend?

I should add that this system will be working in an indutrial environment. It needs to be cost effective and robust. I look forward to all comments and suggestions!

 

[Oldhydroman]

Have you considered trying to match the flows from each of the HPU's?

If you use a quality piston pump (possibly a small radial piston unit) and drive it with an induction motor coupled to a simple speed controller then the flow output from each HPU should be exactly the same. There will be small manufacturing tolerances in the pump which may affect the output flow slightly, but if you have a speed controller on each motor then you can trim this out.

The flow will not vary massivly with pressure (fluid compressibility is about 1/2% per 70 bar) so, if you use leak-free valves in your circuit, then the speeds of the cylinders should be tolerably close.

Stroke transducers on the cylinders are a possible refinement and you can keep the control algorithm simple by utilising the existing positional tolerance and applying a "set-point supervision" technique rather than a full blown PI position control loop. Pick one cylinder/HPU to be the master and make each other cylinder/HPU compare its own position to the "master". If more than 10 mm out of sync then perform a small [automatic] adjustment of the speed of the drive motor for that HPU. Keep the new [adjusted] speed in play until the tolerance is within an inner band, say ±5 mm, and then go back to the "pre-trimmed" speed.

Although it's nice to put the stroke transducers inside the cylinder, lots of industrial applications use external pull-wire sensors ("string pots") and find them adequate. A while ago I came across an encoder version of the "string pot" which would give you a relaively easy way of achieving your desired positional accuracy because you would be working on a straightforward count of pulses rather than comparing analogue voltage (or current) values.

DOL

 

[Mitchell21]

Thanks for your reply Oldhydroman.

Sorry, I unintentionally misled when I mentioned it was an industrial application. It's actually a mobile application and each pump will be driven by a small 50cc internal combustion engine. We looked into DC motors for the HPU, but the electric-to-hydraulic efficiency isn't good enough and the currents are too high. I don't imagine trimming the flow-rate via the throttle would be straight-forward. I guess a servo on the throttle could regulate the speed, but the lag could be a killer.

Alternatively, what about trimming the flow with a hydraulic component e.g. a proportional throttle valve (normally open), or a 2-way flow control valve, or a 3-way flow control valve that bleeds off excess flow? (Can these components regulate flow from either direction?)

I have seen those pull-wire sensors and had them in mind, but not the encoder version. Thanks for the tip. And your suggestion for the control algorithm sounds good.

I'm interested in comments on the flow-rate sensor v's cylinder position sensor. I'm exploring the flow sensor solution as it would be mechanically robust, and it eliminates the need for an electrical connection between the HPU and the the cylinder. The same control algorithm approach could be used, but counting pulses from the flow sensor rather than the position sensor. (The flow sensor output comes from gear teeth passing a hall effect sensor, as they're rotated by the passing flow. The manufacturer claims an accuracy of ±0.3%.)

 

[Oldhydroman]

If you used a positive displacement flow meter on each pump outlet then the pulses you get from the device are sort of equivalent to the pulses you would get from a "string-encoder" on the cylinder. The compressibility of the fluid can probably be ignored, but you need to ensure you have leak-free valves in the circuit to avoid unnecessary errors.

For your simple "adjustable" speed control you could use a 2/2 [normally closed] solenoid valve connected between the pump outlet and tank - with a small "bleed-off" needle valve in series. Then set the "master" HPU to be a little slower than all the others, but set the opening of the "bleed-off" needle valves so that when the bleed-off solenoid was enegised the flow from the slave HPU's would reduce to a little less than the master HPU.

Your control algorithm could go like this: start all cylinders running together by energising the appropriate DCV solenoid on each HPU - but the master cylinder would be a fraction slower than each of the others. When any of the slave cylinders reached, say, 10 mm ahead of the master, the bleed-off solenoid valve would energise and that particular cylinder would slow down a bit and allow the master to catch up. When any cylinder was, say, 10 mm behind, the bleed-off solenoid valve would be de-energised and then that slave cylinder would start to catch up again. You could also build in an outer control loop so that if the cylinder got too far ahead you de-energised the main DCV for a few seconds. By this means you have a fine control and a coarse control of the cylinder speed. I know it's coarse but it's a scheme that I've used before and it's quite robust.

If you were unsure about the speed regulation of your little engines you could use a three-way pressure compensated flow control valve on each pump outlet: then you wouldn't need a fancy radial piston pump either - you could manage with a robust [and forgiving] gear pump (although your fuel efficiency would be lower and, depending on the duty cycle, you might need a cooler).

DOL

 

[kcj]

Calling doctor Peter N.

I don't think you will be accurate enough with the open loop ideas.

Closing the loop on flow has the advantage as you say of keeping wiring away from the ends of the hoses and cylinders and at a safer spot. Integrating flow vs time would give you close position, but it is still open loop to the actual load. Are the cylinders quite tight, and will they get enough use that they will later have worn seals and leakage?

Depending on engine rpm alone, without other feedback, will be unpredictable. As the load increases on one unit, the governor droop causes an engine rpm droop thus flow drops off. You could measure engine rpm and integrate that with time to estimate volume sent to the cylinder, but even that does not account for pump efficiency. If instead you are measuring flow, then there is no real need to depend on engine rpm to estimate flow.

With the limited info available, here is what I would consider:
-Feedback inside each of the 4 cylinders. MTS/Temposonics preferred, or linear pots maybe.

-In rough environments, the cabling and connectors are the weak points for us. Far more 'bad cylinders' removed are electrical issues, not sensor issues. To deal with the external abuse, run a small single wire hydraulic hose, with crimped on flare or face seal ends, from a bulkhead fitting mechanically anchored on the cylinder, back to similar bulkhead connnection at the power unit. Run the signal cable from the feedback through this small hose, and bundle it with the two cylinder hoses, also pretty small I presume. Now the signal line is mechanically protected inside this hose.

-Obviously anything closing the loop needs some sort of controller device, but that is common to all the ideas on the table.

-At this low flow, and with +- an inch of tolerance allowed, maybe you don't need a proportional valve at the power unit. Just bang/bang on and off the solenoid valve when the slaves are out of tolerance with the master unit. This assumes that the on/off pulses and spikes don't cause a dynamic issue with the load being lifted.

-Does the load need counterbalances or lock protection at the actual cylinder? What if one of these hoses fails? CB adds difficulty to a control circuit because it is not linear about 0. Takes a certain pressure rise to open the CB or PO check.

 

[Mitchell21]

Thanks for the replies. I appreciate the good suggestions.

I agree that an open loop system will not give reliable performance. We are setting up a test system, so we will close the loop and try both the flow measurement and position measurement options. The rest of the system and the method of control (i.e. the bleed-off approach and algorithm) could be the same. The three-way pressure compensated flow control valve on each pump outlet would help too, Oldhydroman.

Measuring the cylinder position directly is certainly more accurate and conventional. I like the idea of protecting the sensor cable from the cylinder with a hose and fitting, however in our case, the HPU and cylinder need to be disconnected for transporting. So a robust plug and cable would be necessary.

In comparison, I know that deriving the cylinder position by measuring and intergating the flow is open to error, especially as things wear. We would have the flow meter as the last component on the HPU before the QD coupling, so only flow to the cylinder would be counted. There will need to be a safety valve on the cylinder which is a potential leakage path, and the piston seal is another. I'll try to quantify the error this may cause. Given the tolerance we have - maybe we'd get away with it.

I take your point KCJ about the non-linearity that POCVs or CB valves will introduce. I'm hoping that the slow, quasi-static nature of our system will help us out. I'll find that out soon enough!

Thanks again guys for your advice.

 

[Oldhydroman]

Although it is a little unconventional to use integrated flow as a substitute for direct position measurement, you would not be the first to do this. It is a not-uncommon practice in sub-sea applications and also in steelworks applications. And it is done for the exact same reasons as in your application: to avoid having a stroke transducer in the cylinder.

The CB or POCV valves definitely create a discontinuity in the relationship between pressure and force but not so much between integrated flow (volume) and cylinder rod position. Since you will be counting the volume into the cylinder full bore on the way up, and counting the volume out of the full-bore on the way down then the presence of the load holding valves shouldn't upset your calculations too much.

There is, however, one little wrinkle you should bear in mind. Standard "over-centre" (or "counterbalance") valves often have a "bleed-through pilot" which could have a slight effect on your accuracy. I presume you are thinking of having something like an over-centre valve on the full bore port of the cylinder to hold the supported load up in the air when you no longer want to be lifting or lowering. When bringing the load down this valve is piloted open by a pressure signal from the annulus port. Any air trapped in the pilot line upsets the required rate of change of signal pressure so a typical over-centre valve has a deliberate leak between the pilot port and the outlet port (not the load holding port) to act as a continuous bleed to remove the trapped air and to avoid having a stagnant slug of oil in that pilot line. For you this means that, when lowering, a tiny portion of the full bore flow will have come from the pilot line "bleed through", and, when raising, a tiny portion of the flow you put into the full bore will go back up through the pilot line.

To overcome the inaccuracy this creates make sure you use a load holding valve with a "sealed" pilot ... but do then put in some provision for bleeding the pilot line: a pressure test point, a Safe-T-Bleed point, or a small needle valve connected between the pilot port and the outlet port (which you can use to bleed the pilot line by trying to lower the already fully lowered cylinders once you have connected your QD's).

You might want to consider a form of load holding valve with no pilot assist - this works best if your load doesn't vary much. But such a scheme will not allow the cylinder full bore to completely depressurise when the load is fully down. While the QD's are disconnected the trapped pressure might bleed through and make it difficult to reconnect at some later time. Overcome this by using the type of QD that can be connected under pressure or make sure there is no load on the cylinder before disconnecting.

If you are concerned about dynamic instability with all the bang-bang operation of the solenoids then you might want to look at soft-switching valves, but if they don't all switch at the same rate then you will get an instant position error every time you start or stop the movement. The error may not be significant but all these insignificant errors can eventually add up to something which is significant.

DOL

 

[Mitchell21]

Thanks Oldhydroman. I appreciate your advice. I was thinking about sealed pilot pistons as the way to go. I'll make sure we have a way to bleed those pilot lines. I know that the piston seals we will use are very good at holding pressure. If they ever leaked enough to upset the control system, then there's a bigger problem! Although we're neglecting compressibility, I'm optimistic that integrating the flow will be a reasonable approach. We're preparing to build some hardware, so I'll let you know when we've tried it.

 

[israelkk]

If the payload is not guided it means that the cylinders are fixed to the ground and the piston rods are the actual guides. How you account for the side loading on the piston rods? especially when at fully extended position or close to it? Cylinders are not built to carry side loads. The piston may friction locked due the side loads.

 

[Mitchell21]

The cylinders are custom designed for this application. They have good overlap between rod and barrel at full extension, and plenty of bearing area on the wearbands.

 

[PNachtwey]

25mm over 1.75m is pretty easy to achieve. It is the distance between the cylinders that make this project expensive.
Feedback devices should be used as kcj suggested.

Is it possible to have a small electric variable speed motor drive bi directional pump? Then a hydraulic motion controller like the ones we sell could easily keep 4 axes synchronized. It would be efficient but kind of costly compared to using a gear flow divider.



Peter Nachtwey
Delta Computer Systems

http://www.deltamotion.com