hydraulic closed loop with dead time

[0vidiu]

I have a project where I need to use a proportional valve to control hydraulic system.
Before, I used open loop and now I try closed loop PID with feedforward velocity and acceleration.
In open loop the maxim velocity depends with load, in closed loop this is ok. But in closed loop take some time to find stop position.
The accuracy of stop position is +7 mm maxim, ok for me.

In my PLC I writing a simple trapezoidal trajectory generator and I use feedback position PID control with feedforword velocity
and acceleration.



My loop is:

PLC Modicon M340 Schneider with time execution code 5 ms
analog output 0-10V, update 1 ms
proportional Valve 4WRKE Rexroth https://www.boschrexroth.com/ics/content/UpToDate/PDF/re29075_2013-08.pdf
hose size DN 16, hose lenght aprox. 4 m between proportional valve and cylinder
cylinder Area 1=31 cm^2, Area 2 = 15 cm^2, stroke aprox. 1 m, (flow 70 l/min, presure 120 bar)
load 100 to 300 kg
feedback position is rotary encoder XCC3510PS84CBN and bus Can Open update 2 ms with system resolution 2 count / mm



I have attached a plot. Some problems I have with my system tuning:
The system has dead time and can't track very well position.

1) For different stroke the system needs readjusting parameters for optimal settlig time.
I use Velocity feedforward , Proportional, Deadband and little Acceleration feedforward .
I've tried to use Integral term for reduced settling time, but Integral needs readjusting with stroke. I don t use Derivativ.
Any suggestion ?

2) The system has need different gains for Acceleration feedforward and Deceleration feedforward?


Any ideas what can be improved? I would appreciate any help with this issue.
Thanks in advance!

 

[Peter Nachtwey]

The hose length of 4m is a killer.
Is the BoschRexroth valve a P or L type? It had better be the L type but this is not your main problem.

The valve should be mounted ON THE CYLINDER. If not then pipe should be used between the valve and the cylinder, NOT HOSE! Hose expands and contracts with pressure changes so the oil volume will change here first before moving the piston. Also, you are right about the dead time. It takes 1 millisecond for the pressure waves to travel 1.2m although the delay looks longer than that. It is best to mount the valve on the cylinder. Also, is there an accumulator? Hydraulic pumps are very slow at responding to pressure changes. If you are using a pressure compensated pump the swash plate will move slowly relative to the move you want to make. An accumulator will store oil under pressure so the hydraulic system will respond quickly.

The acceleration and deceleration feed forwards should be the same. It looks like you may have the P type spool which is bad.
The extend and retract feed forwards should be different if the piston areas each side are different.

2 counts/mm is not very good. Typically you should be using Balluff or Tempsonic rods with 5 micron per count resolution or better. The feed back is also slow and not synchronous to the scan and that makes using the derivative gain difficult.

My hat is off to you for getting this far with what you have got but you are fighting a tough battle.
Get rid of the hose and at least half your problem will go away.
Most hydraulic designers do not know how to design a servo hydraulic system.

 

[osmanmom]

additional info,

http://deltamotion.com/pdf/technical/tn_dos and donts of hydraulic system design.pdf

http://insidepenton.com/machinedesign/AdvHydMotionDesignPractices-Paso.pdf

 

[0vidiu]

Is the BoschRexroth valve a P or L type?
L type.

The valve should be mounted ON THE CYLINDER.
I know but the project is very ... very ... very difficult , rather impossible to change, because is already builded.


It takes 1 millisecond for the pressure waves to travel 1.2m although the delay looks longer than that.
For 4 m hose result 5 ms delay. My system has 100 ms delay. Why?


Also, is there an accumulator?

The project comprise 4 hydraulics movements with 4 valve 4WRKE each. One side of project include tank, pump, accumulator, valves block, and another side is machine with hydraulic cylinders.
Both sides are linked by hoses 4 m. This solution is easy for services intervention.


Acceleration and deceleration feedrorward?
In my plot to the same settings and the same gain acceleration and deceleration , the system is different. More gain for acceleration is good, but affects deceleration .


The extend and retract feed forwards should be different
Yes, for extend and retract, I have different feedforward gains.


Now I'm looking for improvement of the system in the closed loop who is more stable but is slower than open loop, otherwise I have to do step back in open loop.