For the purposes of this series of articles we will use databased on a symmetrical, double-ended actuator with a total stroke ofaround 100mm and an active piston area of around 1 inch. Piston massmay be specified by the manufacturer, or can becalculated from a knowledge of the geometry of the actuator. For apiston of this size, a mass of around 9 Kg would be typical.
The DSP controller used in this design is a TMS320C28xDSP with a 32-bit fixed-point DSP core, a low latency interruptmechanism, and an instruction set including “atomic” instructions.Pertinent information on the actuator and servo-valve is summarized in Table 1 and Table 2 , below .
|Table1: Actuator data|
|Table2: Servo-valve data|
Other useful data is derived from knowledge of the system. Leakageand frictional effects could be modelled using coefficient valuesestimated from empirical data, as shown in Table 3, below.
|Table3: Miscellaneous system data|
Matlab Simulation results
A simple test which yields useful information about the performance ofthe system is to apply a small step signal to the input and monitor theresponse. In Simulink, thiscan be simply achieved using a 'scope' block to monitor command andresponse signals.
The vertical axis is graduated in volts measured at the erroramplifier. The controller is scaled to a range of +/- 10V, so for a100mm stroke actuator, a step input of +2V corresponds to a pistondisplacement of +10mm from the central position.
Controller gain terms are adjusted and the step test repeated totune the actuator response as required. The plot shown in Figure 12 below represents asatisfactory compromise between rise time and overshoot, and thecorresponding controller settings would serve as a useful startingpoint for the control engineer whentesting the real system.
|Figure12. Simulated Actuator Step Response|
By configuring the actuator chamber pressures as test points inSimulink, a 'floatingscope' block may be used to monitor these andother signals during the step response test.
The graph below in Figure 13 shows the behavior of the chamber pressures during the step responsesimulation. Chamber B pressure (P b ) is shown as a continuous line, andchamber A pressure (P a )a dashed line. The vertical axis is in Pascals (N/m ). The slight asymmetry resultsfrom the change in chamber volumes as the piston is displaced to itsnew position.
|Figure13. Step Response of Piston Chamber Pressures|
Figure 14 below shows thestep response of a real high-performance linear actuator withdimensions similar to those describe earlier. The sharp rising andfalling edges and minimal overshoot represent the optimum response thatcan be obtained with a PID control strategy and a good qualityactuator.
|Figure14. Step Response of High-Performance Linear Actuator|
The dynamic performance of the system is limited by the capacity ofthe hydraulic power supply as well as the performance of theservo-valve. This is illustrated by Figure15 below , in which a large sinusoidal command input is appliedand the frequency gradually increased.
|Figure15. Frequency Response of High-Performance Linear Actuator|
The curve shows two high frequency asymptotes: the first occurs atabout 8 Hz and is caused by the limited flow capacity of the powersupply. The second, at about 40 Hz, is the high frequency responselimit of the servo-valve.
The importance of friction in a high performance actuator isdemonstrated by the following two displacement/timegraphs below . Both showthe behavior of a linear servo-actuator when subjected to alow-frequency, low amplitude sinusoidal command input. The first in Figure 16 showsa low friction actuator, and the second in Figure 17 illustrates an actuatorwith higher frictioncaused by tighter piston and end cap oil seals.
|Figure16. Low Frequency Test on Low Friction Linear Actuator|
|Figure17. Low Frequency Test on Linear Actuator with Significant Friction|
Response losses caused by friction in the actuator can be reduced tosome extent by the addition of “dither” to the controller output. Thisis a relatively high frequency, constant amplitude oscillation, whichkeeps the valve spool in constant motion and reduces the break-awayforce needed to overcome anystatic friction present in the system.
The use of Simulink tomodel anddesign the system allows enhancements such as this to be easily addedto the digital controller.
The principal non-linear effects in hydraulic systems arise from thecompressibility of hydraulic fluid, the complex flow properties of theservo-valve, and internal friction in the actuator. These depend onphysicalfactors which are difficult to measure accurately and for this reasonsimulation results should be supported by experimental testing wheneverpossible.
Conventional feedback control techniques work well in cases wheredead reckoning of the above factors is possible or where theirinfluence is sufficiently small that they can safely be ignored.However, for true high-performance control advanced digital controltechniques are required and in this the DSP excels.
The performance of a high quality hydraulic actuator is verydependent on the servo-controller and DSPs lend themselves well toimplementing real-time control algorithms necessary. Moreover a DSPallows the designer of a PID or servo controller to implement advancedcontrol strategies, including multi-variable and complex controlalgorithms using modern intelligent methods such as neural networks andfuzzy logic.
Also available is the ability to perform adaptive control, in whichthe algorithm dynamically adapts itself to match variations in systembehavior. A DSP-based PID controllerallows the developer to implementcomplex topologies such as multi-axis control where synchronization ofmultiple force patterns is required and perform diagnostic monitoring,including frequency spectrum analysis to identify mechanical vibrationsand predict failure modes.
As pressure grows for faster time to market of new products withmore stringent safety features, increasing emphasis is being placed onthe use of high-level tools and software for application development.These afford a level of abstraction from the processor core, and permitrapid applicationdevelopment and re-use of material.
For complex control applications, the ability to simulate controllerand plant behavior at the design phase is invaluable, and the use ofembedded auto-code generation and validation features affords thedesigner the ability to move quickly and easily from simulation toprototyping. This is a powerful advantage and we may expect the trendtowards integrated system simulation and code development to continuein the coming years.
To read Part 1 go to “Thebasics of electro-hydraulic servo actuator systems.”
To read Part 2, go to “Usingmodelling tools to simplify hydraulic PID system design.”
Richard Poleyis Field Application Engineer at TexasInstrumentswith focus on digital control systems .
1 . Richard Poley, DSP Controlof Electro-Hydraulic Servo Actuators, TIApplication Report SPRAA76
2. Herbert E. Merritt,Hydraulic Control Systems, Wiley, 1967
3. M.Jelali and A.Kroll,Hydraulic Servo Systems – Modelling,Identification & Control, Springer, 2003
4. Electrohydraulic Valves ATechnical Look, Moog Technical Paper
5. Moog 760 Series Servovalves,product datasheet
6. R.H.Maskrey and W.J.Thayer,A Brief History of ElectrohydraulicServomechanisms, Moog Technical Bulletin 141, June 1978
7 . T. P. Neal, PerformanceEstimation for Electrohydraulic ControlSystems, Moog Technical Bulletin 126, November 1974
8. W.J.Thayer, TransferFunctions for Moog Servovalves, Moog TechnicalBulletin 103, January 1965
9. J.C.Jones, Developments inDesign of Electrohydraulic ControlValves, Moog Technical Paper, November 1997
10. D.C.Clarke, Selection &Performance Criteria forElectrohydraulic Servovalves, October 1969
11 . M.D'Amore andG.Pellegrinetti, Dissecting High-PerformanceElectrohydraulic Valves, Machine Design,April 2001
12. D.DeRose, The ExpandingProportional and Servo Valve Marketplace, Fluid PowerJournal, March/April 2003
13. D.DeRose, Proportional andServo Valve Technology, Fluid PowerJournal, March/April 2003
14. D. Caputo, Digital MotionControl for Position and Force Loops NFPATechnicalPaper I96-11.1, April1996
15. B.C.Kuo and F.Golnaraghi,Automatic Control Systems, Wiley, 2003
16. J.Schwarzenbach andK.F.Gill, System Modelling & Control,Edward Arnold, 1992
17. J.B. Dabney and T.L.Harman, Mastering Simulink, Pearson PrenticeHall, 2004
18. Digital ControlApplications With The TMS320 Family SelectedApplication Notes (SPRA019)
19. Implementation of PID andDeadbeatControllers with the TMS320Family ( SPRA083)
20. Gene F. Franklin, J. DavidPowell & Michael L. Workman, DigitalControl of Dynamic Systems,Addison-Wesley, 1998
21 .TMS320F2812 Data Manual(SPRS174)
22. Signal Conditioning an LVDTusing a TMS320F2812 DSP (SPRA946)