Auto-Calibration and Control Applied to Electro-Hydraulic Valves PATRICK OPDENBOSCH By
Download ReportTranscript Auto-Calibration and Control Applied to Electro-Hydraulic Valves PATRICK OPDENBOSCH By
Auto-Calibration and Control Applied to Electro-Hydraulic Valves By PATRICK OPDENBOSCH Graduate Research Assistant Manufacturing Research Center Room 259 (404) 894 3256 [email protected] April 11, 2006 Sponsored by: HUSCO International and the Fluid Power Motion Control Center MOTIVATION MOTION CONTROL Electronic approach Use of solenoid Valves Energy efficient operation New electrohydraulic valves Conventional hydraulic spool valves are being replaced by assemblies of 4 independent valves for metering control High Pressure Low Pressure Spool Valve Spool piece Spool motion Piston April 11, 2006 Piston motion 2 MOTIVATION MOTION CONTROL Electronic approach Use of solenoid Valves Energy efficient operation New electrohydraulic valves Conventional hydraulic spool valves are being replaced by assemblies of 4 independent valves for metering control Low Valve motion Pressure High Pressure April 11, 2006 Piston motion 3 MOTIVATION Poppet type valve Pilot driven Solenoid activated Internal pressure compensation Virtually ‘zero’ leakage Bidirectional Low hysteresis Low gain initial metering PWM current input April 11, 2006 Coil Cap Modulating Spring Input Current Coil Armature Control Chamber Pressure Compensating Spring U.S. Patents (6,328,275) & (6,745,992) Electro-Hydraulic Poppet Valve (EHPV) Adjustment Screw Pilot Pin Armature Bias Spring Main Poppet Forward (Side) Flow Reverse (Nose) Flow 4 MOTIVATION VALVE CHARACTERIZATION Flow Conductance Kv Kv Q Q K P1 P2 K P 2 V 2 V P1 P2 or Q K V P1 P2 sgn P1 P2 Q FULLY TURBULENT CHARACTERIZATION April 11, 2006 5 MOTIVATION EHPV Forward Flow Conductance Coefficient Measurement 100 1.5044 1.3565 1.2074 1.0584 1.4308 1.2818 1.1326 0.98395 90 FORWARD MAPPING 80 Kv [LPM/sqrt(MPa)] 70 60 50 40 30 20 10 Side to nose 0 0 0.2 0.4 0.6 0.8 1 Pressure Differential [MPa] 1.2 1.4 1.6 1.8 Forward Kv at different input currents [A] REVERSE MAPPING EHPV Reverse Flow Conductance Coefficient Measurement 120 1.507 1.3587 1.2091 1.0594 1.4333 1.2838 1.134 0.9845 100 Kv [LPM/sqrt(MPa)] 80 60 40 20 Nose to side 0 0 0.2 0.4 0.6 0.8 Pressure Differential [MPa] 1 1.2 1.4 Reverse Kv at different input currents [A] April 11, 2006 6 VELOCITY COMMAND TABLE 1.00 0.80 0.60 Velocity [kph] MOTIVATION 0.40 0.20 0.00 -0.20 -0.40 -0.60 -0.80 -1.00 0 1 2 3 4 5 Applied Voltage [V] Obtain (Operator) desired speed, n HUSCO’S CONTROL TOPOLOGY Calculate desired flow, nAB Q US PATENT # 6,732,512 & 6,718,759 Read port pressures, Ps PR PA PB Calculate equivalent KvEQ Determine Individual Kv KvB Hierarchical control: System controller, pressure controller, function controller April 11, 2006 KvA Determine input current to EHPV isol=f(Kv,P,T) 7 MOTIVATION Constant Temperature (T = 30 C) P 120 T Kv [(LPM)/sqrt(MPa)] 100 80 60 EXPERIMENTAL DATA 40 20 Kv 0 1.5 5 1 4 3 0.5 2 1 isol 0 Input [A] 0 dP [MPa] T = 20 C Kv 1.4 1.2 Input [A] 1 0.8 T INTERPOLATED AND INVERTED DATA isol 0.6 0.4 0.2 0 5 4 100 3 P April 11, 2006 80 60 2 40 1 dP [MPa] 20 0 0 Kv [(LPM)/sqrt(MPa)] 8 MOTIVATION Flow conductance online estimation 100 80 60 40 20 Online inverse flow conductance mapping learning and control Effects by input saturation and timevarying dynamics Maintain tracking error dynamics stable while learning 120 Kv [(LPM)/sqrt(MPa)] Accuracy Computation effort Constant Temperature (T = 30 C) 0 1.5 5 1 4 3 0.5 2 1 0 Input [A] 0 dP [MPa] T = 20 C 1.4 1.2 1 Input [A] 0.8 0.6 0.4 0.2 0 5 4 Fault diagnostics 100 3 80 60 2 40 1 dP [MPa] 20 0 0 Kv [(LPM)/sqrt(MPa)] How can the learned mappings be used for fault detection April 11, 2006 9 PRESENTATION OUTLINE FLOW CONDUCTANCE ESTIMATION Reported work Approaches ONLINE FLOW CONDUCTANCE MAPPING LEARNING AND CONTROL Fixed inverse mapping Learning mapping response April 11, 2006 FUTURE WORK CONCLUSION 10 FLOW CONDUCTANCE ESTIMATION REPORTED WORK O'hara, D.E., (1990), Smart valve, in Proc: Winter Annual Meeting of the American Society of Mechanical Engineers pp. 95-99 Book, R., (1998), "Programmable electrohydraulic valve", Ph.D. dissertation, Agricultural Engineering, University of Illinois at Urbana-Champaign Garimella, P. and Yao, B., (2002), Nonlinear adaptive robust observer for velocity estimation of hydraulic cylinders using pressure measurement only, in Proc: ASME International Mechanical Engineering Congress and Exposition pp. 907-916 Liu, S. and Yao, B., (2005), Automated modeling of cartridge valve flow mapping, in Proc: IEEE/ASME International Conference on Advanced Intelligent Mechatronics pp. 789-794 Liu, S. and Yao, B., (2005), On-board system identification of systems with unknown input nonlinearity and system parameters, in Proc: ASME International Mechanical Engineering Congress and Exposition Liu, S. and Yao, B., (2005), Sliding mode flow rate observer design, in Proc: Sixth International Conference on Fluid Power Transmission and Control pp. 69-73 April 11, 2006 11 FLOW CONDUCTANCE ESTIMATION O'hara (1990), Book (1998) Concept of “Inferred Flow Feedback” Requires a priori knowledge of the flow characteristics of the valve via offline calibration Squematic Diagram for Programmable Valve April 11, 2006 12 FLOW CONDUCTANCE ESTIMATION Garimella and Yao (2002) Velocity observer based on cylinder cap and rod side pressures Adaptive robust techniques Parametric uncertainty for bulk modulus, load mass, friction, and load force Nonlinear model based Discontinuous projection mapping Adaptation is used when PE conditions are satisfied April 11, 2006 13 FLOW CONDUCTANCE ESTIMATION Liu and Yao (2005) Flow rate observer based on pressure dynamics via sliding mode technique. Needs piston’s position, velocity, rode side pressure, and cap side pressure feedback Affected by parametric uncertainty in the knowledge of effective bulk modulus April 11, 2006 14 FLOW CONDUCTANCE ESTIMATION Liu and Yao (2005) Modeling of valve’s flow mapping Online approach without removal from overall system Combination of model based approach, identification, and NN approximation Comparison among automated modeling, offline calibration, and manufacturer’s calibration April 11, 2006 15 FLOW CONDUCTANCE ESTIMATION PS APPROACHES Model based Physical sensor INCOVA based Learning based Pump QB- QA+ KvB- M KvA+ PB PA KvP KvB+ KvAPR QB KvT QB+ QA- QA QL FL=0 PB Tank PA m x x VB0 AB AA VA0 Hydraulic Piston EHPV - Wheatstone Bridge used for motion control of hydraulic pistons April 11, 2006 16 FLOW CONDUCTANCE ESTIMATION PS MODEL BASED Object oriented Offline identification Online identification Customization Pump QB- QA+ KvB- M KvA+ PB PA KvP KvB+ KvAPR QB KvT QB+ QA- QA QL FL=0 PB Tank PA m x x VB0 AB AA VA0 Hydraulic Piston EHPV - Wheatstone Bridge used for motion control of hydraulic pistons April 11, 2006 17 FLOW CONDUCTANCE ESTIMATION PS PHYSICAL SENSOR Pump Position sensor Position/velocity sensor Venturi type flow meter Efficiency compromise Sensor safety compromise Design compromise Cost QB- QA+ KvB- M KvA+ PB PA KvP KvB+ KvAPR QB KvT QB+ QA- QA QL FL=0 PB Tank PA m x x VB0 AB AA VA0 Hydraulic Piston EHPV - Wheatstone Bridge used for motion control of hydraulic pistons April 11, 2006 18 FLOW CONDUCTANCE ESTIMATION n INCOVA BASED Relies on expected pressures for given commanded speed PR KvB QB R @ AA AB PB AB QA PA KvA PS Power Extension Mode (PEM) AA Actual System x&cmd AB = K vEQ PEQ K vA = mK vB PEQ RPs PA PB PR K vEQ PEQ K vA K vB K va R 3 K vB 2 2 Equivalent System April 11, 2006 19 FLOW CONDUCTANCE ESTIMATION n INCOVA BASED Relies on expected pressures for given commanded speed PR KvB QB R @ AA AB PB AB QA PA Power Extension Mode (PEM) PS AA KvA Actual System x&cmdAA = K vA PS - PˆA x&cmdAB = K vB PˆB - PR PEQ KEQ Equivalent System April 11, 2006 20 FLOW CONDUCTANCE ESTIMATION n INCOVA BASED Relies on expected pressures for given commanded speed PR KvB QB R @ AA AB PB AB QA PA Power Extension Mode (PEM) æ (x&cmd + e )AA = ççç1 + çè æ (x&cmd + e )AB = ççç1 + çè P = Pˆ + D x&= i i April 11, 2006 i dA ö 2 ÷ ÷ & x A - D AK V2A ( ) ÷ cmd A ÷ ÷ K VA ø dB ö 2 ÷ ÷ & x A + D BK V2 B ( ) ÷ cmd B ÷ ÷ K VB ø x&cmd + e K Vi = K Vi + di PS AA KvA Actual System PEQ KEQ Equivalent System 21 FLOW CONDUCTANCE ESTIMATION LEARNING BASED PS Assumptions: Pump bulk modulus is sufficiently high M Variable volume is sufficiently small. Negligible temperature change Negligible leakage QB- QA+ KvB- KvA+ PB PA KvP KvB+ KvAPR QB KvT QB+ QA- QA QL FL=0 PB Tank PA m Chamber pressure equation dP 1æ ¶P ö ÷ ç = ç ÷ ÷ dt v çè ¶ r ÷ øT é æ¶ r êr (Qi - Qo - v&) - T& v çç êë è¶ T æ¶ r ö dP v çç ÷ » 0 = éër (Qi - v&)ù ÷ û ÷ è¶ P øT dt April 11, 2006 x x VB0 AB AA VA0 Hydraulic Piston ö ù ÷ ú ÷ ÷ øP ú û EHPV - Wheatstone Bridge used for motion control of hydraulic pistons 22 FLOW CONDUCTANCE ESTIMATION LEARNING BASED æ¶ r ö dP v çç ÷ » 0 = éër (Qi - v&)ù ÷ û ÷ è¶ P øT dt Q A = A Ax& Let K = K (x ) @ K VA (isol ) h @ PS - PA Then K h = A A x& Differentiation yields &= AA (u + d) K&h + K h&= AAx& April 11, 2006 23 FLOW CONDUCTANCE ESTIMATION LEARNING BASED &= AA (u + d) K&h + K h&= AAx& &= PAAA - PBAB - FL - f f - mg sin q mx& Let Then FL = 0 q = 0 &= PAAA - PBAB - fˆf - D f mx& f f = fˆf + D f Let &= x& (P A - P A - fˆ ) - D u @ (P A - P A - fˆ ) 1 m A 1 m A B A B A d @- B 1 m 1 m f B f f Df How good is this approximation? April 11, 2006 24 FLOW CONDUCTANCE ESTIMATION LEARNING BASED &= AA (u + d) K&h + K h&= AAx& Assume that the “sup” norm of K is bounded, and that K is continuous on the compact set A: K (x ) : A Ì ¡ K Then : A + ® BÌ ¡ + @ sup K (x ) < ¥ xÎ A 20 800 15 700 10 600 K&(x ) - W T F&(x ) < eK ,2 5 0 -5 500 400 300 -10 200 -15 100 -20 400 April 11, 2006 Flow Conductance Kv [LPH/sqrtMPa] K (x ) - W T F (x ) < eK ,1 Flow Conductance Error Actual-NLPN [LPH/sqrtMPa] Actual NLPN 450 500 550 600 650 700 Input Solenoid Current [mA] 750 800 850 900 0 400 450 500 550 600 650 700 Input Solenoid Current [mA] 750 800 25 850 900 FLOW CONDUCTANCE ESTIMATION LEARNING BASED Actual system &= AA (u + d) K&h + K h&= AAx& Let the observer be ˆ T F&(x )h + W ˆ T F (x )h&= A uˆ W A Let the error be e @ A A (uˆ - u ) Then (W T April 11, 2006 ˆ T )(F&(x )h + F (x )h&) = e + A d + O (e , e ) - W A K ,1 K ,2 26 FLOW CONDUCTANCE ESTIMATION LEARNING BASED SIMULATIONS PS PS Eta PA PA eta uest Eta-CONVERTER acc u isol isol u [mm/s] KA Vcmd KA [LPH/sqrtMPa] Vcmd KA-OBSERVER KA HYDRAULIC MODEL April 11, 2006 27 FLOW CONDUCTANCE ESTIMATION LEARNING BASED SIMULATIONS plots (d = 0) 1000 1.8 Actual Estimated 1.6 800 600 1.2 Eta [sqrtMPa] Flow Conductance [LPH/sqrtMPa] 1.4 400 200 1 0.8 0.6 0.4 0 0.2 -200 0 10 20 30 40 Time [sec] 50 60 70 0 80 60 0 10 20 30 40 Time [sec] 50 60 70 80 0 10 20 30 40 Time [sec] 50 60 70 80 60 40 40 20 20 Commanded Speed [mm/s] Piston Acceleration [mm/s/s] Actual Estimated 0 -20 -40 -60 0 -20 -40 0 April 11, 2006 10 20 30 Time [sec] 40 50 -60 28 FLOW CONDUCTANCE ESTIMATION LEARNING BASED Experimental data (offline) 1.4 1200 1.2 1000 1 800 Flow Conductance [LPH/sqrtMPa] Eta [sqrtMPa] Actual Estimated 0.8 0.6 0.4 0.2 0 600 400 200 0 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 -200 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 Note: Signals low-pass filtered at 5Hz April 11, 2006 30 FLOW CONDUCTANCE ESTIMATION LEARNING BASED How small is d? &= AA (u + d) K&h + K h&= AAx& The error is (W T ˆ T )(F&(x )h + F (x )h&) = e + A d + O (e , e ) - W A K ,1 K ,2 e @ A A (uˆ - u ) d depends on how well we know the friction model April 11, 2006 31 FLOW CONDUCTANCE ESTIMATION LEARNING BASED Actual Data & f f = PAAA - PBAB - mx& April 11, 2006 32 FLOW CONDUCTANCE ESTIMATION LEARNING BASED Friction model* fˆf = l 1 exp (l 2x&) + l 3PA - l 4PB + l 5x& Error from Velocity Independent Model 1.5 1200 1000 800 600 400 200 0 -200 -400 -600 -800 Friction Force Error [N] Friction Force [N] FRICTION FORCE VELOCITY INDEPENDENT MODEL 1.0 0.5 0.0 -0.5 -1.0 -1.5 -60 -40 -20 0 20 Piston Velocity [mm/sec] ff * 40 60 -60 -40 -20 0 20 40 60 Piston Velocity [mm/sec] ffMOD Bonchis, A., Corke, P.I., and Rye, D.C., (1999), A pressure-based, velocity independent, friction model for asymmetric hydraulic cylinders, in Proc: IEEE International Conference on Robotics and Automation pp. 17461751 April 11, 2006 33 FLOW CONDUCTANCE ESTIMATION LEARNING BASED Friction model* fˆf = l 1 exp (l 2x&) + l 3PA - l 4PB + l 5x& Experimental Data 1500 Friction Force [N] 1000 500 0 -500 -1000 -1500 -2000 -2500 -150 -100 -50 0 50 100 150 Piston Velocity [mm/sec] * Bonchis, A., Corke, P.I., and Rye, D.C., (1999), A pressure-based, velocity independent, friction model for asymmetric hydraulic cylinders, in Proc: IEEE International Conference on Robotics and Automation pp. 17461751 April 11, 2006 34 PRESENTATION OUTLINE FLOW CONDUCTANCE ESTIMATION Reported work Approaches ONLINE FLOW CONDUCTANCE MAPPING LEARNING AND CONTROL Fixed inverse mapping Learning mapping response April 11, 2006 FUTURE WORK CONCLUSION 35 MAPPING LEARNING & CONTROL PS PUMP CONTROL Pump Single EHPV Feedback compensation (discrete PI controller) Feedforward compensation (lookup table) QB- QA+ KvB- M KvA+ PB PA KvP KvB+ KvAPR QB KvT QB+ QA- QA QL FL=0 PB Tank PA m x x VB0 AB AA VA0 Hydraulic Piston EHPV - Wheatstone Bridge used for motion control of hydraulic pistons EHPV for pump control April 11, 2006 36 MAPPING LEARNING & CONTROL PUMP CONTROL Single EHPV Feedback compensation Feedforward compensation PRESSURE CONTROL EHPV (SINGLE CARTRIDGE) Patrick Opdenbosch November 9, 2005 ver 1.0 Sampling: 10 msec PSD[MPa] 3 PS_DES FEEDFORWARD COMPENSATION sw2 DISCRETE PID uk SENSORS ek 1 isol isol_m isol Target Scope Id: 1 Target Scope Id: 2 FFWD 1 1 Pdes i_COIL 1 isolm EHPV CONTROL 0 Pdes sw1 PS CAN-AC2-PCI B1 CAN 1 / CAN 2 Standard / Extended Setup Target Scope Id: 3 1 PS 1 PR 1 Target Scope Id: 4 MEASUREMENTS PR err Pump pressure control scheme April 11, 2006 37 MAPPING LEARNING & CONTROL PUMP CONTROL 12 1500psi 1450psi 1300psi 1000psi 800psi 400psi 1400 1300 1200 Coil Current [mA] Single EHPV Feedback compensation Feedforward compensation 1500 1100 1000 900 SUPPLY PRESSURE [MPa] 10 800 8 6 700 4 600 2 500 1 2 3 4 0 1500 5 6 7 Desired Supply Pressure [MPa] 8 9 10 11 1300 MA NU AL 1000 SE T 800 PR E SS UR E 400 [P SI ] 500 600 700 800 900 1400 1500 1200 1300 1000 1100 Feedforward mapping COIL CURRENT [mA] Measured mapping Pump pressure control scheme April 11, 2006 38 MAPPING LEARNING & CONTROL PUMP CONTROL PID Response: Manual Setting at 1000psi & KI = 30 KP = 350 7 5 Pressure [MPa] Single EHPV Feedback compensation Feedforward compensation 6 4 PSdes PS PR atm 3 2 PID Response: Manual Setting at 1000psi & KI = 30 KP = 350 6 1 2.5 5 3 3.5 4 Time [sec] Closed loop step response Pressure [MPa] 4 PSdes PS PR atm 3 2 1 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Time [sec] Closed loop tracking response April 11, 2006 39 MAPPING LEARNING & CONTROL FIXED TABLE CONTROL Desired Actual 50 Velocity [mm/s] Pump control + INCOVA control No adaptation of inverse Kv mapping Same inverse Kv mapping for all valves 100 0 -50 -100 -150 400 200 0 2 4 6 Time [sec] 8 0 2 4 6 Time [sec] 8 10 6 7 8 9 10 7 8 9 10 6 800 0 5 Time [sec] 200 800 200 4 7 1000 400 3 400 1000 600 2 600 0 10 1 8 5 0 2 4 6 Time [sec] 8 10 Pressure [MPa] B+ Input Current [mA] 600 0 A- Input Current [mA] 800 B- Input Current [mA] A+ Input Current [mA] 800 0 4 PSET PS PA PB PR ATM 3 600 2 400 1 200 0 0 2 4 6 Time [sec] 8 10 0 0 1 2 3 4 5 Time [sec] 6 Fixed Set Pump Pressure April 11, 2006 40 MAPPING LEARNING & CONTROL 250 Desired Actual FIXED TABLE CONTROL 1000 1000 800 600 400 200 0 2 4 6 Time [sec] 8 800 0 0 2 4 6 Time [sec] 8 10 Velocity [mm/s] -50 -100 -150 -200 -250 0 2 3 4 5 Time [sec] 6 7 8 9 10 PSET PS PA PB PR ATM 7 0 2 4 6 Time [sec] 8 6 10 600 5 4 3 400 2 200 0 1 0 2 4 6 Time [sec] 8 10 0 April 11, 2006 1 9 200 800 200 0 400 1000 400 50 8 1000 600 100 600 0 10 150 Pressure [MPa] 800 0 A- Input Current [mA] B+ Input Current [mA] 1200 B- Input Current [mA] A+ Input Current [mA] Pump control + INCOVA control No adaptation of inverse Kv mapping Same inverse Kv mapping for all valves 200 0 1 2 3 4 5 Time [sec] 6 7 8 Pump Margin Control 9 10 41 MAPPING LEARNING & CONTROL FIXED TABLE CONTROL VELOCITY ERRORS Inaccuracy of inverse tables Physical limitations/constraints 50 45 40 Steady State Speed Error V - Vdes [mm/s] Pump control + INCOVA control No adaptation of inverse Kv mapping Same inverse Kv mapping for all valves 35 30 25 20 15 10 5 0 10 20 30 40 50 60 70 80 90 100 Commanded Retract Speed [mm/s] Velocity Errors with Pump Margin Control and Fixed Inverse Tables April 11, 2006 42 MAPPING LEARNING & CONTROL LEARNING APPLIED TO NONLINEAR SYSTEM K Î [0, K MAX ] K k 1 F K k , isol,k isol,k Î [0,1500mA ] CONTROL DESIGN Tracking Error: Error Dynamics: ek = K k - d Kk F F i di o e , e ek 1 sol, k k k sol,k K k d K , d i k isol,k d d k sol,k Kk , isol,k ek 1 d J k ek d Qk isol,k d isol,k April 11, 2006 43 MAPPING LEARNING & CONTROL LEARNING APPLIED TO NONLINEAR SYSTEM K Î [0, K MAX ] K k 1 F K k , isol,k isol,k Î [0,1500mA ] CONTROL DESIGN Error Dynamics: ek 1 d J k ek d Qk isol,k d isol,k Deadbeat Control Law: isol,k = disol,k - dQk- 1 dJ kek April 11, 2006 44 MAPPING LEARNING & CONTROL LEARNING APPLIED TO NONLINEAR SYSTEM K Î [0, K MAX ] K k 1 F K k , isol,k isol,k Î [0,1500mA ] CONTROL DESIGN Deadbeat Control Law: isol,k = disol,k - dQk- 1 dJ kek Proposed Control Law: isol,k = g (d K k ) isol,k isol,k isol,k Q J e 1 k 1 k 1 k April 11, 2006 %T F (d K ) D i% = W sol,k k k 45 MAPPING LEARNING & CONTROL Nominal inverse mapping uk Qk11J k 1ek uk uk isol,k Qk11 J k 1ek isol,k isol,k Inverse Mapping Correction NLPN dx uk xk PLANT k Adaptive Proportional Feedback April 11, 2006 Jacobian Controllability Estimation 46 MAPPING LEARNING & CONTROL MODELING: Single Valve Command 8000 Actual Desired 1 Fac Kv d Kv d 7000 isol [mA] ERROR [LPH/sqrtMPa] 6000 Flow Conductance [LPH/sqrtMPa] Kv m isol isol Kv 5000 Kv [LPH/sqrtMPa] 10 1 4000 Vm PS PS [MPa] Q positive 8 INVERSE VALVE CONTROL 3000 PA Q [LPH] PA [MPa] EHPV A+ 2000 1000 0 1200 -1000 0 5 10 15 20 25 30 NLPN Prop Nom 35 Time [sec] 1000 12 Solenoid Input Current [mA] 800 10 Relative Error [%] 8 6 600 400 200 0 4 -200 2 0 0 5 10 15 20 25 30 35 Time [sec] 0 5 10 15 20 Time [sec] April 11, 2006 25 30 35 E rel @ (K - d K ) dK 47 MAPPING LEARNING & CONTROL MODELING: Full system STANDARD METERING EXTEND & STANDARD METERING RETRACT FUNCTIONS (w/ Pump Pressure Control) EHPV LEARNING CONTROL SIMULATION Patrick Opdenbosch March 06, 2006 ver 1.0 PS [MPa] PMAX =280 8sec => Vmax =50 6sec => Vmax =50 4sec => Vmax =80 2sec => Vmax = 200 Command PMAX =200 8sec => Vmax =30 6sec => Vmax =50 4sec => Vmax =80 2sec => Vmax = 150 v el P [MPa] P PR [MPa] ATM 1 Fac x [mm] pos PS v _des PB M v el PR Ps CPS PS PS Kv B- PA PUMP MARGIN CONTROL Kv BKv A+ PUMP P PR Tank v _des Kv B+ iB- iA+ iA+ Kv APB PB [MPa] PA PR INCOVA PA [MPa] QB PS iA- iA- iB+ iB+ Vm QA PB PA PA PB iB- Kv AKv A+ v [mm/s] pos Kv B+ QS EHPV CONTROL Fload INCOVA CONTROL EHPV WHEATSTONE QR Q [LPH] xdot PA F PB Piston Friction April 11, 2006 Friction [kN] 48 MAPPING LEARNING & CONTROL MODELING: Full system 60 Actual Desired 40 Piston Speed [mm/s] 20 Supply, Piston, and Return Pressures 0 6 -20 PS PA PB PR atm 5 -40 0 10 20 30 Time [sec] 40 50 60 Actual and Commanded Speeds Pressure [MPa] 4 -60 3 2 1 0 April 11, 2006 0 10 20 30 Time [sec] 40 50 60 49 MAPPING LEARNING & CONTROL MODELING: Full system (Solenoid Currents) Solenoid B- Solenoid B+ 900 900 Nom NLPN Prop 700 700 600 600 500 400 300 500 400 300 200 200 100 100 0 0 -100 -100 0 10 20 30 Time [sec] 40 50 Nom NLPN Prop 800 Solenoid Input Current [mA] Solenoid Input Current [mA] 800 60 0 10 20 Solenoid A- 50 60 900 Nom NLPN Prop 800 700 700 600 600 500 400 300 500 400 300 200 200 100 100 0 0 0 10 April 11, 2006 20 30 Time [sec] 40 50 Nom NLPN Prop 800 Solenoid Input Current [mA] Solenoid Input Current [mA] 40 Solenoid A+ 900 -100 30 Time [sec] 60 -100 0 10 20 30 Time [sec] 40 50 60 50 MAPPING LEARNING & CONTROL EXPERIMENTAL: Learning applied to retract motion Valve motion Low Pressure High Pressure Piston motion April 11, 2006 51 MAPPING LEARNING & CONTROL EXPERIMENTAL: (30 mm/s commanded) ADAPTIVE Retract Control: Pump Margin Pressure Control ADAPTIVE Retract Control: Pump Margin Pressure Control 370 100 360 80 350 60 340 40 330 20 Velocity [mm/s] Position [mm] Desired Actual 320 310 0 -20 300 -40 290 -60 280 -80 270 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 -100 5 0 0.5 1 50 9 40 8 30 7 20 6 5 PSET PS PA PB PR ATM 4 3 April 11, 2006 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 4 4.5 5 -20 -40 1 3.5 -10 1 0.5 3 0 -30 0 2.5 Time [sec] 10 2 0 2 ADAPTIVE Retract Control: Pump Margin Pressure Control 10 Velocity Error: Vdes - V [mm/s] Pressure [MPa] ADAPTIVE Retract Control: Pump Margin Pressure Control 1.5 4 4.5 5 -50 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 52 MAPPING LEARNING & CONTROL EXPERIMENTAL: 800 B- Nominal Input Current [mA] A- Nominal Input Current [mA] 800 600 400 200 0 0 1 2 3 Time [sec] 4 200 0 1 2 3 Time [sec] 4 5 0 1 2 3 Time [sec] 4 5 200 B- NLPN Input Current [mA] A- NLPN Input Current [mA] April 11, 2006 400 0 5 150 100 50 0 -50 600 0 1 2 3 Time [sec] 4 5 150 100 50 0 -50 53 MAPPING LEARNING & CONTROL EXPERIMENTAL: Learning applied to all four (4) EHPVs Valve motion Low Pressure High Pressure Piston motion April 11, 2006 54 MAPPING LEARNING & CONTROL 100 Desired Actual ADAPTIVE TABLE CONTROL 60 40 Velocity [mm/s] Pump margin control + INCOVA control NLPN approximation of inverse Kv mapping using 4 NLPN 80 20 0 -20 -40 -60 -80 220 -100 200 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 Velocity Performance 180 20 15 140 10 120 Velocity Error: Vdes - V [mm/s] Position [mm] 160 100 80 60 40 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 0 -5 5 Piston Displacement: Retraction -10 -15 -20 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 Velocity Errors April 11, 2006 55 MAPPING LEARNING & CONTROL 160 ADAPTIVE TABLE CONTROL Desired Actual 120 100 Velocity [mm/s] Pump margin control + INCOVA control NLPN approximation of inverse Kv mapping using 4 NLPN 140 80 60 40 20 300 0 0 250 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 4.5 5 Velocity Performance 200 140 120 150 100 100 50 0 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 4.5 5 Piston Displacement: Extension Velocity Error: Vdes - V [mm/s] Position [mm] 0.5 80 60 40 20 0 -20 0 0.5 1 1.5 2 2.5 Time [sec] 3 3.5 4 Velocity Errors April 11, 2006 56 PRESENTATION OUTLINE FLOW CONDUCTANCE ESTIMATION Reported work Approaches ONLINE FLOW CONDUCTANCE MAPPING LEARNING AND CONTROL Fixed inverse mapping Learning mapping response April 11, 2006 FUTURE WORK CONCLUSION 57 FUTURE WORK Investigate online application of observer Complete velocity error comparison between system’s response under fixed inverse tables and adaptive inverse tables Study convergence properties of adaptive proportional input and its impact on overall stability Improve learning applied to 4 EHPVs by NLPN + adaptive proportional feedback Incorporate fault Diagnostics capabilities along with mapping learning April 11, 2006 58 PRESENTATION OUTLINE FLOW CONDUCTANCE ESTIMATION Reported work Approaches ONLINE FLOW CONDUCTANCE MAPPING LEARNING AND CONTROL Fixed inverse mapping Learning mapping response April 11, 2006 FUTURE WORK CONCLUSION 59 CONCLUSIONS Discussed several approaches to the flow conductance estimation problem Presented a learning method for estimating flow conductance Presented performance of the INCOVA control system under constant and margin pump control for fixed inverse valve opening mapping Presented Simulations and experimental results on applying learning control to the Wheatstone Bridge EHPV arrangement April 11, 2006 60