DESIGN AND REALIZAION OF PROGRAMMABLE EMULAOR OF MECHANICAL LOADS Milan Žalman, Radovan Macko he Slovak University of echnology in Bratislava, Slovakia Abstract: his paper concerns design and realization of programmable ulator of mechanical loads for motors. he aim of this paper is to build integrated test equipment, which will allow realistic testing of drive parameters and mainly quality of control structures on controlling different kinds of loads. For this purpose create userfriendly software which will allow ulation of physical syst in realtime using xpc arget under Matlab 6.. Copyright 5 IFAC eywords: Load Emulation, Mechanical Loads, Drive Control. INRODUCION Electric drives have a wide range of application in all industries. hey are used to drive, translation, stroke, positioning etc. hese processes often contain linear, nonlinear or time varying inertial loads. herefore a dand for examination of motor characteristics and testing of motors has come up. For testing, research and development purposes it is not possible to have all kinds of different mechanical loads in laboratory due to financial and space reasons. Convenient solution is to ulate mechanical load electrically, using electromotor. his allows experimental testing of nonlinear, adaptive, robust and intelligent control algorithms and testing of motor properties. Aim of this paper is to design and physically realize load ulator, so the user can easily choose type of load and set up desired parameters and verify his own design of control algorithm.. BASIC LOAD EMULAION MEHODS In load ulation, the aim is to produce an OLF which may be given by any relation between input M m and output ω. It is perhaps best to start with the simplest load ulation which can be written by ω ( s G ( s ( M m ( s s B and B is the ulated inertia and friction and it is going to be defined by user. Now we need to produce load machine control structure so that the relation between speed ω(s and the electrical driving motor torque M m (s is given by equation (.. Inverse Model Method Principle of this method will be explained using Figure (. ω (s M m (s s G ( s (s M l G ( s Fig. Inverse model In order to ulate mechanical load G the easiest way is to use Inverse model. G (s represents the mechanical load to be ulated, G (s its inverse model, s represents the syst of motor and load machine and G (s its inverse model. For example: s ( s B
G ( s s B ( is moment of inertia (kgm and B is viscousfriction coefficient (Nms. According to Figure ( we can say that equation ( is valid. M m(s M z(s s B ω(s ω( ( z B ( z B M z( zoh G h(s Fig. Sampled inverse model syst G ( In practice the load machine controller will be realized in a discrete form, which means we have to consider sampling of signals. herefore we transform the continuous syst into discrete syst as you can see in Fig. using sample period, G h (s is the zero order hold. If this control is implented in discrete form, output of inverse model can be very noisy due to derivative terms in inverse function, and it is necessary to resolve stability of the syst. From Figure ( we can write M m ω ( (4 Gh G ( hen we can derive G ( as M z z G ( ( ( (5 ω( erm G h (ss after discretization becomes ( z (/ B D G h ( Z (6 z s ( / B s B( z D D e,, B B B hen OLF (G h (G ( is ( ( D( z α OLF (7 z( z D α (8 he closed loop characteristic equation can be written as ( ( D α z z D ( ( D 4444 4444 4 444444 β β (9 In general B will be close to zero. herefore we use simplifications: D e D ( After substitution into characteristic equation β and β becomes β and β ( ake note that β and β are independent off the viscous friction value B, which means that stability is not affected by value of B. If we consider that B is small or zero, the characteristic equation becomes ( z z ( his indicates that the syst is unstable when / > or >. If we consider that and B is not zero or small the characteristic equation becomes B z z ( In this case the syst is unstable if B / >. he syst can be stabilized using discrete filter of the form (4 to lower the noise of M z and by doing this increase the use of this method. z ( γ (4 z γ Using this filter we stabilized the syst and increased the area of use significantly, however, the zero pole structure of desired mechanical load is violated which leads to unsatisfactory results in a closed loop control syst. For this reason is the Inverse model method which is used in previous researches (Collins and Huang, 994; Sandholdt, et.al., 996; Betz, et.al., 994; Newton, et.al., 995 not in place for our ulator since we want to preserve the real mechanical dynamics during ulation. M m ( G ( ω ( G comp ( G h (s M z (s G r ( G h (s s ω(s Fig. Block syst of open loop method
. Open loop ulation Method Consider the syst in Figure (. Based on Motor orque M m and required load to be ulated, ideal speed ω is computed. his is the speed at which would the real motor rotate with the desired load. his ideal speed is compared with real speed ω and modified by G comp. Based on the difference the controller G r ( derives load torque M z for the load machine. he total of these two torques from both machines (load and motor, forces the syst to travel at the desired ideal speed ω. he reason why this method is called open loop is that, the transfer function ω/m m is controlled in open loop (Akpolat et.al., 998. he function of G comp is to compensate all parts of the open loop except the ulated load function, so that the open loop discrete transfer function, ω(/m m ( becomes equal to discretized equivalent of G (s.he transfer function G (s includes mechanical parameters of both motors. herefore the transfer function of G comp is Gr ( Gcomp ( (5 Gr ( Z{G h (ss} and G r ( is discrete PI controller. Controller G r ( is given by r ( z Ar G ( (6 r z In order to complete the control structure of Open loop method we need to define G comp. az az a Gcomp( z (7 b z b z Where the constants are given as b, b A r, b ( D D a a a Ar he delay which we added to G comp is compensated by z, which we multiplied the load transfer function with. If the ulated load is to be that of inertia and friction B, the discretized load transfer function G becomes ( D z * Z z s s B B( z D (8 D e Compensative term G comp cancelled all parts of open loop syst except of ulated load transfer function so, that the final reduced syst matches syst in Figure (4. ω d( ω( G c ( G h (s Sample and hold M m(s Fig. 4 Closed loop speed control syst ω(s s B G (s Now we can apply speed or position control methods to verify the quality of ulation desired loads. his method preserves the dynamics of chosen load and is suitable for realization of our ulator. Notice that block G comp is independent of G. his is very handy since we can calculate it once and then we can choose different kind of loads without the need to calculate it again. Programmable load ulator in general is equipment which consists off controlled load machine and is tightly mechanically coupled with the motor (tested motor as shown on Figure (5. Load machine with special control (substitutes ulates technological load. Desired load type is obtained using a suitable control of load machine. control ested motor ω, M m M z load motor Fig. 5 General sche of load ulator. EXPERIMENAL AND SIMULAION RESULS Physically is ulator realized as a syst of two tightly coupled 4W DC motors. Both motors are fed by transistor converter. Figure (6 shows Basic simulation/experimental syst used to compare simulation and experimental results of load ulator using described method with different kind of loads. Fig. 6 Basic simulation/experimental syst his syst is universal. o change the type of ulated load you simply change G in block gen Mz as shown on Figure (7. Fig. 7 load torque generator structure (linear load We created a simulation model of the motor with defined load so that we have some data to compare the experimental results with. First we checked the functionality of the ulator by doing an experiment
in which we applied linear load to the motor, G was times the value of syst inertia and B was equal to B. We also applied static friction (M z.5nm in the model as in the ulator because it causes probls to motion control systs. he result of this experiment is presented on Figure (8 and Figure (9. with modified block gen M z from Figure (. 9 8 Exp. speed Fig. Gen Mz structure for ulation of mass syst with flexible coupling ω [rad/s] 7 6 5 4 5 4 99 Simulation 98.6.8..4.5.5.5 Fig. 8 Comparison of speed responses for, B B and M z.5 Nm From these two figures you can see that ulation works very good and the simulation responses are very close to experimental responses of speed and torque. [A] Im 5 4 Exp. torque Simulation.5 4.5 5.8..4.5.5.5 Fig. 9 Comparison of torque responses for, B B and M z.5 Nm As you can see the only difference between simulation end experimental results are notable at the end when the motor is stopping and speed is zero. his difference is caused only in simulation by static friction because the exact model was not used. However this is only simulation probl when the angular speed is close to zero and the experimental results are satisfactory and correct. One of the most complicated types of load is mass syst connected with flexible coupling, which is possible to find in various applications in praxis. In order to check, if the ulator is able to act also as mass syst, we used block diagram from Figure (6 [rad/s] ω 9 8 7 6 5 4 95 9 85.5.5.5.5.5.5 4 4 ω real ω real ω simulated ω simulated Fig. Comparison of speed responses for, C t. Nm.s/rad and zero dumping coefficient In next experiment we ulated twoinertia syst without dumping (most problatic case, the torsional spring constant C t.nm.s/rad and inertia ratio R /. We applied step of external load (5% of the motor rated torque disturbance at t.5s. For speed control of this syst we used controller designed with the method of pole placent by identical dumping coefficients (Zhang and Furusho,. he coefficients used were ξ ξ.865. Figure ( and Figure ( shows the comparison of experimental and simulation results of ulation two inertia load. Im [A] 4 smulation real current.5.5.5.5 4 4.5 Fig. Comparison of torque responses for, C t. Nm.s/rad and zero dumping coefficient
As you can see the simulation results are very close to results of experimental syst. Load speed (ω control was indirect, in other words in closed loop was controlled only angular speed of (ω motor. Even under these conditions ulation gives satisfactory results. In general load ulator can be realized with any pair of motors. What we need to know is mechanical parameters of such motor couple ( and B, information about actual angular speed and information of motor torque M m. his ulator as presented can be used to ulate different kind of linear or nonlinear loads. We can reduce the area of use to schatic diagram on Figure (. Motor fixed flexible Gearing Load mechanism Mz f (ω, ϕ,, t, par. ω f (t, par. Fig. Loads able to be ulated. HW AND SW REALIZAION Emulator was realized with DC motors. he control strategy was ployed using Matlab real time and xpc target libraries. he model was created on the PC with windows and Matlab and then compiled using C compiler into executable file. hese files were executed on second PC with xpc operating syst. ransfer between the computers is automatic via Ethernet adapters. his method allows us to ulate complicated loads with high computing power needs. he process of creating, compiling, transferring, running and getting measured data is quite complicated. herefore in order to ease the way this ulator is used in praxis, user friendly software was created under Matlab. You can see its main window on Figure (4. User interface is very convenient when testing a drive with same kind of loads only different parameter setup. here are 4 predefined load types: linear centrifugal fan winding machine 4 mass syst with flexible 4. CONCLUSIONS In this paper programmable load ulator for dynamic ulation of linear and nonlinear mechanical loads has been presented. Dynamics of desired load is preserved during ulation. Realized tool will help future student generations to design and test robust, intelligent, adaptive or fuzzy control algorithms for controlling nonlinear, twoinertia or time varying loads. Emulation and intelligent control of two inertia systs with flexible coupling and static friction will be the subject of future research. REFERENCES Collins, E.R. and Y. Huang (994. A programmable Dynamometer for esting Rotating Machinery Using hreephase Induction Machine. IEEE ransactions on Energy Conversion 9,, 5 57. Sandholdt, P., E. Ritchie,.. Pederson and R.E. Betz (996. A Dynamometer Performing Dynamical Emulation of Loads with NonLinear Friction. IEEE Int. Symposium on Industrial Electronics,, 87878. Betz, R.E., H.B. Penfold and R.W. Newton (994. Local Vector Control of an AC Drive Syst Load Simulator. IEEE Conference on Control ApplicationsProceedings,, 776. Newton, R.W., R.E. Betz and H.B. Penfold (995. Emulating Dynamical Load Characteristics Using a Dynamic Dynamometer. Proceedings of Int. Conference on Power Electronics and Drive Systs,, 46547. Akpolat, Z.H., G.M. Asher and.c. Clare (998. Emulation of high bandwidth mechanical loads using vector controlled AC dynamometer. Proceedings 8 th Int. Power Electronics and Motion Control Conference, 5, 8. Akpolat, Z.H., G.M. Asher and.c. Clare (999. Experimental dynamometer ulation of nonlinear mechanical loads. IEEE, ransactions on industry applications, vol.5, No.6. Zhang, G. and. Furusho, (. Speed Control of woinertia Syst by PI/PID Control. IEEE, ransactions on industrial electronics, vol. 47, No.. Fig. 4 Main window of software interface.