The Colteye of Engineering University of Wisconsin-Madison Wisconsin Electric Machines and Power Electronics Consortium RESEARCH REPORT 8-8 Synthesis of State Variable Controllers for Industrial Serva Drives Robert D. Lorenz University of Wisconsin 1415 yonnson orive Madison, Wisconsin 53706 WEMPEC Department of Electrical and Computer Engineering 1415 Johnson Drive Madison, Wisconsin 53706 May 1986of State Variable Controllers {for Industrial Servo Drives Robert D. Lorenz University of Wisconsia-Madison \widely used in industrial servo drives. Furhermora, avan has had formal ‘exposure to modern control theory, classical control techniques, rather than stato variable approaches, are ‘offen salacted for industal automation applicatione, ‘There ‘are many possible reasons for this lack of ulization, However, two major issues are @ lack of intuton in the ‘current state variable techniques and a lack of a good field procedure This paper presents a new approach which provides: 4 strong intuitive link and also an easily implemented field Procedure. The approach is demonstrated on @ motion ‘controt servo crve application, Trial results are presented Which indicate substantial promise for the approach when, ‘compared to classical techniques. ‘Backoround + Existing Approach to State Variable Modal Analysis ‘The general state vaiiable apprvact tor modeling and analysis of systam dynamics is based on a matrix Fepresentation of the system differential equations.{1] The ‘goneral form ofthis equation is described by the following Totavonsnips Xe FX + Gu 0 Y= cx @ and where: X & X aro tho state vector and ts derivative F isthe state transition matrix Gis the system input matrix U__ isthe system input vector Y isthe system output vector © is the system output defining matrix I form of this state variable Figura 1 shows the ger ‘modal in block diagram form. u +k x v =le thot b Figure 1 Block Diagram of State Variable Mode! To add a state variable controller to this model one system dynamic model with a controller modal UAKX @ where: K eth stato gain vootor ‘This implios that the system states are measured and fed back with gain applied to each term. The analysis of the systurn response then follows from a matrix analysis of the resulting equation. The final resut is of the matnomatically {familiar Eigenvalue form. Using this conventional state variable approach then implies synthesis of the controller via selection of gains to achieve desired closed loop Eigenvalues. + Limitations of tha Existing Apprmach tn Stata Variahia Controller Synthesis ‘Tha existing approach to state variable controler synthesis, albeit analytically correct, has some signifcant mations tor practicing sorrel engineers, One ofthe most important limitations ve the lack of intuition, at least as compared to the classical control synthesis techniques of root locus or Bode and Nyquist frequency response plot Techniques. (ne matnx notation, azinouga simpinying the representation, Is also a further abstraction trom the individual systom equations which have physical meaning Thus. ft, 100. dearades some of the intuition of ohysical ‘magnitudes. ‘Another important limitation of the existing approach to state variable controller synthesis is the lack of a fraotablo Weld tuning approach. Given fold data, tho Eigenvalue solver Is not generally availabe to the field engineer. Furthermore, simultaneous installation of new gain torms via a naw gain vector makas it more citicut to IGolate and debug problems wnicn may occur ‘Development of a Naw State Variable Controller ‘Synthesis Aporoach The general goals of a new approach to state ‘variable controller synthesis should include: Improved intuition for tha dasignar ‘9280 of use comparable to classical techniques Insensiivty to parameter values ‘applicability to field tuning needs. ‘A general approach which achieves the above goals has been developed and Implemented in teaching and laboratory environments. In order to present this approach 8 simple de motor current loop example will be carried out as part of the development. +The first stop in achieving greater intuition isto return to the Individual state equations. For tha de motor ‘example the governing state equations are:ie. Ri-Kor 6 (4) go Kri-b @ 6) 122 at In matrix form thoee etate equations appear a¢ follows: oi Ro Kel | fo lar] .[t oT iG da} Kr .2 * a om a || 2 (6) ‘These equations may also be represented in a block | FOP as fe Rie ble i my k "State Variable Feedback" ke Figure 4 System Block Diaaram with ‘State Variable Controlor In matrix equation form the closed loop system state fai | [A Keks Ko Kate ka jax} Po EY bey, 5 + da Kr a at a v Hey? (7 + By comparison of the state variable block diagram form with the closed loop state equation matrices, itis apparent that a parallel structure exists between the physical state one forme a dimancional analyeie of the de motor example, than the units of the Ky state variable controlier gain are seen to be ohms, and thus, the gain KyK is seen to be a form of effective rasistance addad by the state variable controller, Similarly, the units of the Ka state variable controlar gain aro seen to be the same as the back emt Constant, Ke, which the controler state feedback parallels. This parallel structure leads to a direct and intuitive Interpretation of the controlar state feedback as follows. ‘To tune the current loop of the de motor example, one needs to estimate how low the effective electrical time constant (= L/R ) should be for the application. From this estimate and the armature inductance “L*. it is then ible to directly calculate the required resisiance wrich state variable controler “resistance” gain Ky should be sotto.{t should be noted that offen the state controller “resistance” gain (or other gains also) will be much larger than the physical state feedback ‘resistance’ (or oth physical state feedback gain). In such cases. the physical state gain parameter may be significanty in error or may ‘vary without substantially affecting the closed loop control arlormanca ‘The back emf state variable gain Kp can be similarly set by observing that the back emf voltage acts as a isturbance to the current regulating 1o0p. I the state variable controller back emf" gain Ka is set such that it ‘cancels the effect of the physical state feedback gain Ke, ve. ran fe ® then the current regulator will be dynamically and statically decoupled from the back emf disturbance. In effect, the State variable controller is generating a voltage command ‘which would instantanaously procuce ine required pack ‘emt terminal voltage without requiring any error in the current state variable control loop. This case demonstrat ‘cancallation which may be verv directly imolomented via a ‘State variable controller. + Tho dynamic *correctnese" of the state vatiahla evirant Control foop stands in marked contrast to the dynamic ‘errors inherent in the classical proportion/ntegral (PI) Controller which is almost universaly used in industriat Serve uilves (Uull Ue aid ac dives). Due to the (gration, and thus, lfiite de gain ofthe Pl controker, the ‘current loop is not atfected by de currents. However, for dynamic content of de drives and for ac currents as required in ac drives (brushless dc and ac synenronous ‘and induction motor drives) state feedback (or state feed forward) is required to achieve dynamic correctness of the ‘currant regulation (2) ‘Summan_of the Developed. General Anoioach for ‘Synthesis of State Variable Controllers ‘Anew, general state variable development approach thas been demonstrated to have the following principal elements: writing out the sta differential equations ‘drawing the state variable form of block diagram Identitying the physical system state feedback gain terms setting up a parallel structure of controller stat feedback gain terms having the same physical dimensions ae the physical system ciate feedback gain torms applying block diagram or matrix state equations ‘analysis to establish the desired physical system ‘equivalent closed loop state feedback gain terms. variable form of the system ‘A second application of this method will provide yet ‘another means of demonstrating the above principles. ‘Anplcation of this State Variable Approach to Velocity and Position Loags for Motion Control Servo Drives Given modem command feed forward algorithms, a primary function of the closed loop controller in motion ‘control applications is to provide disturbance rejection. ‘The davelopad stata variable synthasis tachniqua may ba ily employed to implement the desired disturbance fejection attributes, The following example will ‘demonstrate such en application: ‘Givon a de motor (or fold orianted ac synchronous or {induction motor) with @ position sensor and a pwm power amplifier tuned as proviously discussed with a state variable current regulator. For such an application the cive may be viewed as a torque controlled source. { 3} With this assumption, a simplified state variable dynamic ‘model of the motor position and velocity would b 4 ew: ® = 70 w+ Rrer (10) In ganaral a motion control system is not connected to its reference frame with a spring. However, In order to best demonstrate the physical meaning of the state feedback terms, itis possible to tamporanly add an imaginary spring, Keg The rosuing etato variable dynamia madal nt tha ‘motor would then be: 8 a 5 ° a) BO Kye = b O+ Kro. en 3 Ter (12) In matrix form these state equations appear as follows: Ea t]}o} fo a| .|° jaa] [Ke *)g|°* at T el (13) In state vanabie block aiagram form the system would bu ‘as shown in Figure 5, Figure 6 Ctate Variable Block Diagram of MotorFrom the state variable form of the system block diagram it is apparent that physical system state feedback gains exist {or both camping and stitfness. Thus, in the closed loop ‘system state variable model one should expect that terme which directly produce damping and stiffness. should ‘@ppear, Such is the case as shown in Figure 6. Figure 6 Single Axis System Block Diagram with ‘State Variable Controlier In matrix equation fun thy vlusud foop system state variable model becomes: 00 Se ° 1 He] Jo da] KJ? at 7 ) Doth te state variapie block alagram model and the ‘mattix form demonstrate the role of state feedback in directly producing the effect of both damping and sttiness. However, unlike the current loop example, both of the Present example's physical state feedback gains a eneraly very small or zero (sttiness is defined to be zero it the spring is removed). In that case the controller is roviding all the relevant damping and all the stifnoss ‘Thus, given a disturbance rejection erterion, is possible to directly and independently manipulate stitiness and damping. It should be noted that the state variable feedback gain stitiness and damping terms also are defining the bandwidth and damping characteristics of the ‘command path. Thus, even for command driven systems, the dynamic properties of the state variable closed loop sysiem are easily detined using this state variable synthesis technique + Itis significant to note that with classical techniques tha various loop gains do not individually define either static or dynamic siiiness. Thus, the state variable approach presented provides @ unique level of design intuition eyond that readily available using classical techniques. ‘Development of a Field Tuning Aparoach ‘The field tuning procedura daveloped for thie etato variable synthesis approach is consistent with that used for classical approaches, but is inherently easier to implement. ‘The general procedure Is: A” Setup and pertorm all cancollation terms ist, such as the back emt cancellation, B. Tune each state loop individually going trom the innermost stato variable loop to the outermost state loop, ie. first current, than velocity. then displacement. This procedure is inherently easier to implement than classical techniques for at least three major reasons. 1. Disturbance responses are basically being nulled (zeroed) with the cancellation terme Nuling ‘operations are gen 9 fo measure than are ‘xact performance values. 2. Each loop being tuned only has one costficient to adjuet. In claeeical PI or PID approaches twe and three costficients must be modified simultaneously In defined relationships, 3. Precise knowledge of physical state gain values are iy Wot Hee. For cancellation terms, the null detection ‘assures that the correct state feedback gain has ‘been selected, independent of the any estimated or specification sheet values of the physica’ state feedback gain terms. For physical state feedback gain terms which ara baing aligmantad hy the stato cantraller gaine, the fact that the controller state feedback gains will ally be larger than the physical state foodback {Gains makos those estimates less critical, ‘Teal Evaluation of this Stata Variable Synthesis Anoroach In order to demonstrate the viabilty of this approach {or tuning of motion contro aves a laboratory cnva system With all necassary specification data was provided to two groups of two graduate students who were asked to independently tune the drive. The students were to use both the state variable approach and a classical Pl type approach. The laboratory drive consisted of a de motor (parameters are in the appendix) with attached dc tachometer and angular ancodar, an amplifier with current {feedback and an M6809 control computer complete with all ‘necessary interfaces. The students wore asked to tune the ‘current loop for a 1kHz bandwidth, and to achieve 10 Hz postion loop with critical damping. + Experimental Results of the Trial ‘The results of the current loop tuning were very dramatic. None of the students correctly tuned the PI current loop on their first try. All of the students tuned the slate variable controller correctly to obtain the 1000 Hz bandwidth on their fist try. The numerical results indicates that the precalcuiated values based on specification sheet values were in error by more than 25%, ‘The results of the position and velocity loop tuning wore almost a2 dramatic. The students sorectiy completed! the state variable tuning in lass than one 2-hour session,including time to key in their control software, The Pl Controller took more than two 2-hour sessions to complete. Such trial results are hardly sufficient to reach firm ‘conclusions in the use of this state variable synthesis approach, but the approach does appear to show promise. ‘Summary and Conclusions ‘A now stato variable synthesis procedure has been Presented and demonstrated on a motion controt servo rive. This appreach poceecsos the following atributos: + It provides physically meaningful insight into the design of state variable controllers, by moving eyond a purely mathematical formulation of state variables + itis teoretically sound and vet very intuitive + itlonds itself toa straightforward field tuning Procedure + itis inherently less sensitive to physical parameter ‘errors than are classical techniques + 1s capable oF ana was nett wo Bynamicaly worrUCt fecoupling of interacting toms + it is not an inherently more costly solution than classical tehcniques. + the results of ongoing use and proliminary trials show pmmica af greater canimilar synthacie afiiancy and ‘accuracy. ‘References 1.) Brogan. W. L._. "Modem Control Theary*, Quantum Publishers, 1974, New York, New York 2) Novotny, D. W. , Lorenz, R. D., “An Introduction to Fiold Orientation and High Performance AC Drives", IEEE Tutorial Publication for 1985 IEEE IAS Annual Meeting, Toronto, Ont, Oct. 6-7, 1985. 3) LoenrkeJ. M., Lorenz, R.D., Novoiny, 0. W. , “Torque Characteristics of Field Oriented Induction Machines", in Cont. Rec. 1985 CAMC, pp. 101-107. ‘Anpencs ‘The nameplate data and parameters of the machine used {or the experimental work are: Mow Lee oF R= 2.05 Ohm Kg = 0.084 Viradsec i= 442 x10 A Mech J 25x 10°4 Kg-m? v= 0.09x 10° rewrauisee Amplifier Kg=54VN