Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 8, No. 4, August 2018, pp. 2180 2198 ISSN: 2088-8708 2180       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     Contr olling a DC Motor thr ough L ypauno v-lik e Functions and SAB T echnique Alejandr o. Rinc ´ on 1 , F abiola. Angulo 2 , and Fr edy . Hoy os 3 1 F acultad de Ingenier ´ ıa y Arquitectura, Uni v ersidad Cat ´ olica de Manizales, Grupo de In v estig aci ´ on en Desarrollos T ecnol ´ ogicos y Ambientales-GIDT A, Email: arincons@ucm.edu.co 2 F acultad de Ingeniera y Arquitectura - Departamento de Ingenier ´ ıa El ´ ectrica, Electr ´ onica y Computaci ´ on - Percepci ´ on y Control Inteligente - Bloque Q, Campus La Nubia, Manizales, 170003 - Colombia,Email: f angulog@unal.edu.co 3 F acultad de Ciencias, Escuela de F ´ ısica, Grupo de In v estig aci ´ on en T ecnolog ´ ıas Aplicadas - GIT A, Uni v ersidad Nacional de Colombia, Sede Medell ´ ın, Email: feho yosv e@unal.edu.co Article Inf o Article history: Recei v ed: October 10, 2017 Re vised: March 14, 2018 Accepted: May 2, 2018 K eyw ord: Rob ust Adapti v e Control L yapuno v-lik e Function Permanent Magnet DC Motor Rapid control prototyping ABSTRA CT In this paper , state adapti v e backstepping and L yapuno v-lik e function me thods are used to design a rob ust adapti v e controller for a DC motor . The output to be controlled is the motor speed. It i s assumed that the load torque and inertia moment e xhibit unkno wn b ut bounded time-v arying beha vior , and that the measurement of the motor speed and motor current are corrupted by noise. The controller is implemented in a Rapid Control Proto- typing system based on Digital Signal Processing for dSP A CE platform and e xperimental results agree with theory . Copyright c 2018 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Alejandro Rinc ´ on Santamar ´ ıa Uni v ersidad Cat ´ olica de Manizales Cra. 23 No. 60-63, Manizales, Colombia T elephone: (57-6) 8933050 Email: arincons@ucm.edu.co 1. INTR ODUCTION DC motors are commonly used in industry applications [1], [2]. A significant dif ficulty for DC-motor control design is the unkno wn time-v arying nature of its parameters [3], [4]. An important frame w ork for motor control design is the state adapti v e backstepping (SAB) technique presented in [5], as can be noticed from [6], [7]. Nussbaum g ain techniques are usually incorporated in the SAB control frame w ork in order to handle the ef fect of unkno wn time-v arying parameters and to impro v e the rob ustness of the system. Rob ust SAB control schemes incorporate a compensation term in the control la w , and some modification in the update la w , for instance the modification, see [8]. Ne v ertheless, upper or lo wer bounds of the plant model coef ficients ha v e to be kno wn to guarantee asymptotic con v er gence of the tracking error to a residual set of user -defined size. In [9] a non-adapti v e state backstepping control scheme is de v eloped. The resulting time deri v ati v e of the L yapuno v function in v olv es an unkno wn, time v arying b ut bounded term, whose upper bound is unkno wn, such that the backste pp i ng states remain bounded and con v er ge to residual set whose size depends on plant parameters. Nussbaum SAB control schemes are based on the controllers presented in [10], [11], as can be noticed from [12], [13], [14]. In turn, the controllers in [10], [11] are based on the Uni v ersal Stabilizer that w as originally introduced in [15] and discussed in [16]. The main dra wback of the Nussbaum g ain technique is that the result- ing upper bound of the transient beha vior of the tracking error depends on inte gral terms that in v olv e Nussbaum functions [12], [17], [14], [10]. Some controllers that use this technique usually present som e of the follo wing dra wbacks: i) some upper or lo wer bounds of plant model parameters ha v e to be kno wn in order to guarantee the con v er gence of the tracking error to a residual set of user -defined size [14], [10], b ut the control designs in [12], [13] indicate that a proper design w ould relax this dra wback, and ii) the control or update la ws in v olv e signum type J ournal Homepage: http://iaescor e .com/journals/inde x.php/IJECE       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     DOI:  10.11591/ijece.v8i4.pp2180-2198 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2181 signals, as can be noticed from [17]. In [18], a linear induction motor is considered. The friction force and unkno wn time-v arying model parameters lead to a lumped bounded uncertain term whose upper bound is unkno wn. The goal is to control the mo v er position. The dra wback is that the identification error is assumed constant in the definition of the L yapuno v function. In [6], a synchronous motor dri v en through A C/DC rectifiers and DC/A C in v erters is considered. It is assumed that the motor parameters e xperience unkno wn time-v arying b ut bounded beha vior . The goal is to control the motor speed, the rectifier output v ol tage and the d component of the stator current; the tracking error con v er ges to a residual set whose size depends on unkno wn motor parameters and user -defined controller parameters. There- fore, if upper bounds of unkno wn motor parameters are kno wn and controller parameters are properly chosen, the size of the residual set can be user -defined also. Other w orks address the problem of designing a controller for dif ferent motors [19, 20, 21]; despite the f act that the controlled systems operate as it is e xpected, the main disadv antages of these w orks are: the size of the output error cannot be determined, and no analyses of the system beha vior inlcuding measurement noise are presented. In [22] tw o coupled controllers are designed: a Linear Quadratic Gaussian (LQG) and a MRAS-based Learning Feed-F orw ard Controller (LFFC); e v enthough the simulation results demonstrate the potential benefits of the proposed controlled, the LQG algorithm may f ail to ensure closed-loop stabil ity when v ariations in the uncer - tainties are lar ge enough. In [23] a plant model in cont rollable form with unkno wn v arying b ut bounded parameters is considered, being the upper bounds of such parameters unkno wn. A SAB control scheme is de v eloped, and o v er - comes the main dra wback of the Nussbaum g ain method, as a result, the transient error is upper bounded by an unkno wn constant that does not depend on inte gral terms. Ne v ertheless, the control scheme is only v alid for plant models in “companion form”. The method is based on the L yapuno v-lik e function technique appearing in [24], [25]. In addition to the undesired unkno wn time-v arying nature of plant parameters, other important issue is the measurement noise. T racking performance can be de graded, e v en if the controller is rob ust ag ainst modeling uncertainty and disturbances (cf. [26], [27]). Some of the main techniques to tackle the ef fect of measurement noise are: high g ain observ ers, interv al observ ers, filter theory and the technique de v eloped in [9]. High g ain observ ers are useful to estimate system states and output deri v ati v es (see [28], [29]). In [28], a nonlinear plant model in state-space form and a plant model in controllable form, are considered, respecti v ely . Both plant models in v olv e kno wn constant coef ficients. The real output is defined as the first state, and is measured, whereas the other states are not. The output measurement is e xpressed as the sum of the real output plus a bounded measurement noise parameter . The observ er depends on the dif ference between the noisy output measurement and the output estimate. The stability analysis indicates that the state estimation error con v er ges to a residual set whose upper bound depends on the magnitude of both the measurement noise and the observ er parameters. Such upper bound has a global minimum for some v alue of the observ er g ain (see [28] and [29]). The main dra wbacks of the design are: the size of the residual set is unkno wn, so that the upper limit of the steady st ate of the state estimation error is unkno wn, and second, the coef ficients of the plant model are required to be kno wn. Interv al observ ers pro vide an upper and a lo wer bound for each unme asured state v ariable (s ee [30], [31]). The main disadv antage of the interv al observ ers is that se v eral upper and lo wer bounds of the plant model parameters are required to be kno wn. In [9], it is assumed that the measurements of the plant states are corrupted by noise and are described by a measurement model, which consists of a polynomial with respect to the real state v ector , with de gree one and unkno wn b ut bounded time-v arying coef ficients, being the time deri v ati v es of such coef ficients unkno wn and bounded. If each measurement model is dif ferentiated with respect to time, the time deri v ati v e of the state mea- surement is a linear polynomial of de gree one with respect to the time deri v ati v e of the real state v ector . The states resulting from the backstepping state transform ation are defined in terms of the noisy measurements ins tead of the real st ates. T o compute the time deri v ati v e of each quadratic function of the backstepping procedure, each measurement model is dif ferentiated. In [32], the state adapti v e backstepping (SAB) of [5] is used as the basic frame w ork for controlling a DC permanent magnet motor whose v oltage is supplied by a b uck po wer con v erter , b ut the ef fect of noisy measurement is not tak en into account. In the present paper , significant modifications are incorporated to the control scheme presented i n [32], in order to handle the ef fect of measurement noise: i) a measurement model is used to define the relationship between the motor current ( i a ) and the motor speed ( W m ), and their corresponding noisy measurements ( i a j m and W m j m , ii) the states of the backstepping state transformation are defined in terms of the noisy measurements, and Contr olling a DC Motor thr ough L ypauno v-lik e Functions and SAB ... (Alejandr o Rinc ´ on) Evaluation Warning : The document was created with Spire.PDF for Python.
2182 ISSN: 2088-8708 iii) the dif ferentiation of each quadratic-lik e function with respect to time in v olv es the dif ferentiation of each mea- surement model. The controller is implemented in a digital platform to carry out the real e xperiments. The main contrib ution of this paper respect to close ones are: a) the relationship between the real and the measured state is tak en into account in the control design procedure: in the definition and dif ferentiation of the backstepping states as well as in the definition and dif ferentiation of the quadratic functions, b) upper or lo wer bounds of the noise model parameters or their combination are not required to be kno wn by the controller , and c) the con v er gence of the tracking error to a residual set of user -defined size is pro v en in presence of noisy measurments. The rest of the paper is or g anized as follo ws. In section 2., the plant model and the control goal are described. In section 3. the controller is designed. In section 4., the bounded nature of the closed loop signals and the con v er gence of the tracking error are pro v en. In section 5., numerical and e xperimental results are presented. Finally , discussion and conclusions are presented in Section 6.. 2. PLANT MODEL AND CONTR OL GO AL The DC motor is represented by the follo wing plant model (see [33]): _ W m = B J eq W m + k t J eq i a ( T f r ic + T L ) J eq _ i a = R a L a i a k e L a W m + 1 L a u (1) The state v ariables are: the armature current i a and the motor speed W m . u is the control input and it corresponds to a v oltage v alue. The system output is y = W m . The model parameters are: the v oltage constant k e [V/rad/s], the armature inductance L a [H], the armature resistance R a [ ], the viscous friction coef ficient B [N.m/rad/s], the inertia moment J eq [kg.m 2 ], the motor torque constant k t [N.m/A], the friction torque T f r ic [N.m], and the load torque T L [N.m]. The follo wing assumptions are made for the model (1): Ai) the parameters T L and J eq are time-v arying, unkno wn and upper bounded by unkn o wn constants, and J eq is positi v e and lo wer bounded by an unkno wn positi v e constant, Aii) parameters B , k t , R a , L a , k e are unkno wn, positi v e and constant, and Aiii) W m and i a are measured, b ut their measurements W m j m , i a j m are noisy and satisfy: W m j m = a 6 W m + a 7 and i a j m = a 8 i a + a 9 (2) where the parameters a 6 , a 7 , a 8 , and a 9 are unkno wn, time-v arying and upper bounded by unkno wn positi v e constants, their time deri v ati v es are unkno wn, time-v arying and bounded by unkno wn positi v e constants and the parameters a 6 and a 8 are positi v e and lo wer bounded by unkno wn positi v e constants. The abo v e e xpressions are based on [9]. F or a simpler control design, the plant model (1) is re written as: _ x 1 = a 1 x 1 + a 2 x 2 a 3 (3) _ x 2 = a 4 x 1 a 5 x 2 + bu (4) x 1 = W m ; x 2 = i a ; u = v c ; y = W m (5) y m = W m j m = a 6 x 1 + a 7 (6) x 2 m = i a j m = a 8 x 2 + a 9 (7) where a 1 = B J eq ; a 2 = k t J eq (8) a 3 = ( T f r ic + T L ) J eq ; a 4 = k e L a ; a 5 = R a L a ; b = 1 L a (9) so that a 1 , a 2 , a 3 , a 4 , a 5 , and b are positi v e, a 1 , a 2 , and a 3 are time-v arying, a 4 , a 5 and b are constant, and y m and x 2 m are the noisy measurements of W m and i a , respecti v ely . Assumption Ai implies that the parameters a 1 , a 2 , a 3 are unkno wn and time -v arying, b ut the y are upper bounded by unkno wn constants. Assumption Aii implies that a 2 > 0 , and the parameters a 4 , a 5 , and b are unkno wn and constant. Assumption Aiii impli es that the states x 1 = y and x 2 are unkno wn b ut their measurements y m and x 2 m are kno wn, the parameters a 6 , a 7 , a 8 , a 9 , _ a 6 , _ a 7 , _ a 8 , and _ a 9 are unkno wn and time-v arying b ut bounded, and the parameters a 6 and a 8 are positi v e. In summary , the system (1) satisfies the follo wing properties: Pi) a 1 , a 2 , a 3 , a 4 , a 5 , b , a 6 , and a 8 are positi v e, Pii) the parameters IJECE V ol. 8, No. 4, August 2018: 2180 2198 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2183 a 1 , a 2 , a 3 , a 6 , a 7 , a 8 , a 9 , _ a 6 , _ a 7 , _ a 8 , and _ a 9 are unkno wn, time-v arying and upper bounded by unkno wn constants, Piii) the parameters a 2 , a 6 , a 8 , and 1 =a 8 are positi v e and lo wer bounded by unkno wn positi v e constants, Pi v) the parameters a 4 , a 5 , and b are unkno wn, positi v e and constant, and Pv) the v alues of y = x 1 = W m and x 2 = i a are unkno wn, whereas their measurements y m and x 2 m are kno wn. T aking into account the model, the control goal ca n be defined as follo ws. Let the follo wing reference model: y d = a m 1 _ y d a mo y d + a mo W mr ef (10) where W mr ef is the user –defined reference v alue, and a m 1 and a mo are user -defined positi v e constants. Hence, the desired output y d is pro vided by (10) subject to i) a mo and a m 1 are user -defined b ut positi v e and constant, and ii) W mr ef is user -defined and bounded b ut it may be time-v arying. Therefore, equation (10) is a stable reference model. The tracking error is defined as: e ( t ) = y m ( t ) y d ( t ) = W m j m y d (11) e = f e : j e j C be g (12) where y m is defined in (6), y d is the desired o ut put and is pro vided by (10), e is a residual set whose s ize is defined by C be which is an user -defined positi v e constant. The goal of the control design is to formulate a control la w and an update la w for the plant model (1), subject to assumptions Ai to Aiii, such t hat: CGi) the tracking error e asymptotically con v er ges to the residual set e , CGii) the control and update la ws do not in v olv e discontinuous signals, CGiii) the control la w and the updated parameters are bounded, and CGi v) all the closed loop signals are bounded. 3. CONTR OL DESIGN In this section, a controller for the plant (1) is de v eloped, it tak es into account the assumptions Ai to Aiii, and the goals CGi to CGi v stated pre viously . The procedure is similar to that in [32], b ut there are se v eral dif fer - ences due to the presence of measurement noise. Therefore, the procedure omits the steps that are quite similar to those in [32]. The state adapti v e backstepping (SAB) presented in [5] is used as control frame w ork, b ut important modifications are incorporated i n order to tackle the ef fect of unkno wn time-v arying plant model coef ficients and measurement noise. The controller design procedure is or g anized in the follo wing steps: i) define the first ne w state z 1 and dif ferentiate it with respect to time, ii) define a quadratic function V z 1 that depends on z 1 , and dif ferentiate it with respect to time, iii) e xpress the terms that in v olv e time-v arying coef ficients, as functions of upper constant bounds, and parameterize such bounds as function of parameter and re gression v ectors, i v) e xpress the parameter v ector in terms of updating error v ector and update parameter v ector , and define the second ne w state z 2 , v) dif ferentiate z 2 with respect to time, define a quadratic function V z that depends on z 1 and z 2 , and dif ferentiate it with respect to time, vi) e xpress the terms that in v olv e time v arying coef ficients as function of upper constant bounds, and parameterize in terms of parameter and re gression v ectors, vii) e xpress the parameter v ector in terms of updating error v ector and updated parameter v ector , and formulate the control la ws, and viii) formulate the L yapuno v-lik e function, dif ferentiate it with respect to time and formulate the update la ws. Step 0. In this step, the model (3)-(4) is e xpressed in terms of y m , x 2 m , using the noise models (6)-(7). Dif ferentiating (6) and using (3), yields: _ y m = ( _ a 6 a 6 a 1 ) x 1 + a 2 a 6 x 2 a 3 a 6 + _ a 7 (13) solving (6) and (7) for x 1 and x 2 and substituting into the abo v e e xpression we obtai n the basic e xpression for _ y m : _ y m = ( _ a 6 a 6 a 1 ) a 6 y m + a 2 a 6 a 8 x 2 m a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 : (14) Dif ferentiating (7) with respect to time and incorporating (4), yields: _ x 2 m = a 4 a 8 x 1 + ( _ a 8 a 5 a 8 ) x 2 + a 8 bu + _ a 9 (15) Contr olling a DC Motor thr ough L ypauno v-lik e Functions and SAB ... (Alejandr o Rinc ´ on) Evaluation Warning : The document was created with Spire.PDF for Python.
2184 ISSN: 2088-8708 solving (6) and (7) for x 1 and x 2 and incorporating in the abo v e equation we obtai n the basic e xpression for _ x 2 m which is _ x 2 m = a 4 a 8 a 6 y m + _ a 8 a 5 a 8 a 8 x 2 m + a 8 bu + a 4 a 8 a 7 a 6 +( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 (16) Remark 3..1 In or der to consider the noisy y m and x 2 m instead of the r eal b ut unknown values x 1 and x 2 , in the r emaining pr ocedur e equations (14) and (16) ar e used instead of equations (3) and (4) . Step 1. In this step, the first state v ariable is defined and dif ferentiated with respect to time. The state v ariable z 1 is defined as the tracking error: z 1 = e = y m y d (17) where y d is pro vided by (10). Dif ferentiating (17) with respect to time and using (14), yields: _ z 1 = _ y m _ y d (18) _ z 1 = a 1 a 6 + _ a 6 a 6 y m + a 2 a 6 a 8 x 2 m + ( a 1 a 6 _ a 6 ) a 7 a 6 a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 _ y d (19) Step 2. In this step, a quadratic function that depends on z 1 is defined and dif ferentiated with respect to time. Such quadratic form is defined as: V z 1 = (1 = 2) z 2 1 (20) Dif ferentiating (20) with respect to time, using (19) and adding and substracting c 1 z 2 1 yields: _ V z 1 = z 1 _ z 1 = c 1 z 2 1 + z 1 a 2 a 6 a 8 x 2 m + z 1 a 1 a 6 + _ a 6 a 6 y m + ( a 1 a 6 _ a 6 ) a 7 a 6 a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 + c 1 z 1 _ y d (21) The term c 1 z 2 1 has been added to obtain asymptotic con v er gence of the tracking error later . The unkno wn and time v arying beha vior of the bounded parameters a 1 , a 2 , a 3 , a 6 , _ a 6 , a 7 , _ a 7 , a 8 , and a 9 is a significant obstacle for the controller design. F or this reason, the terms that in v olv e such coef ficients will be e xpressed as function of upper constant bounds. Step 3. Recall that a 1 , a 2 , a 3 , a 6 , a 7 , a 8 , a 9 , _ a 6 , and _ a 7 are time-v arying, unkno wn and bounded. In this step, the terms that in v olv e such time-v arying parameters are e xpressed as function of upper and lo wer constant bounds, and such bounds are parameterized in terms of parameter and re gression v ectors. The term that in v olv es the brack ets in (21) yields: z 1 a 1 a 6 + _ a 6 a 6 y m + ( a 1 a 6 _ a 6 ) a 7 a 6 a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 + c 1 z 1 _ y d 10 j y m jj z 1 j + 11 j z 1 j + j c 1 z 1 _ y d jj z 1 j (22) where 10 , 11 are unkno wn positi v e constants such that a 1 a 6 + _ a 6 a 6 10 ( a 1 a 6 _ a 6 ) a 7 a 6 a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 11 (23) As mentioned in [32], Y oung’ s inequal ity must be applied to (22), such that the j z 1 j term leads to z 2 1 , to allo w a proper definition of z 2 . Arranging (22), and applying Y oung’ s inequality ([34]), yields: z 1 a 1 a 6 + _ a 6 a 6 y m + ( a 1 a 6 _ a 6 ) a 7 a 6 a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 + c 1 z 1 _ y d c a 10 c a j y m jj z 1 j + c a 11 c a j z 1 j + c a j c 1 z 1 _ y d jj z 1 j c a (24) c 2 a 2 + 2 11 2 c 2 a z 2 1 + c 2 a 2 + 2 10 2 c 2 a y 2 m z 2 1 + c 2 a 2 + ( c 1 z 1 _ y d ) 2 z 2 1 2 c 2 a (25) IJECE V ol. 8, No. 4, August 2018: 2180 2198 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2185 where c a is a posi ti v e constant that should be chosen to fulfill certain conditions that will be defined later . Substi- tuting into (21), yields: _ V z 1 c 1 z 2 1 + 3 c 2 a 2 + z 1 a 2 a 6 a 8 x 2 m + 2 11 2 c 2 a z 2 1 + 2 10 2 c 2 a y 2 m z 2 1 + ( c 1 z 1 _ y d ) 2 z 2 1 2 c 2 a (26) The terms that in v olv e x 2 m , z 2 1 , x 2 1 z 2 1 , ( c 1 z 1 _ y d ) 2 should be grouped in a ne w kno wn state v ariable z 2 , according to the procedure in [5]. Ne v ertheless, the unkno wn time-v arying beha vior of parameter a 2 a 6 =a 8 poses a significant obstacle. T o remedy such situation, a positi v e constant lo wer bound of a 2 a 6 =a 8 will be used. Property Pi mentions that a 2 , a 6 , and a 8 are positi v e, wher eas property Piii implies that a 2 , a 6 , and 1 =a 8 are lo wer bounded by unkno wn positi v e constants. Therefore, 0 < l 12 a 2 a 6 a 8 (27) where l 12 is an unkno wn positi v e constant lo wer bound. Incorporating the constant l 12 into (26) and arranging, yields: _ V z 1 c 1 z 2 1 + 3 c 2 a 2 + z 1 a 2 a 6 a 8 x 2 m + 1 2 c 2 a 2 11 l 12 l 12 z 2 1 + 1 2 c 2 a 2 10 l 12 l 12 y 2 m z 2 1 + 1 2 c 2 a 1 l 12 l 12 z 2 1 ( c 1 z 1 _ y d ) 2 (28) = c 1 z 2 1 + 3 c 2 a 2 + z 1 a 2 a 6 a 8 x 2 m + ' > 1 1 2 c 2 a l 12 z 2 1 (29) where ' 1 is a kno wn re gression v ector and 1 is an unkno wn constant parameter v ector gi v en by: ' 1 = [1 ; y 2 m ; ( c 1 z 1 _ y d ) 2 ] > (30) 1 = 2 11 l 12 ; 2 10 l 12 ; 1 l 12 > ; (31) Step 4. In this step, the unkno wn parameter v ector 1 is e xpressed in terms of an updating error v ector and an updated parameter v ector , and then a ne w state v ariable z 2 is defined. The parameter v ector 1 can be re written as: 1 = ^ 1 ~ 1 (32) where ~ = ^ 1 2 11 l 12 ; 2 10 l 12 ; 1 l 12 > (33) where ^ 1 is an updated parameter v ector pro vided by an updating la w that will be defined later , and ~ 1 is an updating error . Substituting (32) into (29), yields: _ V z 1 c 1 z 2 1 + 3 c 2 a 2 + z 1 a 2 a 6 a 8 x 2 m + ' > 1 ^ 1 1 2 c 2 a l 12 z 2 1 ' > 1 ~ 1 1 2 c 2 a l 2 z 2 1 As can be noticed from [23] and [24] the updated parameter ^ 1 is non-ne g ati v e, so that j ^ 1 j = ^ 1 for ^ 1 ( t o ) 0 . The accomplishment of this property will be sho wn later . In vie w of this f act and incorporating the inequality (27) in the term ' > 1 ^ 1 0 : 5 c 2 a l 12 z 2 1 and arranging, yields: _ V z 1 c 1 z 2 1 + 3 c 2 a 2 + z 1 a 2 a 6 a 8 x 2 m + ' > 1 ^ 1 1 2 c 2 a a 2 a 6 a 8 z 2 1 ' > 1 ~ 1 1 2 c 2 a l 2 z 2 1 (34) = c 1 z 2 1 + 3 c 2 a 2 + z 1 a 2 a 6 a 8 z 2 ' > 1 ~ 1 1 2 c 2 a l 2 z 2 1 (35) Contr olling a DC Motor thr ough L ypauno v-lik e Functions and SAB ... (Alejandr o Rinc ´ on) Evaluation Warning : The document was created with Spire.PDF for Python.
2186 ISSN: 2088-8708 where z 2 = x 2 m + ' > 1 ^ 1 1 2 c 2 a z 1 (36) Step 5. In this step, the state v ariable z 2 is dif ferentiated with respect to time, a quadratic function V z is defined as function of z 1 and z 2 , and such function is dif ferentiated with respect to time. Dif ferentiating (36) with respect to time, yields: _ z 2 = _ x 2 m + _ ' > 1 ^ 1 1 2 c 2 a z 1 + ' > 1 _ ^ 1 1 2 c 2 a z 1 + ' > 1 ^ 1 1 2 c 2 a _ z 1 (37) where _ ' 1 = [0 ; 2 y m _ y m ; 2( c 1 z 1 _ y d )( c 1 _ z 1 y d )] > (38) Substituting (18) into (37) and arranging, yields: _ z 2 = _ x 2 m + ' 1 b _ y m + ' 1 c (39) where ' 1 b = 1 2 c 2 a 2 y m ^ 1[2] + c 1 ( c 1 z 1 _ y d ) ^ 1[3] z 1 + ' > 1 ^ 1 (40) ' 1 c = 2( c 1 z 1 _ y d )( c 1 _ y d + y d ) ^ 1[3] 1 2 c 2 a z 1 + ' > 1 _ ^ 1 1 2 c 2 a z 1 ' > 1 ^ 1 1 2 c 2 a _ y d (41) ' 1 b and ' 1 c are kno wn scalar functions, ^ 1[2] and ^ 1[3] are the second and third entries of the v ector ^ 1 , and y d , _ y d , and y d are pro vided by (10). Substituting (14) and (16) into (39) and arranging, yields: _ z 2 = a 4 a 8 a 6 y m + ( _ a 8 a 5 a 8 ) a 8 x 2 m + ( _ a 6 a 6 a 1 ) a 6 ' 1 b y m + a 2 a 6 a 8 ' 1 b x 2 m + a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 ' 1 b + a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 + a 8 bu + ' 1 c (42) The quadratic form that depends on z 1 and z 2 is defined as: V z = (1 = 2)( z 2 1 + z 2 2 ) (43) Dif ferentiating with respect to time, incorporating (35) and (42), and adding and subtracting c 2 z 2 2 , yields: _ V z = z 1 _ z 1 + z 2 _ z 2 = _ V z 1 + z 2 _ z 2 c 1 z 2 1 c 2 z 2 2 + 3 2 c 2 a + z 2 a 2 a 6 a 8 z 1 a 4 a 8 a 6 y m + _ a 8 a 5 a 8 a 8 x 2 m + _ a 6 a 6 a 1 a 6 ' 1 b y m + a 2 a 6 a 8 ' 1 b x 2 m + a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 ' 1 b + a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 + c 2 z 2 + a 8 bu + ' 1 c ' > 1 ~ 1 1 2 c 2 a l 12 z 2 1 (44) IJECE V ol. 8, No. 4, August 2018: 2180 2198 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2187 The control input u is defined as follo ws: u = u a + u b (45) where u a is an user -defined constant. The constant u a w as incorporated in order to a v oid abrupt control input beha vior at the be ginning of the closed loop operation. Therefore, it should be chosen as the open loop v alue of the control input. The control u b is established by the controller design. Substituting (45) into (44), yields: _ V z c 1 z 2 1 c 2 z 2 2 + 3 2 c 2 a + z 2 a 2 a 6 a 8 ( z 1 + ' 1 b x 2 m ) a 4 a 8 a 6 y m + _ a 8 a 5 a 8 a 8 x 2 m + _ a 6 a 6 a 1 a 6 ' 1 b y m + a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 ' 1 b + a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 + a 8 bu a + ' 1 c + c 2 z 2 a 8 bu b z 2 ' > 1 ~ 1 1 2 c 2 a l 12 z 2 1 (46) Step 6. Recall that a 1 , a 2 , a 3 , a 6 , a 7 , a 8 , a 9 , _ a 6 , _ a 7 , _ a 8 are unkno wn and time-v arying. In this step, the terms that in v olv e such parameters are e xpressed in terms of upper bounds, and paramet erized in terms of parameter and re gression v ectors. The term that in v olv es the squared brack ets can be re written as: z 2 a 2 a 6 a 8 ( z 1 + ' 1 b x 2 m ) a 4 a 8 a 6 y m + _ a 8 a 5 a 8 a 8 x 2 m + _ a 6 a 6 a 1 a 6 ' 1 b y m + a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 ' 1 b + a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 + a 8 bu a + ' 1 c + c 2 z 2 j z 2 j [ 13 j y m j + 14 j x 2 m j + 15 j ' 1 b y m j + 16 j z 1 + ' 1 b x 2 m j + 17 j ' 1 b j + 18 + 19 j u a j + j ' 1 c + c 2 z 2 j ] (47) where 13 , 14 , 15 , 16 , 17 , 18 , 19 are unkno wn positi v e constant upper bounds that satisfy: a 4 a 8 a 6 13 ; _ a 8 a 5 a 8 a 8 14 _ a 6 a 6 a 1 a 6 15 ; a 2 a 6 a 8 16 a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 17 a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 18 ; j a 8 b j 19 (48) From equations (46), (47), and as can be inferred from [32] there are t w o significant obstacles. First, the j z 2 j term may lead to discontinuous signals in the definition of u b . This can be remedied by using Y oung’ s inequality . Sec- ond, the unkno wn v arying beha vior of a 8 b mak es it dif ficult for u to eliminate the ef fect of the terms that in v olv e unkno wn time v arying parameters. This can be remedied by incorporating a positi v e constant lo wer bound of a 8 b . From properties Piii and Pi v (see page 2182) it follo ws that a 8 and b are positi v e, b is constant and a 8 is lo wer bounded by an unkno wn positi v e constant. Therefore, 0 < l 20 a 8 b (49) where l 20 is an unkno wn positi v e constant lo wer bound. Re writ ting (47) in terms of a parameter v ector and a Contr olling a DC Motor thr ough L ypauno v-lik e Functions and SAB ... (Alejandr o Rinc ´ on) Evaluation Warning : The document was created with Spire.PDF for Python.
2188 ISSN: 2088-8708 re gression v ector , and incorporating (49), yields: z 2 a 2 a 6 a 8 ( z 1 + ' 1 b x 2 m ) a 4 a 8 a 6 y m + _ a 8 a 5 a 8 a 8 x 2 m + _ a 6 a 6 a 1 a 6 ' 1 b y m + a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 ' 1 b + a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 + a 8 bu a + ' 1 c + c 2 z 2 p l 20 j z 2 j ' > 2 (50) where ' = [ j y m j ; j x 2 m j ; j ' 1 b y m j ; j z 1 + ' 1 b x 2 m j ; j ' 1 b j ; 1 ; j u a j ; j ' 1 c + c 2 z 2 j ] > (51) 2 = 1 p l 20 [ 13 ; 14 ; 15 ; 16 ; 17 ; 18 ; 19 ; 1] > ; (52) ' is a re gression v ector whose entries are kno wn, and 2 is a parameter v ector , whose entries are positi v e, constant and unkno wn, and l 20 is an unkno wn positi v e constant lo wer bound. The constant p l 20 has been incorporated in order to handle the ef fect of the unkno wn time-v arying parameter a 8 b appearing in the term a 8 bu b in (46). Step 7. Since the parameter v ector 2 is unkno wn, in this step it is e xpressed in terms of an updated parameter v ector and an updating error v ector; after of this the control la w is formulated. The parameter 2 can be re written as 2 = ^ 2 ~ 2 (53) where ^ 2 is an updated parameter v ector pro vided by an update la w which is defined in the step 8, and ~ 2 is an updating error v ector gi v en by ~ 2 = ^ 2 1 p l 20 [ 13 ; 14 ; 15 ; 16 ; 17 ; 18 ; 19 ; 1] > (54) Substituting (53) into (50) yields: z 2 a 2 a 6 a 8 ( z 1 + ' 1 b x 2 m ) a 4 a 8 a 6 y m + _ a 8 a 5 a 8 a 8 x 2 m + _ a 6 a 6 a 1 a 6 ' 1 b y m + a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 ' 1 b + a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 + a 8 bu a + ' 1 c + c 2 z 2 p l 20 j z 2 j ' > ^ 2 p l 20 j z 2 j ' > ~ 2 (55) incorporating the inequality (49) and applying Y oung’ s inequality (cf. [34] pp. 123) to the term p l 20 j z 2 j ' > ^ 2 , yields: z 2 a 2 a 6 a 8 ( z 1 + ' 1 b x 2 m ) a 4 a 8 a 6 y m + _ a 8 a 5 a 8 a 8 x 2 m + _ a 6 a 6 a 1 a 6 ' 1 b y m + a 2 a 6 a 9 a 8 a 3 a 6 + _ a 7 ( _ a 6 a 6 a 1 ) a 7 a 6 ' 1 b + a 4 a 8 a 7 a 6 + ( _ a 8 + a 5 a 8 ) a 9 a 8 + _ a 9 + a 8 bu a + ' 1 c + c 2 z 2 c 2 c 2 + 1 2 c 2 c a 8 bz 2 2 ( ' > ^ 2 ) 2 p l 20 j z 2 j ' > ~ 2 (56) where c c is a positi v e constant that satisfies some some conditions that will be defined in the Step 8. Substituting (56) into (46) and arranging yields: _ V z c 1 z 2 1 c 2 z 2 2 + 3 2 c 2 a + c 2 c 2 + a 8 bz 2 u b + 1 2 c 2 c z 2 ( ' > ^ 2 ) 2 ' > 1 ~ 1 1 2 c 2 a l 20 z 2 1 p l 20 j z 2 j ' > ~ 2 (57) IJECE V ol. 8, No. 4, August 2018: 2180 2198 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2189 with the aim to cancel the ef fect of the term a 8 bz 2 (1 = 2) c 2 2 z 2 ( ' > ^ 2 ) 2 , the e xpression for u b is chosen as: u b = 1 2 c 2 c z 2 ( ' > ^ 2 ) 2 (58) In vie w of (45), the control la w for u is: u = u a 1 2 c 2 c z 2 ( ' > ^ 2 ) 2 (59) substituting (59) into (57), yields: _ V z 2 min f c 1 ; c 2 g V z + 3 2 c 2 a + c 2 c 2 ' > 1 ~ 1 1 2 c 2 a l 20 z 2 1 p l 20 j z 2 j ' > ~ 2 The abo v e e xpression implies that the ti me deri v ati v e of the L yapuno v function w ould contain the term (3 = 2) c 2 a + (1 = 2) c 2 c , so that the required ne g ati v eness properties w ould be altered. Therefore, the quadratic function V z is considered, which is a truncated function of V z and v anishes when V z is lo wer or equal than the constant C bv z . The quadratic function V z is defined as: V z = (1 = 2)( p V z p C bv z ) 2 if V z C bv z 0 otherwise (60) C bv z = (1 = 2) C 2 be (61) where V z is defined in (43). The function defined by (60) and (61) has the follo wing properties: V z 0 , V z 3 C bv z + 3 V z and V z and @ V z =@ V z are locally Lipschitz continuous. Dif ferentiating (60) with respect to time, yields: d V z dt = @ V z @ V z _ V z (62) where @ V z @ V z = ( 1 2 p V z p C bv z p V z if V z C bv z 0 otherwise (63) Combining (62) with (60) yields: d V z dt 2 min f c 1 ; c 2 g V z @ V z @ V z + 3 2 c 2 a + c 2 c 2 @ V z @ V z ' > 1 ~ 1 1 2 c 2 a l 20 z 2 1 @ V z @ V z p l 20 j z 2 j ' > ~ 2 @ V z @ V z (64) Step 8. In this step, the L yapuno v-lik e function is formulated and dif ferentiated with respect to time, and the update la ws are formulated. The L yapuno v-lik e function is defined as: V ( x ( t )) = V z + V (65) x ( t ) = [ z 1 ( t ) ; z 2 ( t ) ; ~ > 1 ; ~ > 2 ] (66) V = (1 = 2) l 20 ~ > 1 1 1 ~ 1 + (1 = 2) p l 20 ~ > 2 1 2 ~ 2 (67) where ~ 1 and ~ 2 are defined in (33) and (54) respecti v ely , and V z is defined in (60). The v ector x ( t ) contains the closed loop stat es z 1 ( t ) , z 2 ( t ) , ~ > 1 , ~ > 2 . F or the sak e of simplicity , V ( x ( t )) is represented as V . Dif ferentiating (65) and (67) with respect to time, yields: _ V = _ V z + _ V (68) _ V = l 20 ~ > 1 1 1 _ ^ 1 + p l 20 ~ > 2 1 2 _ ^ 2 (69) Incorporating (64) and (69) into (68), yields: _ V 2 min f c 1 ; c 2 g V z @ V z @ V z + 3 2 c 2 a + c 2 c 2 @ V z @ V z + l 20 ~ > 1 ' 1 1 2 c 2 a z 2 1 @ V z @ V z + 1 1 _ ^ 1 + p l 20 ~ > 2 j z 2 j ' @ V z @ V z + 1 2 _ ^ 2 (70) Contr olling a DC Motor thr ough L ypauno v-lik e Functions and SAB ... (Alejandr o Rinc ´ on) Evaluation Warning : The document was created with Spire.PDF for Python.