IOP Conerence Series: Materials Science and Engineering PAPER OPEN ACCESS Passive Identiication is Non Stationary Obects With Closed Loo Control To cite this article: Valeriy F Dyadik et al 2016 IOP Con. Ser.: Mater. Sci. Eng. 142 012047 View the article online or udates and enhancements. This content was downloaded rom IP address 148.251.232.83 on 27/09/2018 at 16:34
VII International Scientiic Practical Conerence "Innovative Technologies in Engineering" IOP Publishing IOP Con. Series: Materials Science and Engineering 142 (2016) 012047 doi:10.1088/1757-899x/142/1/012047 Passive Identiication is Non Stationary Obects With Closed Loo Control Valeriy F Dyadik 1, Igor S Nadezhdin 1, Alexey G Goryunov 1, Flavio Manenti 2 1 Tomsk Polytechnic University, Institute o Physics and Technology, Deartment o Electronics and Automation o Nuclear Plants, Russia, Tomsk, Lenina St 30, 634050 2 Politecnico di Milano, Diartimento di Chimica, Materiali e Ingegneria Chimica «Giulio Natta» Italy, Milano, Piazza Leonardo da Vinci 32, 20133 E-mail: kun9@list.ru Abstract. Tyically chemical rocesses have signiicant nonlinear dynamics, but desite this, industry is conventionally still using PID-based regulatory control systems. Moreover, rocess units are interconnected, in terms o inlet and outlet material/energy lows, to other neighbouring units, thus their dynamic behaviour is strongly inluenced by these connections and, as a consequence, conventional control systems erormance oten roves to be oor. However, there a hybrid uzzy PID control logic, whose tuning arameters are rovided in real time. The uzzy controller tuning is made on the basis o Mamdani controller, also exloiting the results coming rom an identiication rocedure that is carried on when an unmeasured ste disturbance o any shae aects the rocess behaviour. This aer resents rocedure or identiying technological obect control in a closed loo, i. e. that oerates the automated control system. The variation in the controlled variable, caused by the change o the nonmeasurable disturbance, is considered the initial signal or the identiication rocedure. The arameters o the control obect are ound by otimization method Levenberg-Marquardt. 1. Introduction Nowadays, in chemical and nuclear industries (i.e. integrated searations, extractions crystallization rocesses to uriy U and Pu rom other usion side-comonents) any rocessing ste has a high level o automation. On the one hand, all rocesses are high resonsibility technology (HRT), i.e. high erormance technology with resect o saety level. On the other hand, they are also comlex dynamic systems. Exerience shows that the usually the conventional PID controller is not eective or comlex dynamic systems [1]. A number o researchers have conducted studies to develo a hybrid uzzy PID control logic, whose tuning arameters are rovided in real time. The uzzy controller tuning is made on the basis o Mamdani controller, also exloiting the results coming rom an identiication rocedure. The roblem o identiying has attracted and attracts a lot o research interests and many methods are available or this roblem in literature [2]. A riori inormation about the dynamics o the controlled lant is required or the synthesis o PIDbased uzzy logic controller. Hammerstein and Wiener models may be used to describe comlex dynamics real-lie rocesses [3, 4]. Hammerstein and Wiener models are methodologies constituted by the combination o a static nonlinearity (N) and a linear system (L), resectively in the N-L and L-N Content rom this work may be used under the terms o the Creative Commons Attribution 3.0 licence. Any urther distribution o this work must maintain attribution to the author(s) and the title o the work, ournal citation and DOI. Published under licence by IOP Publishing Ltd 1
VII International Scientiic Practical Conerence "Innovative Technologies in Engineering" IOP Publishing IOP Con. Series: Materials Science and Engineering 142 (2016) 012047 doi:10.1088/1757-899x/142/1/012047 orm. The roblem o identiying N and L rom inut-outut data has attracted and attracts a lot o research interests and many methods are available or this roblem in literature [5]. The nonlinear dynamic system can be aroximated by a linear dynamic system near the oerating oint, which is suicient or PID tuning. It is not a simle task to deine the arameters o the linear dynamic model aroximation in the closed-loo system. In [6, 7] active methods o identiication are roosed; here sine waves in inut are used to excite the Wiener continuous-time system and requency methods are used to determine the unknowns. Unknown additive disturbances create roblems or closed-loo identiication. Good results can be obtained by using MATLAB system identiication toolbox or the identiication o the arameters o the rocess with the use o ARX, ARMAX, BJ state sace, olynomial models and others [8]. 2. Material and methods or assive identiication The roosed method emloys algorithms or the lant identiication couled with uzzy systems such as Mamdani controllers [9]. The layout o a generic automatic control systems lant is resented in Fig. 1 while a scheme o an adative uzzy controller is shown in Fig. 2. Figure 1. Fuzzy adative control system. g : reerence signal; * : non-measurable disturbance; : measurable disturbance; P u : lant control channel; P : lant disturbance channel; P * : lant non-measurable disturbance channel; y: controlled variable; : control error is deined as = g y. y Identiier k, T, τ ε Controller u Figure 2. Adative uzzy controller or automatic control systems. 2
VII International Scientiic Practical Conerence "Innovative Technologies in Engineering" IOP Publishing IOP Con. Series: Materials Science and Engineering 142 (2016) 012047 doi:10.1088/1757-899x/142/1/012047 The oerating rincile o identiier (Fig. 2), is shown in Figure 3. In this aer use the ste disturbance variables, but is not known value and duration o these disturbances. Start identiication. Measurement o the disturbance () The measurement o the controlled variable (y) y Deining the arameters o the controlled obect k, T, τ Measurement o the control action (u) u t 0 t 1 t 2 t 3 Figure 3. The rincile o oeration identiier. The identiication is erormed in the closed-loo system in those oerating conditions where the edge o the transient is reached (see Fig. 4). Figure 4. Transient o the automatic control systems or the identiication o lant arameters in a closed loo. 3
VII International Scientiic Practical Conerence "Innovative Technologies in Engineering" IOP Publishing IOP Con. Series: Materials Science and Engineering 142 (2016) 012047 doi:10.1088/1757-899x/142/1/012047 The variation in the controlled variable, caused by the change o the non-measurable disturbance *, is considered the initial signal or the identiication rocedure. The time instant t 0 where the nonmeasurable ste disturbance * undergoes a ste change is unknown. The time instant t 1 is deined by the deviation threshold o y rom g by y > y g, where y g is the required control accuracy, аnd t 2 is the time instant o the y variable sign change. The arameters o the lant are deined by Levenberg- Marquardt otimization method. In this case, the measured disturbance, the control action u and the controlled variable y (see Fig. 1 4) are sulied to the identiier inut in the time interval whose lower and uer bounds are, resectively, t 2 and t 3. The arameters o the lant (k, T, ) and those related to the disturbance (k, T, ) are otimized, in accordance with the reorted statements, by the iterative rediction-error estimation method (em) by Matlab:,,,,,, m u y y u k T y k T n m 2 S y y n 1 (1) 1 S 0 where y u, y are a digital reresentation o the lant model with lagging or channels u and, S is the otimized unctional, n is the dimension o access. The dynamics o the lant is described as a linear system with the transer unction W u (s) that reresents channel u and the transer unction W (s) that stand or channel : k u s k W ( s) e, W ( s) e T s 1 T s 1 s (2) 3. Results o the assive identiication with closed loo control The roosed the rocedure or identiying controlled obect was used in the identiier, which is used in the develoment o automated control system with uzzy logic. The identiier and the automatic control system was are imlemented in the ackage MATLAB/Simulink. Realized identiier makes determination o the arameters o the transer unctions o the control obect on control and disturbance (2). Table 1 shows the results o identiication o arameters o control obect ater making the system disturbances. Into the system was introduced a ste disturbance with unknown characteristics. Table 1. The results o the identiication. Obect arameters The arameter values o the transer unction o the obect The results o the identiication Control The erturbation Control The erturbation k 4.3 4 4.7 4.4 T 1 90 90 107.8 107.1 T 2 25 - - - T 3 2.5 - - - τ 12 30 62.6 26 4
VII International Scientiic Practical Conerence "Innovative Technologies in Engineering" IOP Publishing IOP Con. Series: Materials Science and Engineering 142 (2016) 012047 doi:10.1088/1757-899x/142/1/012047 The controlled obect is reresented transer unction the third order. As a result, the identiication o the control obect was described by the transer unction o the irst order with delay. The discreancy o transient rocesses ater identiication was 0.3%, the transient rocesses are resented in Fig. 5. Figure 5. The transition rocess ("the downturn", ater the imact o the disturbance) or the identiication arameters o the controlled obect It was also investigated the work identiier at entering the system ste disturbances with noise. As the noise intererence used additive white Gaussian noise. The results o identiication are shown in Table 2. Table 2. The results o the identiication. Obect arameters The arameter values o the transer unction o the obect The results o the identiication Control The erturbation Control The erturbation k 4.3 4 10 8.6 T 1 90 90 239.3 241.7 T 2 25 - - - T 3 2.5 - - - τ 12 30 41.8 28.3 Fig. 6 shows the transition rocess that indicate that an error in the identiication o 1.2 %. 5
VII International Scientiic Practical Conerence "Innovative Technologies in Engineering" IOP Publishing IOP Con. Series: Materials Science and Engineering 142 (2016) 012047 doi:10.1088/1757-899x/142/1/012047 Figure 6. The transition rocess ("the downturn", ater the imact o the disturbance with noise) or the identiication arameters o the controlled obect The use o white noise exlained by its wide distribution in nature and technology. In ractice, most o the signals contain the noise. The use or research ste disturbance with noise, will allow more realistically simulate the identiication rocedure. As seen rom the transient rocesses, the roosed identiier erorms its unctions, even at disturbance with noise. 4. Conclusions As a result o the work was imlemented identiier in structure o system control with uzzy logic. The work was erormed in the ackage MATLAB/Simulink. Identiication o the arameters o the control obect is carried out in a closed loo, with ste disturbances o unknown magnitude and duration. Error occurred while identiy the arameters o the obect was 0.3 %. In studying o the roosed rocedures the identiication, on the disturbance was suerimosed noise. As the intererence was used additive white Gaussian noise. Error occurred while identiication o arameters o the obect at ste disturbance with white noise was 1.2 %. The results demonstrate the ossibility o determining the arameters o the obect in a closed loo, using the roosed identiication rocedure. 5. Acknowledgements This work was unded as art o the Federal government-sonsored rogram «Science» by Tomsk Polytechnic University. Reerences [1] Gao Z., Kong D., Gao C., Chen M. 2013. Modeling and control o comlex dynamic systems. Journal o Alied Mathematics. Article number 151372. DOI: 10.1155/2012/869792 [2] Ikhouane F., Giri F. 2014. A uniied aroach or the arametric identiication o SISO/MIMO Wiener and Hammerstein systems. Journal o the Franklin Institute. 351(3) 1717 27. [3] Biagiola S.I., Figueroa J.L. 2009. Wiener and Hammerstein uncertain models identiication. Mathematics and Comuters in Simulation. 79(11) 3296 313. [4] Demsey E.J., Westwick D.T. 2004. Identiication o Hammerstein models with cubic sline nonlinearities. IEEE Transactions on Biomedical Engineering. 51(2) 237 45. [5] Zhou L., Li X., Pan F. 2013. Gradient based iterative arameter identiication or Wiener nonlinear systems. Alied Mathematical Modelling. 37(16-17) 8203 09. [6] Giri F., Rochdi Y., Chaoui F.Z. 2009. An analytic geometry aroach to wiener system requency identiication. IEEE Transactions on Automatic Control. 54(4) 683 96. [7] Giri F., Rochdi Y., Radouane A., Brouri A., Chaoui F.Z. 2013. Frequency identiication o 6
VII International Scientiic Practical Conerence "Innovative Technologies in Engineering" IOP Publishing IOP Con. Series: Materials Science and Engineering 142 (2016) 012047 doi:10.1088/1757-899x/142/1/012047 nonarametric Wiener systems containing backlash nonlinearities. Automatica. 49(1) 124 37. [8] Rabbani M.J., Hussain K., Khan A.U.R., Ali A. 2013. Model identiication and validation or a heating system using Matlab system identiication toolbox. In: Proceedings o the 1st International Conerence on Sensing or Industry, Control, Communications, and Security Technologies, ICSICCST 2013, Karachi, Pakistan: June 2013. IOP Conerence Series: Materials Science and Engineering. 51(1) DOI: 10.1088/1757-899X/51/1/012022. [9] Xu X.Q. 2014. The alication o MATLAB or uzzy control system simulation. Alied Mechanics and Materials. 494-495 1306 9. 7