A Novel Fuzzy Controller Applicable to Steaming Room - IEEE Xplore

4 downloads 260 Views 1MB Size Report
Engineering campus, Nibong Tebal, 14300, Pulau Pinang,. Malaysia sadeghaminifar@yahoo.com [email protected]. ABSTRACT- This paper describes ...
2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia

A Novel Fuzzy Controller Applicable to Steaming Room Sadegh Aminifar Universiti Sains Malaysia School of Electrical and Electronic Engineering Engineering campus, Nibong Tebal, 14300, Pulau Pinang, Malaysia [email protected]

Arjuna bin Marzuki Universiti Sains Malaysia School of Electrical and Electronic Engineering Engineering campus, Nibong Tebal, 14300, Pulau Pinang, Malaysia [email protected] Three characters which are important regarding these controllers are 1- being nonlinear and unclear behavior of system [3]: Being nonlinear and ill-definition of system under control is a problem of various under control industrial parameters. a lot of industrial process especially our case of study has not clear mathematical model and also if it is even be defined with nonlinear parameters still it is not possible to present a model which cover all parameters 2- uncertain data obtained from sensors [4]: Because of that our sensors are not accurate they are also in a noisy environment there is a need for modeling their uncertainty. On the other hand in practical industrial there is a problem regarding input of controllers. Our input of controller is normally sensors which they are not accurate and also there is inaccuracy in their data. 3- hysteresis: Our actuator here is Industrial on-off valves which they have hysteresis. During using these valves in conditions with high pressure and very high temperature their stiction problems grows and this highly affects on the controller functionality. [5]. High static friction is a common problem in proportional and on-off control valves, which are widely used in the process industry. Recently, there have been many attempts to understand and detect stiction in control valves [6]. In valves changing the threshold voltage which causes hysteresis is usually expected [7].

ABSTRACT- This paper describes a new method of uncertainty description in fuzzy controller by introducing the concept of primary and corrector rule bases in fuzzy controllers and their roll in easing of extracting the optimum control surface by using too lesser rules than traditional fuzzy systems. The power of proposed method to simplify description of complicated systems shown by applying this method to control a steaming room system to reduce input valve amortization and setting the temperature of the autoclave room at defined points which guide us to reach better product quality of artificial stones and prevent misusing energy. Simulations confirm the lesser number of valve switching and better energy saving by applying proposed controller to the under control process compare to traditional counterparts and smoother rout compare to PID controllers. Key words: Fuzzy controller, Primary rule base, corrector rule base, uncertainty, steaming room

I.

PROBLEM STATEMENT

In artificial stone or terrazzo tile maker factories, the strength of the artificial stone or terrazzo tiles highly depend on good process of autoclave steaming room after choosing good composition of raw material and good pressing [1]. A welldesigned controller for such a system for achieving wanted process character directly affects on the quality of product [2]. In this research we want to propose a control system which has good flexibility with uncertain data and compatible with a valve with pre-defined hysteresis model in order to control the temperature of autoclave steam room of industrial stone

manufactures which is a system with high nonlinearity and unforeseeable parameters. The presence of nonlinearities limits the control loop performance [9]. After that we have nonlinear system with unknown

Autoclave Steaming Room With unforeseeable changes

Soft controller which receive uncertain data And applies to a valve

Valve

Data from sensors

Figure 1: Description of problems and research objectives

978-1-4673-3143-2/12/$31.00 ©2012 IEEE

171

2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia

parameters. The vapor which valve injects to ssystem will cause change in temperature. The temperature sensorr output also has nonlinear properties. The total concept of research objective is ddepicted in Fig. 1 This structure presented as proposal for folloowing two main purposes that are product high quality and eenergy efficiency which both are important in industrial. In Fig. 1 we can see one controller which involve with uncertain data of sensors and variation of parameters. This autoclave steam syystem is a system with high nonlinearity and also involves w with sensors and actuators which their parameters changes and are not certain. This research involves with control of such a ssystem with high nonlinearity and in difficulty to model it with claassical methods. Innovations in new methods for control witth malfunctioned elements in real industrial environments are the aim of this paper in order to increase the quality of product andd increase energy efficiency. The autoclave system is controlled by tw wo conventional methods: on–off and PID. Simplicity and loow cost are the advantages of on-off method but to follow the innconstant surfaces are difficult and also amortization of input vvalve is another drawback. In the other method, exact following of the level is an advantage of this method but it is complicated annd expensive and valve amortization is high, too. Another alternative is using fuzzy controller for managing the product process. The main point that we will invvolve in this case is that how we can define membership funcctions which are compatible with linguistic terms that conclude aan output control sphere which is able to control the temperature oof autoclave room so that we get better product quality and lesser wasting of input vapor. This paper wants to supervise the compliccated system via

II.

DESCRIPTION OF O TOTAL SYSTEM AND METH HODOLOGY

The total block diagram of all under u control system is shown in Fig. 2. The autoclave system haas one on-off input valve and one on-off output valve. In this caase, the output valve is set to low output flow just to have the same s pressure as atmosphere pressure without losing significant energy. e This paper considered the pressure as atmosphere pressuree and follows the changes of temperature without changing th he pressure. The input of controller is temperature. This inpu ut is uncertain due to the type of sensors which are used in the autoclave a and steaming room environments. Another uncertaainty is regarding the corresponding word and crisp quanttities. Because of final cooling stage of the autoclave process, it is i necessary to use water tap shown in Fig. 2 in order to satisfy th he optimum conditions. expert human knowledge which desscribes the system with words and control it according the rules which are extracted from knowledge base. Fuzzy type II sy ystems are the existent model for handling these uncertainties [10]. However, there are still variouss deficiencies regarding fuzzy type II systems to model the crisp data as words and handling the uncertainties especially involvin ng a system with high nonlinearities and real probabilisttic space expectation in the output. Our idea is toward to present a corrector fuzzy system which can govern such a system with w high nonlinearity in the presence of various uncertainty typees and hysteresis. The new method which is proposed in this paper to implement controller of the autocclave system has almost all advantages of the other three meth hods that are on-off, PID and conventional fuzzy method. It contrrols the rate of valve opening

Figure 2: The prractical block diagram of the under control system

172

2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia

temperature should start from 25 degree of centigrade and the temperature goes up linearly up to 95 degree of centigrade during 4 hours. After that it must stay in the temperatures around the last temperature for 3 hours. Finally, the cooling stage must start if needed using water tap. But all things are not limited to above descriptions. For example in the first stage of process which related to four first hours, at the beginning the increasing rate of temperature must be slower because the raw material must harden gradually for getting higher resistance and after two hours the increasing rate must intensify because of that the raw material are harder than before. We don’t want to bring all details here, because aim of paper is not to describe the details of chemical and physical properties of terrazzo tiles. For all these information we use field data of factory with consulting with related specialists. The aim of this paper is to show how we use these data and manage them to design our controller. If we consider the described suitable steaming room, we can conclude usually in such processes if we emphasise on human knowledge descriptions, they have two stages: one stage which describe the total and main behaviour of system which is shown in Fig. 3. When we want to describe the system as an expert we describe the main behaviour of system normally with the minimum rules. In our problem this part of system is the area between upper and lower bands shown in Fig. 3. Two: after demonstrating the area using the preliminary description of system, we try to specify the optimum points in corrector axe between both upper and lower bands using secondary descriptions which related to details, one possible curve can be the preferred curve shown in Fig. 3. In order to implement this idea and in order to overcome uncertainty related to describing the words we proposed the system shown in Fig. 4. Fig. 4 shows the Block diagram of controller used to control an on-off valve with measuring temperature and control the duration time that valve is open or closed. If the sign of time is minus it means that for that time valve is closed otherwise is opened. For describing a primary rout we use just three rules as below (which we call them primary set of rules): H-Rule1-If temperature error is very Negative and temperature is (low or medium) then output duration time is high open H-Rule2-If temperature error is very Negative and temperature is high then output duration time is low open H-Rule3-If temperature error is Negative and temperature is low then output duration time is low open. H-Rule4-If temperature error is Negative and temperature is (medium or high) then output duration time is zero (closed or open) H-Rule5-If temperature error is Medium then output duration time is zero (closed or open) H-Rule6-If temperature error is Positive and temperature is (medium or high) then output duration time is zero (closed or open) H-Rule7-If temperature error is Positive and temperature is low then output duration time is low closed

in the condition which is unforeseen variation are high. Valve amortization is minimized. Meanwhile it is simple and inexpensive. The main part of this controller is a fuzzy system which controls the autoclave behavior. This fuzzy system is constructed using min inference engine, triangular fuzzifier, center of gravity (COG) defuzzifier. The controller is able to measure the temperature and set output flow of an on-off valve. Three stages are needed to implement such a system described above. First stage is wordifier _better to say a higher type of fuzzifier which is able to model a word which describes the quantities. Second and third stages include decision maker and defuzzifying stage which provides a probabilistic space in output to limit the decision space toward wanted response. After designing the total system, the system is simulated with measured data of input and approximate model of autoclave system. III.

PROPOSED FUZZY CONTROLLER

A. System definition and rule-base The method of our controlling behaviour here is under base of description and testing the system. Because of this at the beginning we don’t have the curve which shows us the behaviour of system which causes to reach the better quality. As primitive information about the suitable curve that we have to achieve good product quality, there is just descriptive information about the curve which we can illustrate it as shown in Fig. 3. It is necessary to know that this curve is not exact curve, it just show us normally first we have increase of temperature and after that we have a flat zone with minimum temperature changes and finally the temperature must reduce in a specific time. Counting on the practical and field data, the only thing which we have at the first glance is information which is describable with the curve shown in Fig. 3. In [1] says in steaming room the Upper band Preferred behavior Tem 95

Lower band

25 4 hour

7hour

12 hour

Figure 3: The approximate process curve

173

2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia

H-Rule8-If temperature error is very positivee and temperature is high then the output duration time is low closeed

property of the system under controll. V-Rule1: If temperature is very low then the output duration time is low and output band width iss narrow V-Rule2: If temperature is low then the output duration time is medium and output band bandwid dth is narrow V-Rule3: If temperature is med dium the output duration time is medium and output band width is narrow. V-Rule4: If temperature is high h the output duration time is medium and output band width is medium. m V-Rule5: If temperature is very high then the output duration time is very high and output band width is wide. B. membership functions .To implement all steps a controller is needed which contains: primary fuzzy block, corrector fuzzzy block (as shown in Fig. 4) and band width identifier block. A circuit c should also be designed to change the information obtained from sensors into the accepted range by controller. We first f divide the input range to three intervals. These intervals musst be clarified on the base of experienced and scientific knowledg ge of process under control. In the case of our problem and after th hat we assign output partitions which our output here is the duratio on time that valve is open or closed. Because of that the raw output of o controller is the duration of time. The control signal first is applied to a timer system and by each new reading of temperature the output duration time is substituted with new output duration n time. The controller system just in two states sends a command d directly to on-off vale: a- if the sign of recent duration time is opposite of the previous one

Figure 4: structure of proposed controller

H-Rule9-If temperature error is very positivee and temperature is (low or medium) then the output duration timee is high closed After that, we describe the deviation of prrimary rout with another set of rules which we call them correcctor rules. In this case first we consider the response of primarry rules (primary rules), after that, step by step by monitoring thee system behavior or by using detailed knowledge which prepaared earlier with present experts in site, and in each step we definne necessary rule. In this case we used five rules for deviations froom primary rout. Of course we considered three range of deviation due to the

174

2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia

and b- while the duration time is expired. IV.

SIMULATIONS AND DISCU USSION

As described above the on-off valve is set on the minimum effective flow and the injected energy is controolled by the time which valve is open or closed. For achieving this purpose the output membership functions are defined as beloow: Input membership functions for corrrector rules of temperature error are defined as Fig. 5(a) andd we select five singletons as output membership functions. Fig. 5(b), (c) and (d) show open or closed time duration of valve versuus seconds, Input temperature versus centigrade and the rate of chhanges applied to time duration in b, respectively. Considering that the temperature parameter aand its relation to the injection time duration is a corrector relatioon, Fig. 6 shows control surface of process. Fig. 6 (b) shows the final control surface which is complicated than Fig. 6(a) whicch shows primary control surface.

Figure 7: difference values between fin nal and primary control

Figure 8: points which final surface is under primary surface.

We chose one of curves of surfface shown in Fig. 6(b) which shows the time duration to temperatture error curve in the absence of temperature parameter (Input 2)). Using proper ANFIS, this curve extracted with 25 and 10 meembership functions for input and output respectively. While thee same curve here with very lesser numbers and with a descriptiv ve method. In Fig. 7 we see the differen nce values between final and primary control surfaces. Fig. 8 sh hows the points which final surface is under primary surface. Applying final control surfface to autoclave model approximation showed better resultss than primary control surface in point of view of number of swittching of valve and expected output rout. Also, the same comparrison has done with obtained practical data obtained from installed PID controller. Considering that the PID controllers lack facility y to describe unknown system, the results show our controller prov vides better protection of the

Figure 6: Control surface of process: (a) the primary control surface (b) the final control surface

175

2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia

material inside the autoclave room from unwanted temperature shocks. V.

CONCLUSION

In this paper we proposed a new method to handle uncertainty in fuzzy controllers by introducing primary and corrector rule bases. This method enables us to describe a complicated system with lesser rules and fuzzy words than previous used methods. On the other hand we showed the effective ability of this method in order to describe the system on the base of observations of system behavior. Another achievement is that the output surface of proposed fuzzifier is implementable with noticeable lesser number of membership functions and rules than traditional fuzzy controller type I and also the results in point of view of the complexity of calculations and simplicity of implementation and simplicity of system definition for practical usage is better than fuzzy type II. Finally, the paper confirms achieving better results of proposed method by doing simulation on autoclave room model in artificial stone maker factory. And providing more compatible control surface due to better tool for system behavior description which causes saving of energy. ACKNOWLEDGEMENT Authors would like to acknowledge support of Universiti Sains Malaysia and support of USM fellowship. REFERENCES [1]

Saeed Afrashteh, Artificial stones fundamentals, 2010, Akhavan publications, Tehran, Iran. [2] Yan-qing Peng Dept. of Autom., Univ. of Xiamen, Xiamen Jian Luo ; Jin-fa Zhuang ; Chang-qing Wu, “Model reference fuzzy adaptive PID control and its applications in typical industrial processes” Automation and Logistics, 2008. ICAL 2008. IEEE International Conference on, 2008. [3] Horacio J. Marquez, “Nonlinear Control Systems: Analysis and Design” Department of Electrical & Computer Engineering University of Alberta Edmonton, Alberta T6G 2V4, Canada, 2006. [4] Ondrej Linda, Milos Manic, “Uncertainty Modeling for interval type II fuzzy logic systems based on sensor characteristics” IEEE transaction on Fuzzy Systems, 2011. [5] M.A.A. Shoukat Choudhury, N.F. Thornhill, S.L. Shah, “Modelling valve stiction” Control Engineering Practice 13 (2005) 641–658, 2004 . [6] M.A.A. Shoukat Choudhury, S.L. Shah,_, N.F. Thornhill, David S. Shook, “Automatic detection and quantification of stiction in control valves” Control Engineering Practice 14 (2006) 1395–1412, 2005. [7] Guohua Wang Dept. of Ind. Eng. & Manage., Peking Univ., Beijing, China Jiandong Wang. “Quantification of valve stiction for control loop performance assessment” Industrial Engineering and Engineering Management, 2009. IE&EM '09. 16th International Conference on, IEEE resources. [8] Modeling of valve hysteresis, M. Antony, P. Query, Control Engineering Journal, 2010. [9] Youngbae Hwang, Jun-Sik Kim, In-So Kweon “Sensor noise modeling using the Skellam distribution: Application to the color edge detection” Mecatronics, Sciencedirect 2011. [10] J. M. Mendel “Uncertain Rule base Fuzzy Logic Systems: Introduction and new direction” Prentice Hall upper Sadel River, NJ 2001.

176