ADEX Configurator
Contents
1. status of the controller to Ready but it is always possible to return to configuration mode to make changes or even delete a domain 4 UM AC 01 EN ADEX Configurator Elements in the Configuration Window In order to change the default values of each controller configuration double click on the name of the controller The configuration screen will be displayed as shown in Figure 1 All the parameters will have default values EX Standard_C21 ADEX CONTROLLER CONFIGURATION INTERFACE CONFIGURATION DOMAIN 4AP C AP C PV CURRENT DOMAIN PROCESS VARIABLE 1 2x1 CONTROLLER STRUCTURE CONTROLLER STATUS Configuration EXPERT BLOCK gt UDL 1100 LDL lo DRIVER BLOCK e po p 1 23457 cp h AP MODEL ADAPTIVE MECHANISM FIGURE 1 CONFIGURATION WINDOW OF THE ADEX CONTROLLERS Note that any of the MISO controllers selected is by default PV1 and referred to as PROCESS VARIABLE 1 as shown in Figure 1 since it is a single controller The top left hand corner shows the CONFIGURATION DOMAIN and the PV CURRENT DOMAIN for the selected MISO controller The configuration domain box enables the user to select any of the other domains shown in the list when the box is clicked for configuration The ADEX controller blocks are as follows PROCESS EXPERT BLOCK DRIVER BLOCK CONTROL BLOCK and the AP MODEL ADAPTIVE MECHANISM The first three of these blocks display the parameters of the selected MISO contro2. call the response time of the loop a concept which is illustrated in Figure 12 Response time is defined as the time the process variable takes to enter a range from 5 of the value of the increment of PV around its stationary state value in response to a step in the OUT control signal It is assumed that the step in the value of OUT will begin from a stationary state of the process Stationary State 5 of the increment of PV for a step in the OUT process input Response Time FIGURE 12 User Manual 23 UM AC 01 EN ADEX Configurator If the control period is very short with respect to the response time the successive measurements of PV at the different control instants will contribute little information about the variation of PV and even worse may contain measurement noise Moreover the b parameters of the AP model tend to zero as the control period diminishes This makes the validity of prediction more sensitive than is desirable to identification errors and measurement noises Likewise the time delays of the different inputs of the MISO ADEX controller DP which are accounted in control periods may excessively increase lt would be advisable to maintain these time delays under or equal 6 control periods If the control period is significantly long with respect to the response time we cannot correct deviations of the PV produced during a control period until the next control instant arrives and this may take too long
3. determine the rate of change in the functioning of the corresponding MISO controller b If the user inserts O inside this box the rate of change will be determined by the logic executed in the PROCESSOR COL and be transferred via the pin corresponding to the operator If the Driver Block tab is clicked a small window as shown in Figure 8 will appear Default_cauda l 10 1 5 FIGURE 8 PARAMETERS OF THE DRIVER BLOCK This window of the driver block enables the user to set the following parameters PH Prediction Horizon This parameter defines the number of future steps control periods starting from the current control instant and projected along the desired output trajectory of the process The corresponding process control signal for the OUT of the selected MISO system is calculated in such a way that the predicted variable is equal to the desired variable at the end of the prediction horizon TC Time Constant This parameter determines the time constant in control periods of the default desired trajectory generated by the driver block from a second order model with a static gain and damping factor of 1 For example if the TC is equal to 1 5 control periods the desired process output will be delayed nearly 9 control periods required for a variable to reach the set point without overshooting 18 UM AC 01 EN ADEX Configurator ADEX Operator User Manual When the user clicks on the ADEX Operator tab see
4. of the ADEX controller in real time in relation to the selected MISO controller This window is shown in the top right hand of Figure 1 Control option tabs 1 Save Save the configuration of the Controller Help Open the Help screen for controller configuration 3 Done Return to the window with the controller list after saving the controller configuration N Controller Status User Manual In order to monitor the status of each controller once ready for operation and to show if there are any execution errors a configuration window is displayed with an indicator called CONTROLLER STATUS see Figure 18 The user can change the status of the controller by changing this value The next section explains the four possible values Configuration Under this status the internal functions of the controller are being defined configured and so it is not available for operation This is the initial status assigned as soon as the controller is created 7 UM AC 01 EN ADEX Configurator Ready This status indicates that the controller is already configured and ready for operation This status can be set by the user by clicking on the box beside Controller Status shown in the top left corner of Figure 18 whenever the controller configuration process is complete Operation Indicates that the controller is functioning as part of he control and optimization scheme This status is assigned automatically by ADEX
5. positive number the period of control is defined in terms of sample periods 2 If the user types in a zero the control period will be determined by the logic scheme and transferred via the corresponding operator pin in the logic scheme 3 If the user types in a negative number the control instant will occur when the ADEX controller detects a change in the signal sent via the operator pin mentioned above A control instant will take place if it is detected that there has not been a change following the set number of sampling periods since the last control instant In general it would be convenient to use shorter sampling periods in order to receive as much information as possible about the process This information can be used to obtain a convenient filtered process variable FPV from the process variable vector In a SISO process the selection of the control period must take into account the response time of the process for example approximately 95 of the time it takes for a process variable PV to stabilize after the application of a step change in control signal OUT A reasonable value for the control period various normally between 1 10 and 1 40 of the response time In this case since the process response time may be different for different MISO controllers the choice of a common control period needs to be a compromise between both response times Control Block User Manual When the user clicks on the Control Bloc
6. process becomes equal to 2 identification will generate a model like this one PV k 1 k A1 k PV k A2 k PV k 1 0 OUT k 0 OUT k 1 B3 k OUT k 2 2 That is B1 and B2 will have been made equal to O revealing that the actual time delay is equal to 2 Therefore if we want to retain at least one B which is not equal to zero it is advisable to choose DP and N for OUT in the following manner N 2 DP max DP min 1 Where DP min and DPmax are the minimum and maximum limits respectively of the range of possible variation of the actual time delay 25 UM AC 01 EN ADEX Configurator In the AP model the number of actual time delays of OUT over PV which in fact occurs will be defined by the equation DP efectivo DP min n mero de B iguales a O In the most unfavorable case in which the number of actual time delays of OUT over PV would be equal to DP max there would still be a B parameter to identify the process dynamics An analogous consideration is valid for the number of parameters related to other input signals within the AP model of a MISO ADEX controller Initial Values for the AP Model Parameters Let us consider a stable linear single input single output process described by the following equation PV k 1 Az PV k A PV k 1 B OUT k By OUT k 1 3 Assuming initial values equal to zero the process gain can be calculated from the permanent response PVperm to a step i
7. Also if the control period and sampling period are equal we may lose information which would be useful for control purposes since the variation of PV may be captured at intervals which will be too widely spaced ignoring what happens between those intervals For all of the reasons set forth above it is not advisable to choose a control period that is either very short or excessively long relative to response time In practice it is recommended a control period for a MISO ADEX controller between 1 5 and 1 40 of the response time The sampling period should be selected according to filtering criteria so that if strong filtering is suitable the sampling period should be short compared with the control period Within the context of an ADEX multivariable controller including more than one MISO ADEX controller a compromise must be made for the control period selected Noise level User Manual We advise setting the noise level NL as precisely as possible When in doubt it is more prudent to estimate NL on the broad side than on the narrow side However because the control action is moderated when PV is inside the noise level band if the latter is unnecessarily broad it can give rise to unacceptable deviations of PV with respect to the set point SP In general the process variable PV will tend to be in a band which relative to the setpoint will be lower than the noise level band Moreover we must take into account that when the set p
8. COP when the COS is in operation and in communication with the local area network Connecting This status is transitory and assigned automatically by ADEX COP when communication between the controller or the COS and the local area network has been lost Configuration of the AP Domains Process User Manual The parameters which are configured within the Process Block are common to all the MISO controllers and are established in the configuration of the MISO controller for PV1 which determines its value for the whole of the ADEX controller as explained in the following FPERT 4 por PROCESS lau i ES FIGURE 2 PROCESS BLOCK ST Sample Time which determines the frequency with which the ADEX controller receives information particularly the current values of the PV measurements from the output vector of the process the perturbation signal PERT the input vector Al and the vector of operating mode MODE The sample period is the execution time period of the logic schemes which execute the PROCESSOR cycle time and its value is displayed in seconds in this field 8 UM AC 01 EN ADEX Configurator CP Control Period determines the number of sample periods between two control actions generated by the ADEX controller when under automatic control The value can be established through the keyboard or via the logic of the control scheme The procedure is as follows 1 If the user types in a
9. ERT at the k 1 instant and at the previous instants DP DPC and DPD are integers which represent the delay in control periods in which a change in OUT1 OUT2 and PERT respectively produce a change in the value of PV1 The estimated PV1 k k 1 above can differ from PV1 k the measured value of PV1 at instant k with an estimation error of e k as given by k PV1 k PV1 k k 1 3 The values of the parameters Ai Bi Ci and Di at instant k 1 are changed at the instant k by the adaptive mechanism using functions of the form Ai k Ai k 1 a function of e k Bi k Bi k 1 a function of e k Ci k Ci k 1 a function of e k Di k Di k 1 a function of e k 4 12 UM AC 01 EN ADEX Configurator These functions are defined in such a way that e k tends rapidly to zero This adaptation only happens when a statistical criteria indicates the e k is due to a error in the model lack of precision in Ai and Bi and not caused by noise in the measurements or unknown perturbations For more details see references 5 The parameters related to the process variable PV1 shown in the window displayed in Figure 21 are as follows Initial lt Current User Manual NL Noise Level indicating the maximum variations which can be found in the measurement of PV1 while OUT1 OUT2 and PERT are constants and the process is in a steady state These variations can be caused by measurement noise which occurs in the PV and also b
10. Figure 18 the ADEX Operator is displayed as shown in Figure 9 FIGURE 9 DIAGRAM OF THE ADEX OPERATOR It can be seen in the diagram that the ADEX Operator shows I O values of the selected MISO controller in real time which are in this case PV1 SP1 Al1 PERT and OUT1 although the labels do not show the number 1 of the MISO If the MISO controller number 2 was selected the variables would have the number 2 attached PV2 SP2 Al2 and OUT2 In addition it can be seen that the ADEX operator displays two MODE variables The left hand MODE variable is the signal received from the COL while the right hand one is an internal MODE variable which the user can put into Auto by clicking with the mouse If this last option is taken the control mode continues External that is the control signal OUT follows the actual input signal Al generated by the COL but it puts into action a mode Internal AP Control The AP internal mode initiates the function of Adaptive Predictive Control i e the Driver Block the Control Block and the Adaptive Mechanism although the Adaptive Predictive Control signal is not applied yet to the process This mode of internal operation can be understood as a kind of training for the Adaptive Predictive Control before it is applied to the process This allows the user to observe the evolution of the parameters and variables of the controller such as for example see how the Prediction Error PE tends
11. MO can be understood as a group of MISO controllers containing the same number of controllers as variables in the process to be controlled Hence in the configuration process there is always a MISO controller selected for configuration within the ADEX controller An ADEX controller with one PV only one OUT only and with either one or more or no PERTS is a particular case within this general formulation ADEX COP assumes That the user will first configure an AP domain for each of the MISO controllers which will be called AP C C for Central domain i e between upper and lower domains The limits of these domains will by default be those of the corresponding ranges of the PV The user will be able subsequently to configure 1 or 2 AP additional domains for each MISO controller called AP U Upper and AP L Lower taking into account that establishing the limits should correspond to those of AP C For this purpose the user will change to the Configuration Domain to set the relevant limits for that domain The domains defined will always cover the corresponding ranges of the PV Once the desired domains are configured the user can continue to configure the Expert Domains For each of the MISO controllers an EX U Upper domain can be defined with a range above the PV and conversely an EX L Lower domain can be defined with a range below the PV Once the controller configuration is complete the user can change the
12. Ml ESE ES Ml Adaptive Predictive Expert Control ADEX Configurator User Manual April 2008 Edition UM AC 01 EN ADEX Configurator Contents About This Manual inicia did 3 Organization of This Manual cc cccccccccecssssscececeeecseseseeaeeeseeceesesesaeeeeeesesseeseaeeeeeeseseeeees 3 Conventions Used in This Manual ccccnnncncncncnnnnanananananananananananaran ono nono nono conocio nono n on rn cnn nnnnns 3 Related DOCUMENta OM 3 Part Configuration of ADEX Controllers Basie CONCEPUS zs eszet ar tette s dee tek A A o Ta 4 Elements in the Configuration Window ccccceesessceceeeesseseeaeeececeesceesseseeeeeeeseessasaeeeeess 5 Controller Status ui asa 7 Configuration of the AP DomMainNS ccccccccnnnonononnncnnnnnnnnnnonnnnnnnnnnonnnnnnnnnnnnnnnnnnnnnnnnnnnnnaninnnnos 8 TO 8 Control Blok iia A A 9 Adaptive Mecha cris 11 Expert Block acc dad tae cada 15 Driven Block ui rata taa taa e A a A 17 ADEX Operatoria kolega bia E a O ieee Ea leng AA aaia 19 Configuration of other AP DOMAINS ccccccononononnnnnnnnnnnnonnnnnnnoncnnnnnnnnnnnnnnnnnnnnnnonnnnnnnnnninnnns 20 Configuration of the Expert DomMainS cccccncooononcnnnnnnnononnnnnnnnnnnnnonnnnnnnnonnnnnonnonnnnnnnnnnannnos 20 Part II Choice of Structure Variables INTFOAUCION da AA A keen bad a A A AAA 23 Control amp Sampling Period is in 23 NOISE level A A EA A idas 24 Number of AP Model Parameters nusoni enid diiss 25 Initial Values
13. display a screen like that shown in Figure 11 In this case only three blocks to be configured will be active e Process Expert and Control 20 UM AC 01 EN User Manual ADEX Configurator Default_C21 ADEX CONTROLLER CONFIGURATION INTERFACE CONFIGURATION DOMAIN EX U AP C PV CURRENT DOMAIN VARIABLE 1 2a CONTROLLER STRUCTURE COWTROLLER STATUS Ready FIGURE 10 CONFIGURING EXPERT DOMAINS The value of the parameters shown in the Expert Block can be changed by the user in order to determine the upper UDL and lower LDL limits of the Expert Domain selected The user can change the values of the Sampling Time ST and the Control Period CP for the selected domain in the Process Block as has been previously explained Default_C11 EXPE m E4 INCOUT WT OUT 100 E E FIGURE 11 CONFIGURATION OF THE EXPERT DOMAIN If the Control Block is selected two configuration windows are displayed as shown in Figure 11 The window to the right shows the default values of the range and filter variables LV UV and FL in relation to the process variable PV and the ranges and incremental limits LL UL IL in relation to the output variable 21 UM AC 01 EN User Manual ADEX Configurator OUT and perturbations if relevant PERT The user can make changes to all the parameters of the selected domain These have all been described in previous sections The window to the left of Figure 28 shows the pa
14. e derived static gain will be positive To ensure this the value of S is used as a multiplier for the Bi parameters When the derived static gain is negative the parameters inside the AP model may be re initialized The user can disable this internal test by setting the value of S to 0 B1 a B6 Actuales The values of these parameters represent the current value of the adaptation generated by the adaptive mechanism for the corresponding parameters Bi of the AP model The user cannot change this unless the button Current Initial is clicked to reset the current values back to initial B1 to B6 BlaB6 Iniciales The values of these parameters are introduced by the user to provide initial values for the Bi parameters of the AP model before the adaptive mechanism comes into action The user can replace them at any time by clicking on the Initial Current button In Figure 4 shown below again for clarity all the parameters of the AP model are displayed 14 UM AC 01 EN ADEX Configurator Bt Default_Pres ADAPTIVE MECHANISM DPD a CMS CIN CI 1 2 3 4 5 6 PV1 k k 1 A1 k 1 PV1 k 1 A2 k 1 PV1 k 2 2 B1 k 1 OUT1 k 1 DP B2 k 1 OUT1 k 2 DP C1 k 1 OUT2 k 1 DPC C2 k 1 OUT2 k 2 DPC 4 D1 k 1 PERT k 1 DPD D2 k 1 PERT k 2 DPD These parameters correspond to the rows of the AP model shown above Row B refers to the parameters related to OUT1 row C to
15. e mechanism of ADEX controllers will find the appropriate value for the predictive model parameters by itself Nevertheless for reasons of common sense and to facilitate adaptation it is obviously a good idea to initialize the parameters of the predictive model in such a way that the model has a gain that is approximately equal to or greater than the process gain so that the first control signals are correctly moderated when the model is not yet adapted Further the rational choice of other structure variables such as the prediction horizon considered in the next section makes the ADEX controller performance robust and tolerant of errors in the identification of predictive model parameters As a general rule it should be borne in mind that if the control period tends toward zero the Bi parameters of the process will also tend toward zero Moreover when the control period tends toward to the response time the B parameter tends toward the gain G and the rest of the Ai and 27 UM AC 01 EN ADEX Configurator Bi parameters tend toward zero This dependency of the parameter values on the control period is inherent to the mathematical representation of the process and may be observed in parameters identification performed by the adaptive mechanism of ADEX controllers An analogous consideration is valid for the choice of the initial or default values of the parameters related to other inputs signals within the AP model a MISO ADEX control
16. f there is no Expert Domain defined to the limits set for the AP domains If the Expert Block is clicked a small window as shown in Figure 6 will appear Default_cauda a fo FIGURE 6 EXPERT BLOCK PARAMETERS This Expert Block window enables the user to configure the following parameters TR Time of Residence The value introduced in this box determines the number of control periods which the AP model needs to acquire the minimum data I O values sufficient to allow the calculation of a control signal The default value is 1 and in this case the system itself calculates the minimum number of control periods necessary to fill out all the variables of the AP model with significant data before carrying out the calculations If the value assigned to TR is less than the minimum number the AP model will start to carry out calculations 16 UM AC 01 EN Driver Block ADEX Configurator assuming that previous variable values of the AP model are the same as the most recent ones where there is no information available EM Entry Mode The value of this parameter determines which values will be taken by the current parameters of the AP model when the output value of the selected MISO process passes from a particular domain to the selected AP configuration domain There are 4 different options which correspond to the following EM values O Leaving the value at zero default value the Expert Block reinitializes the cur
17. for the AP Model ParameterS oooooconcccccnoccnonnncncnnnnnnnnnnononononononnonononinononoss 26 Prediction Hor ZO cm aia 28 ee AA E E AEE E ida 30 User Manual 2 UM AC 01 EN ADEX Configurator About This Manual This manual describes the tool that enables the user to configure ADEX Controllers and the ADEX Configurator both of which are used in all ADEX systems This manual also presents the methodology for the configuration of the structural variables which determine the dynamic behaviour of the controllers Organization of This Manual This manual is divided into two parts Part Configuration of ADEX Controllers This section explains the ADEX Configurator and the necessary concepts for the configuration of ADEX Controllers Part II Choice of Structure Variables This section explains how to choose the most important structural variables in an ADEX Controller Conventions Used in This Manual d This icon denotes a note which alerts you to important information A This icon denotes a warning which should not be overlooked if the system is to continue to function correctly Related Documentation The following documents contain information you might find helpful as you read this manual ADEX Methodology User Manual 3 UM AC 01 EN Part ADEX Configurator Configuration of ADEX Controllers Basic Concepts User Manual It should be remembered that the structure of the multivariable controllers MI
18. k tab shown in Figure 1 the parameter configuration window shown in Figure 3 appears On the left hand side the window shows a column of variables which form part of the Control Block for the selected MISO controller In this case the variables from top to bottom are PV OUT and PERT where PV and OUT are the corresponding variables of the process and output of the controller MISO and PERT is an input variable which has an influence on the evolution 9 UM AC 01 EN User Manual ADEX Configurator of the process variable Each one of these variables has associated parameters which can be configured by the user in this window EN Default_caudal CONTROL BLOCK FIGURE 3 PARAMETERS DEFINED IN THE CONTROL BLOCK The parameters related to the variable PV of the process are LV Lower value which is the bottom of the range of values measurable by the process UV Upper Value which is the top of the range of values FL Filtered constant The filtered value of the process variable at the sampling instant k FPV k is obtained by measuring the value PV k using FL as follows FPV k FL PV k 1 FL FPV k 1 1 If PV k is contaminated with measurement noise the use of this first order filter 1 produces a better estimate of the process variable FPV k The parameters related to the output OUT of the controller MISO are Los par metros relacionados con la salida OUT del controlador MISO son LL Lower limit for
19. k will have an effect on the process variable PV and as a consequence the application of predictive control will not be possible 28 UM AC 01 EN ADEX Configurator An analogous consideration must also be made for the choice of PH in relation with other input signals with variable time delays within the AP model of a MISO ADEX controller User Manual 29 UM AC 01 EN ADEX Configurator References User Manual 1 J M Mart n S nchez Adaptive Predictive Expert Control System International Patent Application Application N2 PCT USOO 17836 Filling Date June 28 2000 2 J M Mart n S nchez Adaptive Predictive Control System USA Patent N2 4 197 576 1976 3 J M Mart n S nchez Adaptive Predictive Control System CIP European Patent N2 0037579 1980 4 J M Martin Sanchez y J Rodellar Adaptive Predictive Control From de Concepts to Plant Optimization Prentice Hall Internacional 1996 30 UM AC 01 EN
20. ler Prediction Horizon User Manual When the Prediction Horizon PH is small the process variable PV is constrained to follow closely the desired trajectory without taking into account the actual dynamics of the process On the other hand when PH is bigger the process variable can reach the desired values with more flexibility taking into account the actual process dynamics Thus the selection of PH depends on the rigor with which we want the process variable PV to follow step by step a given path or desired trajectory If PH is small or equal to 1 its minimum value the possible measurement noise in PV k can introduce high frequency contents in the control signal OUT Also it is important to consider that the prediction horizon must compensate for the variations of pure time delays when they occur As considered previously in Section 4 for a process time delay which varies between DPmin and DPmax the structure variable DP will be set equal to DPmin and the difference with the actual time delay of the process will be absorbed by a certain number of parameters B which become equal to zero In this case the PH must choose PH gt DPmax DP min 1 8 This is due to the fact that when the actual process time delay becomes equal to DPmax if inequality 8 is not verified the prediction instant k DP PH will be previous to the instant k DP max 1 that is the first instant in which the control action OUT applied at instant
21. ller which are configurable by the user User Manual 5 UM AC 01 EN ADEX Configurator PROCESS Two defaults values are presented ST Sample Time CP Control Period for the AP C domain as shown in Figure 1 These parameters are common to all the MISO controllers and the ADEX controller will always use the values set for PV1 MISO 1 EXPERT BLOCK Two default values are presented UDL Upper limit of the expert domain LDL Lower limit of the expert domain DRIVER BLOCK One default value is presented RC Rate of Change This is the maximum rate at which the variable should change in time When the EXPERT BLOCK DRIVER BLOCK CONTROL BLOCK and the AP MODEL ADAPTIVE MECHANISM are double clicked configuration windows are displayed for each one As is explained in the following sections the user can determine the desired function for each of the blocks by replacing the default values As can be seen the ADEX blocks are represented by option tabs which are permanently active when they are relevant to the domain of the selected configuration or otherwise they remain inactive and are dimmed The principal real time variables of each MISO controller are shown in smaller windows in Figure 18 and are outlined below 1 The MODE variable which together with the variable PV of the current domain determines basically the type of operation of the Expert Block for the corresponding MISO system 2 The MODE variable
22. model The user cannot change them unless the button Current Initial is clicked to revert the current Al A2 and A3 parameter settings to their initial values 13 UM AC 01 EN User Manual ADEX Configurator A1 A2 A3 Iniciales The values of these parameters are those introduced at the very beginning by the user to initialize the AP model prior to putting the adaptive mechanism into action The user can replace the initial values at any time by replacing them with the current values clicking on Initial Current The parameters related to the output OUT1 of the selected MISO controller shown in Figure 4 are as follows DP Delay Periods A change in OUT1 will require a certain number of control periods before a change will start to develop in PV1 This number of control periods minus 1 is what is termed a delay period N A AG The values of these parameters are important for OUT1 and related to the parameters Bi in the AP model equivalent to what has already been described for PV1 S Sign of the static gain of the process which will be made equal to 1 if the response of PV1 is positive and if an increment in the output from the controller OUT1 is also positive Otherwise the S will be made equal to 1 The value of S is used internally to diagnose an undesirable result in the current functioning of the Adaptive Mechanism When the current values of AP model parameters are working satisfactorily th
23. n the control signal OUT step by means of G PV perm OUT step Since eguation 3 will also be verified when the steady state is reached by the process after the application of said step in the control signal it may be written PVperm A1 PVperm A2 PVperm B1 OUTstep B2 OUTstep B3 OUTstep 4 Therefore by dividing both members of 4 by OUT step G A G A gt G B1 B2 B3 5 G B 1 2A 6 User Manual 26 UM AC 01 EN User Manual ADEX Configurator The preceding equation can be useful to us for establishing some initial or default values for the AP model parameters which would comply themselves with a experimental estimation of the process gain As an example let us assume an AP model described by the following equation PV k 1 k A1 k PV k A2 k PV k 1 B1 k OUT k B2 k OUT k 1 7 If the gain of this process is equal to 1 we could take the following as initial or default values for the parameters of the predictive model A1 0 1 B1 0 0 1 A2 0 0 2 B2 0 0 1 In fact these are typical initial or default values for the predictive model of ADEX controllers We can see that equation 6 which relates gain with the model parameters is fulfilled 0 1 0 1 1 1 0 2 If the gain were different from 1 it would be sufficient to multiply all the B parameters by the value of the gain to obtain a predictive model with the appropriate gain In general the adaptiv
24. oint changes the noise band will center itself around the new SP If this change is not greater 24 UM AC 01 EN ADEX Configurator than the width of the noise level band the process variable can remain inside it and as a result the control action will be moderated The result may be that the PV does not converge towards the new SP Number of AP Model Parameters User Manual If the time delays of the MISO ADEX controller inputs are fixed the selection of number of parameters N for the different input output AP model signals is simple in most cases it will be made equal to 2 However if the time delay is variable for the control signal a number of parameters for this signal will be chosen which will allow the adaptive mechanism to adjust the corresponding b parameters to follow the variations of the time delay To illustrate this procedure let us consider a simple single input single output process in which the time delay of PV with respect to OUT can vary between 0 and 2 In this case we could choose the following predictive model PV k 1 k Aa k PV k A2 k PV k 1 B1 k OUT k B2 k OUT k 1 B3 k OUT k 2 1 Where N 2 for PV N 3 for OUT and DP 0 If the actual time delay of the process without considering the inherent time delay of discrete processes is equal to zero the identification mechanism will give the parameter B k a value other than zero Nonetheless if the actual time delay of the
25. rameters of Expert Control A button is displayed which changes the number of parameters used by the expert block In a reduced version only three are used and these are explained as follows in the context of the Expert Block OUT This is the value of the control signal which the ADEX controller under Automatic mode generates first while the selected PV is in the Expert Domain under consideration INCOUT This value will be added to the above control signal periodically if PV1 does not leave the expert domain WT Waiting time This is the value of the time period in seconds which the system will wait prior to carrying out the incremental action in the control signal mentioned above in INCOUT 22 UM AC 01 EN ADEX Configurator Part Il Choice of Structure Variables Introduction As described before ADEX methodology is applied to multivariable processes by decomposing internally the ADEX multivariable controllers into a set of n multi input single output MISO ADEX controllers This section explains how to select the most significant structure variables in a MISO ADEX controller Thus the selection criteria presented here are valid for single input single output and multivariable controllers Control amp Sampling Period The selection of the control period for a MISO ADEX controller should be done if possible in the light of the dynamics of the process variable PV itself and with particular attention to what we
26. rent parameters of the AP model in accordance with the initial values assigned for the selected domain 1 The Expert block will reinitialize the values of the current AP model parameters in accordance with the latest current values of the selected domain 2 If the PV comes from an AP domain the current parameters of the AP model maintain the latest current values from that domain If the PV comes from an Expert Domain the values of the current parameters reset themselves to their initial values automatically 3 If the PV comes from an AP domain the current parameters maintain the latest current values obtained in that AP domain in the same way as described in point 2 If the PV comes from an Expert Domain the values of the parameters also maintain the latest values in the selected domain The Driver Block has a small box inside labeled RC User Manual y SP DPY DRIVER BLOCK RC E FIGURE 7 DRIVER BLOCK 17 UM AC 01 EN User Manual ADEX Configurator RC Rate of Change This parameter limits the rate of change of the trajectory which drives the process output towards the set point The value of the RC is introduced in the form of engineering units per control period The rate of change can be established internally or externally via the control scheme logic The procedure is the following a If the user establishes a positive increment in the engineering units using this tab this increment will
27. the controller output UL Upper limit for the controller output IL Incremental limit for the controller output 10 UM AC 01 EN ADEX Configurator LIL Lower incremental limit for the controller output This reduced control limit is applied when the process is in steady state close to the set point It is important to note that the user must pay special attention to define a variation range for the actual range of the controller when introducing these limits that is a range of variation in which any change in the value of OUTn has a particular and unique effect on the process variable PVn of the process The parameters related to the perturbations are the same as those related to PV except the LIL which does not exist and therefore is substituted by SI Significant increment This is the absolute value of the increment in PERT which if it occurs relieves the controller of the limit LIL although the PV continues close to the SP This allows the controller to respond with a greater control action and thereby reduce the impact that PERT would otherwise have on PV If the MISO controller corresponding to the other PV is selected and the Control Block button clicked the same window will appear but this time with the values of this MISO controller A Warning It is important for the user to make an appropriate selection of SI for OUT2 in such a way that under this absolute value the incremental changes to OUT2 can al
28. those of OUT2 and row Dto those of PERT The row letters are added for illustration The columns are effectively numbered 1 6 shown here for illustration which means for example that parameter C5 is displayed in column 5 of OUT2 and parameter D1 is displayed in column 1 of PERT The delays DP are effectively DPC for OUT2 and DPD for PERT The row B delay period is referred to simply as DP AM Adaptive Mechanism This parameter adjusts the speed of adaptation of the AP model between 0 no adaptation takes place to 1 maximum speed of adaptation with intermediate speeds between Finally it is important to note that if the MISO controller corresponding to PV2 is selected the window shown in Figure 21 will appear the same but with all the parameters corresponding to PV2 Expert Block The variables displayed in the Expert Block are shown in Figure 5 and are as follows User Manual 15 UM AC 01 EN User Manual ADEX Configurator MODE EXPERT BLOCK EXT mi UDL 100 as LDL lo FPERT A 0 FIGURE 5 EXPERT BLOCK UDL Upper Domain Limit This defines the value of the upper limit of the process variable The system uses limits to define appropriate responses to the value of the process variable PV in terms of a corresponding OUT LDL Lower Domain Limit This defines the value of the lower limit of the process variable Both the lower and upper limits will be wider if an Expert Domain is defined or equal i
29. to zero and how the AP model converges on specific values In this way the user can easily adjust the system before the application of the AP control signal and ensure it is working appropriately and correctly Similarly while the controller is under Auto control the operator will be able to activate an Internal Manual MAN mode from the ADEX Operator and an Internal Set point INT SP to control the selected MISO 19 UM AC 01 EN ADEX Configurator controller In the first instance the operator can determine the output OUT of the controller by changing the corresponding field OUT of the ADEX controller from the keyboard In this mode the value of OUT will remain a constant if the operator does not change its value Under the Internal Set point mode the operator can force the value of the Internal Set point INT SP field which will become the new controller set point Configuration of other AP Domains Once the configuration of the AP C domains have been completed the user can configure other AP domains To do this click in the box associated with the configuration domain indicated in Figure 2 shown in the next section The user can configure a new domain such as upper AP U or lower AP L Configuration of the Expert Domains User Manual The user can configure an Expert Domain while in the window as shown in Figure 10 If the user clicks on the selected domain for example EX U the system will
30. ways be compensated by incremental changes in OUT1 under the value of LIL Adaptive Mechanism AP MODEL ADAPTIVE MECHANISM User Manual When the user clicks on the Adaptive Mechanism button shown in Figure 1 the parameter configuration window shown in Figure 4 appears if the MISO controller selected corresponds to that of PV1 The detailed information relating to the variables whose names are shown to the far left is displayed in each line of parameters and these variables can be selected by means of a pop up menu as shown in Figure 4 11 UM AC 01 EN User Manual ADEX Configurator Bt Default_Pres ADAPTI E MECHANISM XI our a CSE CON CEN COST y PERT ebhe fo fon bo b b b b foa Mos f b p fp FIGURE 4 ADAPTIVE MECHANISM PARAMETERS These variables are considered by the Adaptive Predictive model and are used by the MISO controller in the adaptive mechanism defined as follows PV1 k k 1 A1 k 1 PV1 k 1 A2 k 1 PV1 k 2 2 B1 k 1 OUT1 k 1 DP B2 k 1 OUT1 k 2 DP 4 C1 k 1 OUT2 k 1 DPC C2 k 1 OUT2 k 2 DPC 4 D1 k 1 PERT k 1 DPD D2 k 1 PERT k 2 DPD PV1 k k 1 represents the estimated value of PV1 in the control instant k obtained from the previous available instant k 1 As a consequence this estimated PV1 is obtained from the value of the adaptive predictive model parameters AP Ai Bi Ci and Di and the measured values of PV1 OUT1 OUT2 and P
31. which together with the variable PV of the current domain determines basically the type of operation of the Expert Block for the corresponding MISO system 3 The set point SP which is input to the Driver Block and the desired process variable DPV which is output from the Driver Block and input to the Control Block 4 The output from the Control Block OUT to be applied to the Process the perturbation which acts on the process FPERT and the Filtered Process Variable FPV output from the process User Manual 6 UM AC 01 EN ADEX Operator ADEX Configurator 5 The Predicted Process Variable PPV which is output from the AP MODEL and 6 the Prediction Error PE are both shown when the operation is functional The lines showing the interrelationships between the ADEX blocks These lines appear black when the interrelation is active between the blocks for the selected MISO system Otherwise they are dimmed As shown in Figure 1 the MODE variable is EXT or External Only the lines in the upper part of the screen are shown in black and therefore active In this case the Expert Block determines that the Driver Block and the AP Model adaptive mechanism remain inactive It also has determined the functioning of the Control Block which produces an OUT equal to the Al There is an option ADEX Operator shown in Figure 1 which when double clicked displays an Operator scheme showing the values of the Operator I O variables
32. y the dynamic effect of other variables not taken into account in the AP model N Appears in the row of parameters associated with the PV and also in the row of a parameters associated with OUT In the first case it is the number of parameters Ai taken into consideration in the AP model while in the second it is the number of Bi parameters The remainder of the AP model parameters shown but not considered are effectively ignored and treated as though their values were zero A This parameter is also presented in the rows of PV and OUT parameters The value can be 0 disabled or 1 enabled and refers to the enabling of the adaptation mechanism for the Ai and Bi parameters respectively AG Adaptation Gain As in the case of N and A this parameter appears in the PV1 and the OUT1 rows Sometimes it is possible to improve the function of the adaptive mechanism by changing the internal range of the variation of some I O variables in the AP model simply for reasons of adaptation The internal range of variation is defined by default as a percentage above the range of variation in the variable defined in engineering units The parameter AG will change the internal range of variation in PVI or OUT1 multiplying the default by AG to obtain the adjusted range value A1 A2 A3 Actuales The values of these parameters represents the current value of adaptation generated by the adaptive mechanism for the corresponding Ai parameters of the AP
Download Pdf Manuals
Related Search