NAMP PIECE Tier II Worked Examples Module 5 – Controllability Analysis 1 NAMP PIECE Tier II Statement of intent The goal of this tier is to demonstrate various concepts and tools of Controllability Analysis using real examples. Some examples will be given, focusing mainly on Controllability Analysis tools. At the end of Tier II, the student should have a general idea of what is: Relative Gain Array Niederlinski Index Controller design Design of multivariable controllers Steady State Decoupling (Singular Value Decomposition) Module 5 – Controllability Analysis 2 NAMP PIECE 2.0 Inverse of a matrix To obtain the Relative Gain Array (RGA) of a transfer function matrix, first of all it is necessary to review how to obtain the inverse of a matrix. The inverse of a matrix does not exist for all matrices, it exists only if: The matrix is square, and Its determinant is not zero (non-singular matrix). Now given the 3x3 matrix A, it is desired to obtain the inverse of the matrix A. Module 5 – Controllability Analysis 3 NAMP PIECE 1. Determinant of matrix A. The determinant of matrix A is obtained using the cofactors based on any row. In this case the first row is selected. 1 4 0 4 0 1 3 A 2 1 5 2 2 3 5 3 2 1 3 A 0 1 4 5 2 3 2(3 8) 1(0 20) 3(0 5) 5 As it can be seen, the determinant is not equal to zero Module 5 – Controllability Analysis 4 NAMP PIECE In a 3x3 matrix there is no problem to calculate the determinant. Nevertheless it is important to note that the second cofactor is being multiplied by the factor of (-1). This is because the cofactor of each matrix element must be multiplied by the following term: (1)i j where i is the row number and j is the column number of the element for which the cofactor is calculated. It means that for matrix A, the cofactors which are multiplied by (-1) are: C(12), C(21), C(23) and C(32) 2 1 3 A 0 1 4 5 2 3 Module 5 – Controllability Analysis 5 NAMP PIECE 2. Matrix of cofactors (C). Now it is necessary to calculate the cofactors matrix (C) for each element of matrix A. C11 1 4 2 3 5 0 4 C12 5 3 20 1 3 C21 3 2 3 2 3 C22 9 5 3 1 3 C31 1 1 4 2 3 C32 0 4 8 2 1 3 A 0 1 4 0 1 5 2 3 C13 5 5 2 2 1 C23 1 5 2 2 1 C33 2 0 1 And this way the cofactor matrix has been obtained: Module 5 – Controllability Analysis 5 20 5 C 3 9 1 1 8 2 6 NAMP PIECE 3. Adjoint of (A). The Adjoint of A is the transpose of matrix C. 5 20 5 C 3 9 1 1 8 2 1 5 3 adj( A) CT 20 9 8 5 1 2 4. Now the inverse of A is obtained using the determinant and the adjoint of A. 1 3 5 3 1 5 1 CT 1 1 Α adj( A) 20 9 8 4 9 5 Α Α 5 5 1 2 1 1 5 1 5 8 5 2 5 5. Check now that by matrix multiplication that the identity matrix is obtained multiplying A and A-1. 3 2 1 3 1 5 A A 1 0 1 4 4 9 5 5 2 3 1 1 5 Module 5 – Controllability Analysis 1 1 0 0 5 8 0 1 0 5 2 0 0 1 5 7 NAMP PIECE As can be seen, the amount of work is extensive, just to calculate the inverse of a matrix!. Therefore, it will now be shown how to obtain the same inverse matrix A-1 of matrix A now using Excel. 1. Determinant of matrix A, using Excel. Fill the numbers of the matrix in an excel sheet, using a cell for each element of the matrix. Module 5 – Controllability Analysis 8 NAMP PIECE 2. Calculate the matrix determinant, this will be done using cofactors of the first row. Each cofactor must be calculated using the appropriate formula as shown in the Excel formula bar. Pay attention on cofactor C(12) because it must be negative, as it has been shown previously. - + Elements C(11), C(12) and C(13) are the cofactors of each element of the first row, which are elements a11, a12 and a13 according to matrix A: a11 a12 a13 A a21 a31 a22 a32 a23 a33 Module 5 – Controllability Analysis 9 NAMP PIECE 3. Calculate the matrix determinant, using the formula shown on the excel sheet. Remember that the cofactor of element a12 must be negative! The determinant of this matrix is not equal to zero. For this reason it is possible to obtain its inverse matrix. Module 5 – Controllability Analysis 10 NAMP PIECE 4. Calculate the cofactors of matrix A for the second and third rows. Now remember, the cofactors for the elements C(21), C(23), and C(32), are: (select one) POSITIVE NEGATIVE As you can see in the Excel sheet, it is necessary to include the sign for each cofactor, and it arithmetically appears on the cell that calculates the matrix determinant (red one). Do not confuse this sign with the one that has been written in front of the matrix in the excel sheet (blue one). The latter one is just to show the sign of the cofactor. Module 5 – Controllability Analysis 11 NAMP PIECE 5. Create the cofactors matrix C placing each cofactor in the corresponding place of the element of which it has been calculated, as the excel formula bar shows. Do the same with the second and third rows of matrix C. Module 5 – Controllability Analysis 12 NAMP PIECE 6. Calculate the adjoint of matrix A, this will be done just transposing the matrix C (matrix of cofactors). Module 5 – Controllability Analysis 13 NAMP PIECE 6. Calculate the inverse of matrix A (A-1). To do this, divide the Adjoint matrix by the determinant of matrix A, already calculated. And do the same for the remaining elements of matrix A-1 Module 5 – Controllability Analysis 14 NAMP PIECE 7. Check by matrix multiplication that matrix A multiplied by A-1 gives the Identity Matrix. And do the same for the rest of elements of matrix A-1 Module 5 – Controllability Analysis 15 NAMP PIECE As can be seen, even using excel to obtain the inverse of a matrix is still hard work. But there exist some functions in Excel that allow you to obtain the inverse of a matrix rapidly. This will be shown next. Module 5 – Controllability Analysis 16 NAMP PIECE Now it will be shown how to use some Excel functions to manipulate matrices. 1. Determinant of matrix A, using an Excel functions. Fill the numbers, as before, of the matrix in an Excel sheet, using a cell for each element of the matrix. Select the cell where the determinant is to appear. Select function/insert and choose the function mdeterm. Module 5 – Controllability Analysis 17 NAMP PIECE 2. Next, it is necessary to select the range of the matrix data. This way, the determinant of A is easily calculated. Module 5 – Controllability Analysis 18 NAMP PIECE 3. Now the transpose matrix of A will be calculated. Select the cell where the first element a11T is to appear. Select function/insert and choose the function transpose. Next, select the range of the matrix, as the next slide shows. Module 5 – Controllability Analysis 19 NAMP PIECE 4. Select the range of the matrix, as shown below. Once the range of the matrix has been selected press OK. Module 5 – Controllability Analysis 20 NAMP PIECE 5. No values appear in the cell because it is necessary to introduce the formula as an array. To do that, select the range that the transpose matrix will occupy and press F2. Then, press shift+ctrl+enter. The Excel sheet should look like the one below. Module 5 – Controllability Analysis 21 NAMP PIECE 6. It is possible to obtain directly the inverse of a matrix. Select the cell where the first element a11-1 is to appear. Select function/insert and choose the function minverse. Then, select the range of the matrix, as shown on the next slide. Module 5 – Controllability Analysis 22 NAMP PIECE 7. Select the range of the matrix, as shown below. Once the range of the matrix has been selected press OK. Module 5 – Controllability Analysis 23 NAMP PIECE 8. Again no values appear in the cell because it is necessary to introduce the formula as an array. To do that, select the range that the inverse matrix will occupy and press F2. Next, press shift+ctrl+enter. The Excel sheet must look like the one below. Module 5 – Controllability Analysis 24 NAMP PIECE The determinant, transpose and inverse of matrix A can easily be obtained. To verify if the Identity matrix is obtained, the function MMULT of excel is used. Matrix A can be multiplied by Matrix A-1. 9. Multiplying matrices. Select the cell where the element I11 is to appear. Select function/insert and select the function mmult. Module 5 – Controllability Analysis 25 NAMP PIECE 10. Next, select the range of each matrix to be multiplied, as shown below. Once the ranges of the matrices have been selected, press OK. Module 5 – Controllability Analysis 26 NAMP PIECE 11. No values appear in the cell because it is necessary to introduce the formula as an array. To do that, select the range that the identity matrix will occupy and press F2. Next, press shift+ctrl+enter. The Excel sheet should look like the one below. Module 5 – Controllability Analysis 27 NAMP PIECE As it was seen, Excel functions are very helpful to obtain the determinant, the transpose and the inverse of a matrix, even for matrix multiplication. Of course there are several software packages with the availability to work with matrices, but Excel has been chosen because it is available in almost every computer that students have access. These Excel functions are the main tools that will serve to obtain the Relative Gain Array, as it will be shown shortly. Few examples will be covered. Module 5 – Controllability Analysis 28 NAMP PIECE Despite the great help that Excel can provide, some limitations must be specified. These limitations are: Determinant. The size of the array must not exceed 73 rows by 73 columns. Multiplication. The size of the resulting array must not be equal or greater than a total of 5 461 cells. Inverse. The size of the array must not exceed 52 columns by 52 rows Module 5 – Controllability Analysis 29 NAMP PIECE 2.1 Relative Gain Array 2.1.1 Obtain the RGA for the linear model of a distillation column used in separating methanol and water as reported in [1] (see next slide). It is a system with two output variables, two input variables, and one disturbance variable. All variables are defined in terms on deviation variables: y1= overhead mole fraction methanol y2= bottoms mole fraction methanol d = column feed flowrate u1= overhead reflux flowrate u2= bottoms steam flowrate The 2x2 transfer function matrix is: 12.8e s 1 G ( s ) 16.7 s 7s 6.6e 10.9s 1 18.9e 3s 21.0s 1 19.4e 3s 14.4s 1 Module 5 – Controllability Analysis 3.8e 8.1s 14 . 9 s 1 G d ( s) 3.4 s 4.9e 13.2s 1 30 NAMP PIECE Distillation column used in separating methanol and water Overhead reflux flow rate (u1) Feed flow rate (d) Overhead mole fraction methanol (y1) Bottoms steam flow rate (u2) Bottoms mole fraction methanol (y2) It is easy to identify both manipulated and controlled variables. Module 5 – Controllability Analysis 31 NAMP PIECE Problem description. From the transfer function matrix G(s), it is possible to obtain the steady-state gain matrix. The steady-state gain matrix is: 12.8 18.9 K G(0) 6 . 6 19 . 4 Based on this matrix, the Excel functions seen previously can be used to obtain the RGA as the next slide shows. Module 5 – Controllability Analysis 32 NAMP PIECE It is possible to use any Excel sheet shown before, or you can start a new one as the next figure shows. R= Matrix K Matrix R Be careful, because the multiplication of matrices K and R is a multiplication term by term. Module 5 – Controllability Analysis 33 NAMP PIECE The RGA has therefore been obtained very easily: 2.01 1.0 1.0 2.01 The pairing rules recommend pairing 1-1/2-2, which means to use the overhead reflux flowrate to control the overhead mole fraction of methanol, and the bottoms steam flowrate to control the bottoms mole fraction of methanol. The final coupling is shown in the next slide. Module 5 – Controllability Analysis 34 NAMP PIECE Final coupling suggested for RGA from a distillation column used to separate methanol and water LC CC Feed flow rate (d) Overhead reflux flow rate (u1) Overhead mole fraction methanol (y1) Bottoms steam flow rate (u2) LC CC Bottoms mole fraction methanol (y2) Next, the RGA for a 3x3 system will be calculated. Module 5 – Controllability Analysis 35 NAMP PIECE 2.1.2 Obtain the RGA for pilot scale binary distillation column used to separate ethanol and water for which the transfer function matrix is given below [2]. The process variables are (in terms of deviations from their respective steady state values): y1 = overhead mole fraction ethanol u1 = overhead reflux flowrate y 2 = side stream ethanol mole fraction u2 = side stream draw-off rate y3 = Temperature on Tray #19 u3 = reboiler steam pressure 0.66e-2.6s 6.7s+1 y 1 -6.5s y = 1.11e 2 3.5s+1 y 3 -33.68e-9.2s 8.15s+1 -0.61e-3.5s 8.64s+1 -2.3e-3s 5s+1 46.2e-9.4s 10.9s+1 Module 5 – Controllability Analysis -0.0049e-s 9.06s+1 u 1 -0.012e-1.2s u2 7.09s+1 u 0.87 11.61s+1 e-s 3 3.89s+118.8s+ 36 NAMP PIECE Distillation column used in separating ethanol and water Temperature on tray #19 (y3) Overhead reflux flow rate (u1) Feed flow rate (d) Overhead mole fraction ethanol (y1) Side stream draw-off rate (u2) Mole fraction of ethanol in the side stream (y2) Reboiler steam pressure (u3) Module 5 – Controllability Analysis 37 NAMP PIECE Problem description. From the transfer function model, the steady-state gain matrix is: 0.66 -0.61 -0.0049 K = G(0) = 1.11 -2.3 -0.012 -33.68 46.2 0.87 R= Matrix K Module 5 – Controllability Analysis Matrix R 38 NAMP PIECE Again, remember that the multiplication of matrices K and R is a multiplication term by term. The RGA for the 3x3 system has been obtained easily: 1.95 -0.67 -0.27 = -0.66 1.90 -0.23 -0.28 -0.23 1.51 The pairing rules recommend pairing 1-1/2-2/3-3, which means that the overhead mole fraction of ethanol can be controlled using the overhead reflux flowrate. In the same way, the mole fraction of ethanol in the side stream can be controlled using the side stream draw-off rate and the temperature on tray #19 can be controlled using the reboiler steam pressure. The next slide shows the final coupling. Module 5 – Controllability Analysis 39 NAMP PIECE Final coupling suggested for RGA from a distillation column used in separating ethanol and water Overhead mole fraction ethanol (y1) Temperature on tray #19 (y3) CC Overhead reflux flow rate (u1) Mole fraction of ethanol in the side stream (y2) TC CC Feed flow rate (d) Side stream draw-off rate (u2) Reboiler steam pressure (u3) Module 5 – Controllability Analysis 40 NAMP PIECE 2.2 Niederlinski Index Up to now the utility of the RGA has been used to find the appropriate pairing for the process variables. Despite the utility of the RGA, sometimes it is necessary to use the RGA with another important tool such as the NIEDERLINSKI INDEX (NI). The NI is very useful because it allows to identify structurally unstable pairings and as a result to avoid them. Next the use of NI is shown. Module 5 – Controllability Analysis 41 NAMP PIECE 2.2.1 Determine the best pairing using the RGA and the NI for the system with three output variables and three input variables, the steady state matrix is: 1 0.1 1 K 2 3 1 0.1 2 1 u1 u2 u3 y1 y2 y3 Problem description. First of all, since the steady sate matrix is know, the RGA is obtained as before and the next slide shows just the RGA.. Module 5 – Controllability Analysis 42 NAMP PIECE Then, the RGA obtained is: 1.89 3.58 0.70 RGA 3.02 5.60 3.58 0.13 3.02 1.89 Click to Interchange Row 2 and 3 According to the RGA, the only feasible pairing has to involve a negative RGA element, so it is possible to interchange rows 2 and 3. Rows 2 and 3 of RGA have been interchanged. Now the steady state matrix is: 1 0.1 y1 1 K 0.1 2 1 y3 2 3 1 y2 u1 u2 u3 Interchanging the rows was necessary to calculate the Niederlinski Index, because all the pairing elements must be on the diagonal of the K matrix, as next slide shows. Module 5 – Controllability Analysis 43 NAMP PIECE Now the NI shows that the system is not structurally unstable even pairing a negative element. Niederlinski Index Index Niederlinski 1. Obtain the steady state matrix lKl and its determinant. 2. Obtain the diagonal matrix of lKl and its determinant. 3. Obtain the NI ratio: NI G(0) n i 1 gii (0) Module 5 – Controllability Analysis K 0.53 0.27 Diag(K ) 2 44 NAMP PIECE According to these calculations, for the pairing 1-1/2-3/3-2, the system is not unstable, despite the rules of RGA. However if the first loop is opened (y1– u1) or not included in the process model, the resulting subsystem is unstable as will be shown. o First loop of matrix K open (K): 1 0.1 y1 1 2 1 y3 K 0.1 2 K 1 y3 3 1 y2 2 3 1 y2 u1 u2 u3u2 u3 o And the NI for matrix K is: NI K 2 (3) 1 2 2 Reminder: A negative NI indicates that the system is structurally UNSTABLE. Module 5 – Controllability Analysis 45 NAMP PIECE 2.3 Nonlinear Systems Despite that many chemical processes can be adequately represented by linear systems, via linear transfer function models, the majority of chemical processes are inherently nonlinear and sometimes need to use nonlinear models in order to be valid in a wider range of operation. 8 6 4 2 0 -2 -4 -6 5 10 5 0 -5 -5 0 -8 30 20 It is therefore necessary to see how the pairing of input and output variables of nonlinear systems is performed. -10 -10 10 0 Module 5 – Controllability Analysis 0 5 10 15 20 46 25 NAMP PIECE The same information used for the RGA of steady state systems is used for nonlinear systems. This feature can appear for someone a "disadvantage", but it is precisely this "disadvantage" involving only steady state values that can be used to handle nonlinear systems. Disadvantage or advantage??? Take a decision PLEASE !!!!! Next, an example shows that according to the steady state values, the pairing of both manipulated and controlled variables is selected. Module 5 – Controllability Analysis 47 NAMP PIECE 2.3.1 Obtain the RGA for the multivariable system, a stirred mixing tank, consisting in a hot stream and a cold stream which are used to control the liquid level and the tank temperature, and use it to recommend which of the manipulated variables should be used to control the liquid level and the tank temperature. The transfer function matrix is shown below. y1= Liquid level u1= Hot stream flowrate y2= Tank temperature u2= Cold stream temperature 1 k A s+ C 2A C hs G s = TH -Ts k A h s+ C s Ac h s 1 k k A s+ A C s+ 2A h C 2A C hs C s G s = d Tds -Ts TC -Ts k k A Chs s + A Chs s + A h A C hs C s 1 Fds k A Chs s+ A h C s 0 Where k is a constant (see next slide) and Ac is the cross-section area for the tank. Module 5 – Controllability Analysis 48 NAMP PIECE Problem description. A diagram of the tank is shown below. COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) DISTURBANCE (Td, Fd) HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) h Output Flow rate (F, T) F=k(h)½ Next it is necessary to obtain the steady state gain matrix, as next slide shows. Module 5 – Controllability Analysis 49 NAMP PIECE 2 hs k K = G 0 = TH -Ts k hs 2 hs 1 K = TH -Ts k hs 2 hs k TC -Ts k hs 2 hs TC -Ts hs Now with the steady-state gain matrix, it is possible to obtain the RGA, using the full matrix method or, since it is a 2x2 matrix, in a more direct way as seen before. Module 5 – Controllability Analysis 50 NAMP PIECE To calculate the RGA of this 2x2 system it is possible to obtain l11 using the First Principles. In Tier 1 has seen that: 1 l11 l22 1 and, K12 K 21 K11K 22 Substituting the values of the matrix K: 2 hs k = 2 hs k TH -Ts k h TH -Ts s = TC -Ts TC -Ts k h s Module 5 – Controllability Analysis TC -Ts 1 l11 = = TH -Ts TC -TH 1 TC -Ts 51 NAMP PIECE In a similar way it is possible to obtain l12 using: l12 l21 1 According to this the value of l12 is: l12 = TH -Ts TC -Ts TH -Ts 1 TC -Ts = - TH -Ts TC -TH And the RGA for this systems is: Module 5 – Controllability Analysis 52 NAMP PIECE TC -Ts TC -TH Λ= - T -T H s TC -TH - TH -Ts TC -TH TC -Ts TC -TH Here, it is due to mention that the RGA depends only of values of hot and cold streams, and also for the values of the steady state values involved in the steady-sate gain matrix. The values given to hot and cold stream are: TH = 65ºC TC = 15ºC These values are fixed and just the value of Ts will be changed according to different scenarios. Module 5 – Controllability Analysis 53 NAMP PIECE Five different values will given to Ts : Ts > (TH+TC)/2; Ts < (TH+TC)/2; Ts = 55ºC Ts = 25ºC Ts = (TH+TC)/2 ; Ts = 40ºC Ts = TH: Ts = TC; Ts = 65ºC Ts = 15ºC According to this values the system of the stirred mixing tank is: COLD STREAM FLOWRATE (u2) TC = 15ºC TANK TEMPERATURE (y2) TH = 65ºC DISTURBANCE (Td, Fd) HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) h F=k(h)½ Output Flow rate (F, T) Module 5 – Controllability Analysis 54 NAMP PIECE Case 1. Ts > (TH+TC)/2; TC COLD STREAM FLOWRATE (u2) TC= 15ºC Ts = 55ºC TANK TEMPERATURE (y2) DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) LLC h Output Flow rate (F, T) F=k(h)½ From the RGA, the suggested pairing is 1-1/2-2. The physical meaning of this pairing is: since the temperature of cold stream (TC) is farther away from the steady state Click to operating tank temperature, small changes in the cold stream Pairing produce noticeable changes in1-1/2-2 the tank temperature, whereas the temperature of the hot stream (TH) is closer to the operating steady state temperature, it can be used to control the level without causing significant changes in the tank temperature. Module 5 – Controllability Analysis 55 NAMP PIECE Case 2. Ts < (TH+TC)/2; Ts = 25ºC TC COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) TC = 15ºC DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ LLC h Output Flow rate (F, T) From the RGA, the suggested pairing is 1-2/2-1. Again, the physical meaning of this pairing is: since the temperature of the hot stream (TH) is farther away from the Click to steady state operating tank temperature, small changes in the Pairing hot stream produce noticeable1-2/2-1 changes in the tank temperature, whereas the temperature of cold stream (TC) is closer to the operating steady state temperature, it can be used to control the level without causing significant changes in the tank temperature. Module 5 – Controllability Analysis 56 NAMP PIECE Case 3. Ts = (TH+TC)/2; COLD STREAM FLOWRATE (u2) Ts = 40ºC TANK TEMPERATURE (y2) TC = 15ºC DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ h Output Flow rate (F, T) Here the values of the RGA are all equal to 0.5 For this reason it is equally bad to pair 1-1/2-2 than 12/2-1, because the operating temperature is exactly equidistant from both the cold stream temperature and the hot stream temperature. A poor control of the process under this undesirable special condition is obtained. Module 5 – Controllability Analysis 57 NAMP PIECE Case 4. Ts = TH; Ts = 65ºC TC COLD STREAM FLOWRATE (u2) TC = 15ºC TANK TEMPERATURE (y2) DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ LLC h Output Flow rate (F, T) From the RGA, the suggested pairing is 1-1/2-2. It is possible to achieve a perfect control of the level tank, without interacting with the temperature, using the hot stream. Here, the temperature of the hot stream (TH) is the same that Click to the steady state operating tank temperature. For that reason Pairing this stream is used to control the level of the tank, whereas the 1-1/2-2 temperature of the cold stream (TC) is used to control the temperature because a small change of cold stream cause significant changes in the tank temperature. Module 5 – Controllability Analysis 58 NAMP PIECE Case 5. Ts = TC; Ts = 15ºC TC COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) TC = 15ºC DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ LLC h Output Flow rate (F, T) From the RGA, the suggested pairing is 1-2/2-1. It is possible to achieve a perfect control of the tank level, without affecting the temperature, using the cold stream. Here, the temperature of the hot stream (TC) is the same than Click to the steady state operating tank temperature. For that reason Pairing this stream is used to control 1-2/2-1 the level of the tank, whereas the temperature of hot stream (TH) is used to control the temperature because a small change of hot stream cause significant changes in the tank temperature. Module 5 – Controllability Analysis 59 NAMP PIECE These five different analyses have demonstrated that the RGA can indeed be used for nonlinear as well as for linear systems. 10 Pay attention to the fact that the RGA suggests different pairings at different operating conditions. This because even that the analysis has been based on approximate linearized models, this property characteristic of the nonlinear systems is not lost. STEADY STATE 5 0 2 -5 1 -10 30 0 25 20 -1 20 15 10 10 0 5 0 -2 2 1 0 -1 -2 -2 -1 0 2 1 It is as if the nonlinear system was analyzed on different sections, or slides around `fixed` points, or in this case around steady states. Module 5 – Controllability Analysis 60 NAMP PIECE 2.4 Non Square Systems This section discusses another important point about RGA, the selection of variables for Underdefined and Overdefined systems. To do this task, it is first of all absolutely necessary to manipulate the non square system in order to obtain a square system. This is done according to the type of non square systems. The objective in non square systems is pairing, as before, the process variables to minimize the interaction between them. Next slides show how to obtain a square system from a UNDERDEFINED (therefore non square) system. Module 5 – Controllability Analysis 61 NAMP PIECE 2.4.1 Obtain the RGA of a pilot scale binary distillation column used to separate ethanol and water for which the transfer function is given below [2]. In addition, consider the side stream draw-off rate set at a fixed amount and it cannot be changed. Use the same process variables, that in Ex. 2.1.2. Problem description. Now , the side stream draw off rate is not a controlled variable, because it is fixed. For this reason the process model is: y1= overhead mole fraction ethanol y2= ethanol mole fraction in side stream y3= Temperature on Tray #19 0.66e-2.6s 6.7s+1 y1 -6.5s y = 1.11e 2 3.5s+1 y 3 -33.68e-9.2s 8.15s+1 Module 5 – Controllability Analysis u1= overhead reflux flowrate u2= reboiler steam pressure d = column feed flowrate -0.0049e-s 9.06s+1 u1 -0.012e-1.2s 7.09s+1 u2 0.87 11.61s+1 e-s 3.89s+118.8s+ 62 NAMP PIECE Distillation column used in separating ethanol and water Overhead mole fraction ethanol (y1) Temperature on tray #19 (y3) Feed flow rate (d) Overhead reflux flow rate (u1) Mole fraction of ethanol in the side stream (y2) Reboiler steam pressure (u2) Module 5 – Controllability Analysis 63 NAMP PIECE It is impossible to control all three output (ys) variables with only two input variables (us). For that reason it is necessary to select the two most important variables to be controlled, in this case the variables selected have been y1 and y3, as next diagram shows. Overhead mole fraction ethanol (y1) Temperature on tray #19 (y3) Feed flow rate (d) Overhead reflux flow rate (u1) Mole fraction of ethanol in the side stream (y2) Reboiler steam pressure (u2) Module 5 – Controllability Analysis Here the side stream mole fraction of ethanol is taken as the less important of the output variables. 64 NAMP PIECE Once it has been decided to leave the control of the side stream composition out of control scheme, the control model is now: y1= overhead mole fraction ethanol y3= Temperature on Tray #19 0.66e-2.6s 6.7s+1 y 1 y = -33.68e-9.2s 3 8.15s+1 u1= overhead reflux flowrate u2= reboiler steam pressure -0.0049e -s 9.06s+1 u1 0.87 11.61s+1 e-s u2 3.89s+1 18.8s+ This is a square (modified) subsystem. Therefore, now it is possible to perform the RGA analysis and also to obtain the additional relation: 1.11e-6.5s -0.012e-1.2s y2 = u1 + u2 3.5s+1 7.09s+1 Module 5 – Controllability Analysis 65 NAMP PIECE From the subsystem, the steady state gain matrix and the RGA obtained is: According to these values of RGA, a 1-1/2-2 pairing is recommended. It means to use the overhead reflux (u1) to control the overhead composition (y1), and use the reboiler steam pressure (u2) to control Tray #19 temperature (y2). This makes sense. Module 5 – Controllability Analysis 66 NAMP PIECE It must notice that according to the relation: 1.11e-6.5s -0.012e-1.2s y2 = u1 + u2 3.5s+1 7.09s+1 The side stream composition will drift according to the values of the overhead reflux (u1) and the reboiler steam pressure (u2). This is the nature of UNDERDEFINED systems. The previous system showed that it is only possible to achieve arbitrarily good control of two [overhead mole fraction ethanol (y1) and temperature on Tray #19 (y3)] of the three output variables and accept the drift of the third one (composition on the side stream). The strategy to work with an UNDERDEFINED system is to choose a square subsystem by dropping off the excess number of output variables on the basis of economic importance; the subsequent analysis is the same as for square systems. Module 5 – Controllability Analysis 67 NAMP PIECE Next will be show how to deal with Overdefined systems. And this is the real challenge of non square systems, so you must put all your attention… Module 5 – Controllability Analysis …and follow the instructions given in the next example. 68 NAMP PIECE 2.4.2 According to a certain system with two outputs (y1 and y2) to be controlled using two of three available inputs (u1, u2, and u3), which loop pairing is expected to give the best control?. Through pulse testing, the following transfer function model was obtained. 0.5e-0.2s y1 3s+1 y = -0.5s 0.004e 2 1.5s+1 0.07e-0.3s 2.5s+1 -0.003e-0.2s s+1 0.04e-0.03s 2.8s+1 -0.001e-0.4s 1.6s+1 u1 u 2 u3 Problem description. This is a 2x3 system, this implies that only two of the three candidate input variables will be used for control, while the third input variable will have to be set at a fixed value and will therefore be redundant. To determine which variables should be active and which ones should be redundant, first of all possible 2x2 subsystems must be obtained. Module 5 – Controllability Analysis 69 NAMP PIECE Number of subsystems = u! n! u - n! Where: u are the number of input variables and n the number of output variables. According to the previous system: u=3 and n=2 3 21 3! Number of subsystems = = =3 2! 3 - 2 ! 211 Subsystem 1. Utilizing u1 and u2 for control: 0.5e-0.2s y1 3s+1 y = -0.5s 0.004e 2 1.5s+1 Module 5 – Controllability Analysis 0.07e-0.3s 2.5s+1 u1 u -0.2s 2 -0.003e s+1 70 NAMP PIECE From the subsystem 1, the steady state matrix and the RGA are: Subsystem 2. Utilizing u1 and u2 for control: 0.5e-0.2s y1 3s+1 = y 0.004e-0.5s 2 1.5s+1 Module 5 – Controllability Analysis 0.04e-0.03s 2.8s+1 u1 u -0.4s 3 -0.001e 1.6s+1 71 NAMP PIECE From the subsystem 2, the steady state matrix and the RGA are: Subsystem 3. Utilizing u2 and u3 for control: 0.07e-0.3s y1 2.5s+1 = y -0.2s -0.003e 2 s+1 Module 5 – Controllability Analysis 0.04e-0.03s 2.8s+1 u1 u -0.4s 3 -0.001e 1.6s+1 72 NAMP PIECE From the subsystem 3, the steady state matrix and the RGA are: Next slide shows the three RGA values obtained for each subsystem. Module 5 – Controllability Analysis 73 NAMP PIECE RGA for subsystems 1 to 3. 0.843 0.157 Λ1 0.157 0.843 0.758 0.242 Λ2 0.242 0.758 1.4 2.4 Λ3 2.4 1.4 According to these values of RGA for each subsystem, the best possible control is the subsystem 1, because it is closest to subsystem the ideal situation; is somewhat subsystem This involves it pairing u1 with better y1 and than y2 with u2, and 2this andimplies far superior 3. also that u3than is tosubsystem be redundant. Module 5 – Controllability Analysis 74 NAMP PIECE 2.5 Factors Influencing the Loop Pairing As seen in TIER I, there are some factors that affect how the variables are paired. Some of those are: Constraints in the input variable, Time delay, Inverse response, Slow dynamics in the best RGA paring, Timescale decoupling of loop dynamics Next slides show how to pair the process variables according to these factors. Module 5 – Controllability Analysis 75 NAMP PIECE 2.5.1 Take again Ex. 2.3.1, but now, an in-tank heater was added to the stirred mixing tank to control the temperature with the heater power Q. Obtain the RGA for this system if Ts=(TH+TC)/2. The transfer function is: 1 1 0 k k A s+ A c s+ c 2A h 2A h c s c s G s = 1 ρCp TH -Ts TC -Ts k k k A h s+ A chs s+ A c hs s+ c s A h A h A h c s c s c s u1= Hot stream flowrate y = Liquid level 1 y2= Tank temperature u2= Cold stream temperature u3= Heater power Where k is the same constant as in Ex. 2.3.1 and Ac is the cross-section area for the tank. Module 5 – Controllability Analysis 76 NAMP PIECE Problem description. A diagram of the tank with the Heater is show below. COLD STREAM FLOWRATE (u2) HOT STREAM FLOWRATE (u1) HEAT POWER (u3) TANK TEMPERATURE (y2) DISTURBANCE (Td, Fd) h TANK LIQUID LEVEL (y1) Output Flow rate (F, T) F=k(h)½ Now it is an overdetermined system with more than one subsystem to pair. First the RGA for each subsystem will be obtained. Module 5 – Controllability Analysis 77 NAMP PIECE In a similar way as Ex. 2.4.2, there are three different subsystems: Subsystem 1. Utilizing u1 (Hot Stream) and u2 (Cold stream) for control. The steady state gain for this subsystem is the same that ex. 2.3.1: 2 hs k K= TH -Ts k hs 2 hs k TC -Ts k hs And the RGA is the same as obtained in ex. 2.3.1: TC -Ts TC -TH Λ1 = - T -T H s TC -TH Module 5 – Controllability Analysis - TH -Ts TC -TH TC -Ts TC -TH 78 NAMP PIECE Subsystem 2. Utilizing u1 (Hot Stream) and u3 (Heater) for control. The transfer function matrix for this subsystem is: 1 k A s+ c 2A h c s G(s) = TH -Ts A h s+ k c s A h c s And the steady state gain matrix is: 2 hs k K= TH -Ts k hs Module 5 – Controllability Analysis 1 ρCp k A chs s+ A h c s 0 0 1 ρCp k hs 79 NAMP PIECE To calculate the RGA of this 2x2 system it is possible to obtain l11 using the First Principles. Again: 1 l11 l22 1 and, K12 K 21 K11K 22 Substituting the values of the matrix K of this subsystem: TH -Ts (0) k h s = =0 1 2 hs ρCp k k h s Module 5 – Controllability Analysis Therefore l11 is: 1 l11 1 1 0 80 NAMP PIECE And since, l12: l12 l21 1 Finally the value of l12 is: And the RGA for subsystem 2 is : l12 0 1 0 Λ2 = 0 1 Next the RGA for subsystem 3, will be obtained. Module 5 – Controllability Analysis 81 NAMP PIECE Subsystem 3. Utilizing u2 (Cold Stream) y u3 (Heater) for control. The transfer function matrix for this subsystem is: 1 k A s+ c 2A c hs G(s) = TC -Ts A h s+ k c s A h c s And the steady state gain matrix is: 2 hs k K= TC -Ts k hs Module 5 – Controllability Analysis 1 ρCp k A chs s+ A h c s 0 0 1 ρCp k hs 82 NAMP PIECE To calculate the RGA of this 2x2 system it is possible to obtain l11 using the First Principles. Again: 1 l11 l22 1 and, K12 K 21 K11K 22 Substituting the values of the matrix K of this subsystem: TC -Ts (0) k h s = =0 1 2 hs ρCp k k h s Module 5 – Controllability Analysis Therefore l11 is: 1 l11 1 1 0 83 NAMP PIECE And since, l12: l12 l21 1 Finally the value of l12 is: l12 0 1 0 And the RGA for subsystem 3 is : Λ 3 = 0 1 You must noted that the RGA for subsystem 2 and 3 is the same and both are independent of Ts. Module 5 – Controllability Analysis 84 NAMP PIECE RGA for subsystems 1 to 3. TC -Ts TC -TH Λ1 = - T -T H s TC -TH - TH -Ts TC -TH TC -Ts TC -TH 1 0 Λ2 = 0 1 1 0 Λ3 = 0 1 Taking the case where Ts=(TH+TC)/2, the RGA for each subsystem is: Subsystem 1. Subsystem 2. Subsystem 3. 0.5 0.5 Λ1 = 0.5 0.5 1 0 Λ2 = 0 1 1 0 Λ3 = 0 1 Again, note that the RGA for subsystem 1 was obtained in Ex. 2.3.1. Subsystems 2 and 3 are the same and both are independent of Ts. Module 5 – Controllability Analysis 85 NAMP PIECE According to this analysis, the pairing in subsystem 2 involves to use the Hot stream temperature (u1) to control the liquid level (y1) and use the in tank heater (u3) to control the tank temperature (y2): Subsystem 2. COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) DISTURBANCE (Td, Fd) HOT STREAM FLOWRATE (u1) Subsystem 3. LLC h TANK LIQUID LEVEL (y1) HEAT POWER (u3) TC Click to Pairing Subsystem 2 Module 5 – Controllability Analysis F=k(h)½ Output Flow rate (F, T) 86 NAMP PIECE For subsystem 3 the pairing involves to use the Cold stream temperature (u2) to control the liquid level (y1) and use the in-tank heater (u3) to control the tank temperature (y2): Subsystem 3. TANK TEMPERATURE (y2) COLD STREAM FLOWRATE (u2) HOT STREAM FLOWRATE (u1) DISTURBANCE (Td, Fd) Subsystem 3. LLC h HEAT POWER (u3) TC Click to Pairing Subsystem 3 Module 5 – Controllability Analysis TANK LIQUID LEVEL (y1) F=k(h)½ Output Flow rate (F, T) 87 NAMP PIECE But both pairings of these subsystems can become an undesirable pairing control as it will be discussed next. If the in-tank heater can barely achieve the steady state, Ts, at maximum power, there is a major problem. Thus, this subsystem would not be desirable for the regulatory temperature control because, following variations of the other process variable (hot or cold stream), the IN-TANK HEATER has no more power to supply (or extract) heat to keep the new steady state temperature. Module 5 – Controllability Analysis 88 NAMP PIECE Now to overcome the power limitation, a much larger heater is installed in the tank, but as a consequence of this, there is a VERY LARGE TIME DELAY, between the control signal and the actual power delivery. And because of this sluggish closed-loop response in the heater, the best choice for pairing the process variables could be the poor RGA of subsystem 1. Next will be show another factor to considerer in the loop pairing of process variables. Module 5 – Controllability Analysis 89 NAMP PIECE 2.5.2 Now considering a system with a transfer function given as below, obtain the RGA for this system and analyze a unit set point change in (y1) and a diagonal PI controller (Kc1= 4,t I1=0.5; Kc2=-4, t I2=0.3) using the resulting pairing. 2 2 10s+1 s+1 G s = -4 1 s+1 10s+1 Problem description. First of all it is necessary to obtain the steady state gain matrix, as it is shows below. 2 2 G 0 =K= 1 4 Module 5 – Controllability Analysis 90 NAMP PIECE Now, the RGA obtained as before is: According to the RGA, the recommended pairing is y1-u1 and y2-u2. Next step is to analyze a change in the set point. Module 5 – Controllability Analysis 91 NAMP PIECE Since the dynamic simulation of the analyzed system is beyond the scope of this module, only the result of the change in the set point will be display. Module 5 – Controllability Analysis 92 NAMP PIECE As mentioned in last slide, the next graphic shows the closed loop response for a unit set point change in y1 using the pairing suggested for the RGA and a diagonal PI controller. Pairing 1-1/2-2 y1 y2 Set point y1 The performance of this Despite this “not too pairing is not too bad bad” performance, considering that the open the inverse pairing loop time constants on will be analyzed for the diagonal are 10 the same set point minutes change. point Inverse loop pairing involves to take the value ofSet l=0.2 inythe 2 pairing, but it has been mentioned as a situation to avoid !!!!!!!!. Module 5 – Controllability Analysis 93 NAMP PIECE Different pairing also implies to use a different PI controller, for that reason the inverse pairing analysis of a unit set point change in (y1), the new diagonal PI controller is (Kc1= 10,t I1=0.3; Kc2= 20, t I2=0.3). Pairing 1-2/2-1 Set point y1 y1 The reason is that The performance in this case is the control loops are dramatically better than able the to respond so recommended pairing byrapidly the that the RGA, because the open loop interactions that time constants on the diagonal appear more slowly are only 1 minute. are easily dealt with. Set point y2 y 2 Finally in this example, the best loop pairing was obtained using the inverse pairing, than the suggested by the RGA. Module 5 – Controllability Analysis 94 NAMP PIECE After this example, do you fell like this ?... You should not fell like any of The purpose of this example is not to confuse you about how to this, select a because… loop pairing, the purpose is to show you that RGA provides only a guideline to steady state interactions, for that reason, all other engineering considerations must be used together in choosing the loop pairing. Module 5 – Controllability Analysis 95 NAMP PIECE 2.3 Nonlinear Systems Despite that many chemical processes can be adequately represented by linear systems, via linear transfer function models, the majority of chemical processes are inherently nonlinear and sometimes need to use nonlinear models in order to be valid in a wider range of operation. 8 6 4 2 0 -2 -4 -6 5 10 5 0 -5 -5 0 -8 30 20 It is therefore necessary to see how the pairing of input and output variables of nonlinear systems is performed. -10 -10 10 0 Module 5 – Controllability Analysis 0 5 10 15 20 96 25 NAMP PIECE The same information used for the RGA of steady state systems is used for nonlinear systems. This feature can appear for someone a "disadvantage", but it is precisely this "disadvantage" involving only steady state values that can be used to handle nonlinear systems. Disadvantage or advantage??? Take a decision PLEASE !!!!! Next, an example shows that according to the steady state values, the pairing of both manipulated and controlled variables is selected. Module 5 – Controllability Analysis 97 NAMP PIECE 2.3.1 Obtain the RGA for the multivariable system, a stirred mixing tank, consisting in a hot stream and a cold stream which are used to control the liquid level and the tank temperature, and use it to recommend which of the manipulated variables should be used to control the liquid level and the tank temperature. The transfer function matrix is shown below. y1= Liquid level u1= Hot stream flowrate y2= Tank temperature u2= Cold stream temperature 1 k A s+ C 2A C hs G s = TH -Ts k A h s+ C s Ac h s 1 k k A s+ A C s+ 2A h C 2A C hs C s G s = d Tds -Ts TC -Ts k k A Chs s + A Chs s + A h A C hs C s 1 Fds k A Chs s+ A h C s 0 Where k is a constant (see next slide) and Ac is the cross-section area for the tank. Module 5 – Controllability Analysis 98 NAMP PIECE Problem description. A diagram of the tank is shown below. COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) DISTURBANCE (Td, Fd) HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) h Output Flow rate (F, T) F=k(h)½ Next it is necessary to obtain the steady state gain matrix, as next slide shows. Module 5 – Controllability Analysis 99 NAMP PIECE 2 hs k K = G 0 = TH -Ts k hs 2 hs 1 K = TH -Ts k hs 2 hs k TC -Ts k hs 2 hs TC -Ts hs Now with the steady-state gain matrix, it is possible to obtain the RGA, using the full matrix method or, since it is a 2x2 matrix, in a more direct way as seen before. Module 5 – Controllability Analysis 100 NAMP PIECE To calculate the RGA of this 2x2 system it is possible to obtain l11 using the First Principles. In Tier 1 has seen that: 1 l11 l22 1 and, K12 K 21 K11K 22 Substituting the values of the matrix K: 2 hs k = 2 hs k TH -Ts k h TH -Ts s = TC -Ts TC -Ts k h s Module 5 – Controllability Analysis TC -Ts 1 l11 = = TH -Ts TC -TH 1 TC -Ts 101 NAMP PIECE In a similar way it is possible to obtain l12 using: l12 l21 1 According to this the value of l12 is: l12 = TH -Ts TC -Ts TH -Ts 1 TC -Ts = - TH -Ts TC -TH And the RGA for this systems is: Module 5 – Controllability Analysis 102 NAMP PIECE TC -Ts TC -TH Λ= - T -T H s TC -TH - TH -Ts TC -TH TC -Ts TC -TH Here, it is due to mention that the RGA depends only of values of hot and cold streams, and also for the values of the steady state values involved in the steady-sate gain matrix. The values given to hot and cold stream are: TH = 65ºC TC = 15ºC These values are fixed and just the value of Ts will be changed according to different scenarios. Module 5 – Controllability Analysis 103 NAMP PIECE Five different values will given to Ts : Ts > (TH+TC)/2; Ts < (TH+TC)/2; Ts = 55ºC Ts = 25ºC Ts = (TH+TC)/2 ; Ts = 40ºC Ts = TH: Ts = TC; Ts = 65ºC Ts = 15ºC According to this values the system of the stirred mixing tank is: COLD STREAM FLOWRATE (u2) TC = 15ºC TANK TEMPERATURE (y2) TH = 65ºC DISTURBANCE (Td, Fd) HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) h F=k(h)½ Output Flow rate (F, T) Module 5 – Controllability Analysis 104 NAMP PIECE Case 1. Ts > (TH+TC)/2; TC COLD STREAM FLOWRATE (u2) TC= 15ºC Ts = 55ºC TANK TEMPERATURE (y2) DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) LLC h Output Flow rate (F, T) F=k(h)½ From the RGA, the suggested pairing is 1-1/2-2. The physical meaning of this pairing is: since the temperature of cold stream (TC) is farther away from the steady state Click to operating tank temperature, small changes in the cold stream Pairing produce noticeable changes in1-1/2-2 the tank temperature, whereas the temperature of the hot stream (TH) is closer to the operating steady state temperature, it can be used to control the level without causing significant changes in the tank temperature. Module 5 – Controllability Analysis 105 NAMP PIECE Case 2. Ts < (TH+TC)/2; Ts = 25ºC TC COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) TC = 15ºC DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ LLC h Output Flow rate (F, T) From the RGA, the suggested pairing is 1-2/2-1. Again, the physical meaning of this pairing is: since the temperature of the hot stream (TH) is farther away from the Click to steady state operating tank temperature, small changes in the Pairing hot stream produce noticeable1-2/2-1 changes in the tank temperature, whereas the temperature of cold stream (TC) is closer to the operating steady state temperature, it can be used to control the level without causing significant changes in the tank temperature. Module 5 – Controllability Analysis 106 NAMP PIECE Case 3. Ts = (TH+TC)/2; COLD STREAM FLOWRATE (u2) Ts = 40ºC TANK TEMPERATURE (y2) TC = 15ºC DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ h Output Flow rate (F, T) Here the values of the RGA are all equal to 0.5 For this reason it is equally bad to pair 1-1/2-2 than 12/2-1, because the operating temperature is exactly equidistant from both the cold stream temperature and the hot stream temperature. A poor control of the process under this undesirable special condition is obtained. Module 5 – Controllability Analysis 107 NAMP PIECE Case 4. Ts = TH; Ts = 65ºC TC COLD STREAM FLOWRATE (u2) TC = 15ºC TANK TEMPERATURE (y2) DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ LLC h Output Flow rate (F, T) From the RGA, the suggested pairing is 1-1/2-2. It is possible to achieve a perfect control of the level tank, without interacting with the temperature, using the hot stream. Here, the temperature of the hot stream (TH) is the same that Click to the steady state operating tank temperature. For that reason Pairing this stream is used to control the level of the tank, whereas the 1-1/2-2 temperature of the cold stream (TC) is used to control the temperature because a small change of cold stream cause significant changes in the tank temperature. Module 5 – Controllability Analysis 108 NAMP PIECE Case 5. Ts = TC; Ts = 15ºC TC COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) TC = 15ºC DISTURBANCE (Td,Fd) TH = 65ºC HOT STREAM FLOWRATE (u1) TANK LIQUID LEVEL (y1) F=k(h)½ LLC h Output Flow rate (F, T) From the RGA, the suggested pairing is 1-2/2-1. It is possible to achieve a perfect control of the tank level, without affecting the temperature, using the cold stream. Here, the temperature of the hot stream (TC) is the same than Click to the steady state operating tank temperature. For that reason Pairing this stream is used to control 1-2/2-1 the level of the tank, whereas the temperature of hot stream (TH) is used to control the temperature because a small change of hot stream cause significant changes in the tank temperature. Module 5 – Controllability Analysis 109 NAMP PIECE These five different analyses have demonstrated that the RGA can indeed be used for nonlinear as well as for linear systems. 10 Pay attention to the fact that the RGA suggests different pairings at different operating conditions. This because even that the analysis has been based on approximate linearized models, this property characteristic of the nonlinear systems is not lost. STEADY STATE 5 0 2 -5 1 -10 30 0 25 20 -1 20 15 10 10 0 5 0 -2 2 1 0 -1 -2 -2 -1 0 1 It is as if the nonlinear system was analyzed on different sections, or slides around `fixed` points, or in this case around steady states. Module 5 – Controllability Analysis 110 2 NAMP PIECE 2.4 Non Square Systems This section discusses another important point about RGA, the selection of variables for Underdefined and Overdefined systems. To do this task, it is first of all absolutely necessary to manipulate the non square system in order to obtain a square system. This is done according to the type of non square systems. The objective in non square systems is pairing, as before, the process variables to minimize the interaction between them. Next slides show how to obtain a square system from a UNDERDEFINED (therefore non square) system. Module 5 – Controllability Analysis 111 NAMP PIECE 2.4.1 Obtain the RGA of a pilot scale binary distillation column used to separate ethanol and water for which the transfer function is given below [2]. In addition, consider the side stream draw-off rate set at a fixed amount and it cannot be changed. Use the same process variables, that in Ex. 2.1.2. Problem description. Now , the side stream draw off rate is not a controlled variable, because it is fixed. For this reason the process model is: y1= overhead mole fraction ethanol y2= ethanol mole fraction in side stream y3= Temperature on Tray #19 0.66e-2.6s 6.7s+1 y1 -6.5s y = 1.11e 2 3.5s+1 y 3 -33.68e-9.2s 8.15s+1 Module 5 – Controllability Analysis u1= overhead reflux flowrate u2= reboiler steam pressure d = column feed flowrate -0.0049e-s 9.06s+1 u1 -0.012e-1.2s 7.09s+1 u2 0.87 11.61s+1 e-s 3.89s+118.8s+ 112 NAMP PIECE Distillation column used in separating ethanol and water Overhead mole fraction ethanol (y1) Temperature on tray #19 (y3) Feed flow rate (d) Overhead reflux flow rate (u1) Mole fraction of ethanol in the side stream (y2) Reboiler steam pressure (u2) Module 5 – Controllability Analysis 113 NAMP PIECE It is impossible to control all three output (ys) variables with only two input variables (us). For that reason it is necessary to select the two most important variables to be controlled, in this case the variables selected have been y1 and y3, as next diagram shows. Overhead mole fraction ethanol (y1) Temperature on tray #19 (y3) Feed flow rate (d) Overhead reflux flow rate (u1) Mole fraction of ethanol in the side stream (y2) Reboiler steam pressure (u2) Module 5 – Controllability Analysis Here the side stream mole fraction of ethanol is taken as the less important of the output variables. 114 NAMP PIECE Once it has been decided to leave the control of the side stream composition out of control scheme, the control model is now: y1= overhead mole fraction ethanol y3= Temperature on Tray #19 0.66e-2.6s 6.7s+1 y 1 y = -33.68e-9.2s 3 8.15s+1 u1= overhead reflux flowrate u2= reboiler steam pressure -0.0049e -s 9.06s+1 u1 0.87 11.61s+1 e-s u2 3.89s+1 18.8s+ This is a square (modified) subsystem. Therefore, now it is possible to perform the RGA analysis and also to obtain the additional relation: 1.11e-6.5s -0.012e-1.2s y2 = u1 + u2 3.5s+1 7.09s+1 Module 5 – Controllability Analysis 115 NAMP PIECE From the subsystem, the steady state gain matrix and the RGA obtained is: According to these values of RGA, a 1-1/2-2 pairing is recommended. It means to use the overhead reflux (u1) to control the overhead composition (y1), and use the reboiler steam pressure (u2) to control Tray #19 temperature (y2). This makes sense. Module 5 – Controllability Analysis 116 NAMP PIECE It must notice that according to the relation: 1.11e-6.5s -0.012e-1.2s y2 = u1 + u2 3.5s+1 7.09s+1 The side stream composition will drift according to the values of the overhead reflux (u1) and the reboiler steam pressure (u2). This is the nature of UNDERDEFINED systems. The previous system showed that it is only possible to achieve arbitrarily good control of two [overhead mole fraction ethanol (y1) and temperature on Tray #19 (y3)] of the three output variables and accept the drift of the third one (composition on the side stream). The strategy to work with an UNDERDEFINED system is to choose a square subsystem by dropping off the excess number of output variables on the basis of economic importance; the subsequent analysis is the same as for square systems. Module 5 – Controllability Analysis 117 NAMP PIECE Next will be show how to deal with Overdefined systems. And this is the real challenge of non square systems, so you must put all your attention… Module 5 – Controllability Analysis …and follow the instructions given in the next example. 118 NAMP PIECE 2.4.2 According to a certain system with two outputs (y1 and y2) to be controlled using two of three available inputs (u1, u2, and u3), which loop pairing is expected to give the best control?. Through pulse testing, the following transfer function model was obtained. 0.5e-0.2s y1 3s+1 y = -0.5s 0.004e 2 1.5s+1 0.07e-0.3s 2.5s+1 -0.003e-0.2s s+1 0.04e-0.03s 2.8s+1 -0.001e-0.4s 1.6s+1 u1 u 2 u3 Problem description. This is a 2x3 system, this implies that only two of the three candidate input variables will be used for control, while the third input variable will have to be set at a fixed value and will therefore be redundant. To determine which variables should be active and which ones should be redundant, first of all possible 2x2 subsystems must be obtained. Module 5 – Controllability Analysis 119 NAMP PIECE Number of subsystems = u! n! u - n! Where: u are the number of input variables and n the number of output variables. According to the previous system: u=3 and n=2 3 21 3! Number of subsystems = = =3 2! 3 - 2 ! 211 Subsystem 1. Utilizing u1 and u2 for control: 0.5e-0.2s y1 3s+1 y = -0.5s 0.004e 2 1.5s+1 Module 5 – Controllability Analysis 0.07e-0.3s 2.5s+1 u1 u -0.2s 2 -0.003e s+1 120 NAMP PIECE From the subsystem 1, the steady state matrix and the RGA are: Subsystem 2. Utilizing u1 and u2 for control: 0.5e-0.2s y1 3s+1 = y 0.004e-0.5s 2 1.5s+1 Module 5 – Controllability Analysis 0.04e-0.03s 2.8s+1 u1 u -0.4s 3 -0.001e 1.6s+1 121 NAMP PIECE From the subsystem 2, the steady state matrix and the RGA are: Subsystem 3. Utilizing u2 and u3 for control: 0.07e-0.3s y1 2.5s+1 = y -0.2s -0.003e 2 s+1 Module 5 – Controllability Analysis 0.04e-0.03s 2.8s+1 u1 u -0.4s 3 -0.001e 1.6s+1 122 NAMP PIECE From the subsystem 3, the steady state matrix and the RGA are: Next slide shows the three RGA values obtained for each subsystem. Module 5 – Controllability Analysis 123 NAMP PIECE RGA for subsystems 1 to 3. 0.843 0.157 Λ1 0.157 0.843 0.758 0.242 Λ2 0.242 0.758 1.4 2.4 Λ3 2.4 1.4 According to these values of RGA for each subsystem, the best possible control is the subsystem 1, because it is closest to subsystem the ideal situation; is somewhat subsystem This involves it pairing u1 with better y1 and than y2 with u2, and 2this andimplies far superior 3. also that u3than is tosubsystem be redundant. Module 5 – Controllability Analysis 124 NAMP PIECE 2.5 Factors Influencing the Loop Pairing As seen in TIER I, there are some factors that affect how the variables are paired. Some of those are: Constraints in the input variable, Time delay, Inverse response, Slow dynamics in the best RGA paring, Timescale decoupling of loop dynamics Next slides show how to pair the process variables according to these factors. Module 5 – Controllability Analysis 125 NAMP PIECE 2.5.1 Take again Ex. 2.3.1, but now, an in-tank heater was added to the stirred mixing tank to control the temperature with the heater power Q. Obtain the RGA for this system if Ts=(TH+TC)/2. The transfer function is: 1 1 0 k k A s+ A c s+ c 2A h 2A h c s c s G s = 1 ρCp TH -Ts TC -Ts k k k A h s+ A chs s+ A c hs s+ c s A h A h A h c s c s c s u1= Hot stream flowrate y = Liquid level 1 y2= Tank temperature u2= Cold stream temperature u3= Heater power Where k is the same constant as in Ex. 2.3.1 and Ac is the cross-section area for the tank. Module 5 – Controllability Analysis 126 NAMP PIECE Problem description. A diagram of the tank with the Heater is show below. COLD STREAM FLOWRATE (u2) HOT STREAM FLOWRATE (u1) HEAT POWER (u3) TANK TEMPERATURE (y2) DISTURBANCE (Td, Fd) h TANK LIQUID LEVEL (y1) Output Flow rate (F, T) F=k(h)½ Now it is an overdetermined system with more than one subsystem to pair. First the RGA for each subsystem will be obtained. Module 5 – Controllability Analysis 127 NAMP PIECE In a similar way as Ex. 2.4.2, there are three different subsystems: Subsystem 1. Utilizing u1 (Hot Stream) and u2 (Cold stream) for control. The steady state gain for this subsystem is the same that ex. 2.3.1: 2 hs k K= TH -Ts k hs 2 hs k TC -Ts k hs And the RGA is the same as obtained in ex. 2.3.1: TC -Ts TC -TH Λ1 = - T -T H s TC -TH Module 5 – Controllability Analysis - TH -Ts TC -TH TC -Ts TC -TH 128 NAMP PIECE Subsystem 2. Utilizing u1 (Hot Stream) and u3 (Heater) for control. The transfer function matrix for this subsystem is: 1 k A s+ c 2A h c s G(s) = TH -Ts A h s+ k c s A h c s And the steady state gain matrix is: 2 hs k K= TH -Ts k hs Module 5 – Controllability Analysis 1 ρCp k A chs s+ A h c s 0 0 1 ρCp k hs 129 NAMP PIECE To calculate the RGA of this 2x2 system it is possible to obtain l11 using the First Principles. Again: 1 l11 l22 1 and, K12 K 21 K11K 22 Substituting the values of the matrix K of this subsystem: TH -Ts (0) k h s = =0 1 2 hs ρCp k k h s Module 5 – Controllability Analysis Therefore l11 is: 1 l11 1 1 0 130 NAMP PIECE And since, l12: l12 l21 1 Finally the value of l12 is: And the RGA for subsystem 2 is : l12 0 1 0 Λ2 = 0 1 Next the RGA for subsystem 3, will be obtained. Module 5 – Controllability Analysis 131 NAMP PIECE Subsystem 3. Utilizing u2 (Cold Stream) y u3 (Heater) for control. The transfer function matrix for this subsystem is: 1 k A s+ c 2A c hs G(s) = TC -Ts A h s+ k c s A h c s And the steady state gain matrix is: 2 hs k K= TC -Ts k hs Module 5 – Controllability Analysis 1 ρCp k A chs s+ A h c s 0 0 1 ρCp k hs 132 NAMP PIECE To calculate the RGA of this 2x2 system it is possible to obtain l11 using the First Principles. Again: 1 l11 l22 1 and, K12 K 21 K11K 22 Substituting the values of the matrix K of this subsystem: TC -Ts (0) k h s = =0 1 2 hs ρCp k k h s Module 5 – Controllability Analysis Therefore l11 is: 1 l11 1 1 0 133 NAMP PIECE And since, l12: l12 l21 1 Finally the value of l12 is: l12 0 1 0 And the RGA for subsystem 3 is : Λ 3 = 0 1 You must noted that the RGA for subsystem 2 and 3 is the same and both are independent of Ts. Module 5 – Controllability Analysis 134 NAMP PIECE RGA for subsystems 1 to 3. TC -Ts TC -TH Λ1 = - T -T H s TC -TH - TH -Ts TC -TH TC -Ts TC -TH 1 0 Λ2 = 0 1 1 0 Λ3 = 0 1 Taking the case where Ts=(TH+TC)/2, the RGA for each subsystem is: Subsystem 1. Subsystem 2. Subsystem 3. 0.5 0.5 Λ1 = 0.5 0.5 1 0 Λ2 = 0 1 1 0 Λ3 = 0 1 Again, note that the RGA for subsystem 1 was obtained in Ex. 2.3.1. Subsystems 2 and 3 are the same and both are independent of Ts. Module 5 – Controllability Analysis 135 NAMP PIECE According to this analysis, the pairing in subsystem 2 involves to use the Hot stream temperature (u1) to control the liquid level (y1) and use the in tank heater (u3) to control the tank temperature (y2): Subsystem 2. COLD STREAM FLOWRATE (u2) TANK TEMPERATURE (y2) DISTURBANCE (Td, Fd) HOT STREAM FLOWRATE (u1) Subsystem 3. LLC h TANK LIQUID LEVEL (y1) HEAT POWER (u3) TC Click to Pairing Subsystem 2 Module 5 – Controllability Analysis F=k(h)½ Output Flow rate (F, T) 136 NAMP PIECE For subsystem 3 the pairing involves to use the Cold stream temperature (u2) to control the liquid level (y1) and use the in-tank heater (u3) to control the tank temperature (y2): Subsystem 3. TANK TEMPERATURE (y2) COLD STREAM FLOWRATE (u2) HOT STREAM FLOWRATE (u1) DISTURBANCE (Td, Fd) Subsystem 3. LLC h HEAT POWER (u3) TC Click to Pairing Subsystem 3 Module 5 – Controllability Analysis TANK LIQUID LEVEL (y1) F=k(h)½ Output Flow rate (F, T) 137 NAMP PIECE But both pairings of these subsystems can become an undesirable pairing control as it will be discussed next. If the in-tank heater can barely achieve the steady state, Ts, at maximum power, there is a major problem. Thus, this subsystem would not be desirable for the regulatory temperature control because, following variations of the other process variable (hot or cold stream), the IN-TANK HEATER has no more power to supply (or extract) heat to keep the new steady state temperature. Module 5 – Controllability Analysis 138 NAMP PIECE Now to overcome the power limitation, a much larger heater is installed in the tank, but as a consequence of this, there is a VERY LARGE TIME DELAY, between the control signal and the actual power delivery. And because of this sluggish closed-loop response in the heater, the best choice for pairing the process variables could be the poor RGA of subsystem 1. Next will be show another factor to considerer in the loop pairing of process variables. Module 5 – Controllability Analysis 139 NAMP PIECE 2.5.2 Now considering a system with a transfer function given as below, obtain the RGA for this system and analyze a unit set point change in (y1) and a diagonal PI controller (Kc1= 4,t I1=0.5; Kc2=-4, t I2=0.3) using the resulting pairing. 2 2 10s+1 s+1 G s = -4 1 s+1 10s+1 Problem description. First of all it is necessary to obtain the steady state gain matrix, as it is shows below. 2 2 G 0 =K= 1 4 Module 5 – Controllability Analysis 140 NAMP PIECE Now, the RGA obtained as before is: According to the RGA, the recommended pairing is y1-u1 and y2-u2. Next step is to analyze a change in the set point. Module 5 – Controllability Analysis 141 NAMP PIECE Since the dynamic simulation of the analyzed system is beyond the scope of this module, only the result of the change in the set point will be display. Module 5 – Controllability Analysis 142 NAMP PIECE As mentioned in last slide, the next graphic shows the closed loop response for a unit set point change in y1 using the pairing suggested for the RGA and a diagonal PI controller. Pairing 1-1/2-2 y1 y2 Set point y1 The performance of this Despite this “not too pairing is not too bad bad” performance, considering that the open the inverse pairing loop time constants on will be analyzed for the diagonal are 10 the same set point minutes change. point Inverse loop pairing involves to take the value ofSet l=0.2 inythe 2 pairing, but it has been mentioned as a situation to avoid !!!!!!!!. Module 5 – Controllability Analysis 143 NAMP PIECE Different pairing also implies to use a different PI controller, for that reason the inverse pairing analysis of a unit set point change in (y1), the new diagonal PI controller is (Kc1= 10,t I1=0.3; Kc2= 20, t I2=0.3). Pairing 1-2/2-1 Set point y1 y1 The reason is that The performance in this case is the control loops are dramatically better than able the to respond so recommended pairing byrapidly the that the RGA, because the open loop interactions that time constants on the diagonal appear more slowly are only 1 minute. are easily dealt with. Set point y2 y 2 Finally in this example, the best loop pairing was obtained using the inverse pairing, than the suggested by the RGA. Module 5 – Controllability Analysis 144 NAMP PIECE After this example, do you fell like this ?... You should not fell like any of The purpose of this example is not to confuse you about how to this, select a because… loop pairing, the purpose is to show you that RGA provides only a guideline to steady state interactions, for that reason, all other engineering considerations must be used together in choosing the loop pairing. Module 5 – Controllability Analysis 145