Solution Optimizer - Universal Technical Systems
Contents
1. Ta 6 Mazimize Minimize Value of Cancel By Changing alpha Bounds Subject to the Constraints Add Optimize Options Reset All Help Delete 7 5 Part 2 Equation Solving Select the target variable and the variable s that the Optimizer will modify the decision variables by using the selection buttons We will set the value of Equal To to Maximize the default No constraints are required for this problem Clicking the Optimize button causes the model to solve repeatedly under control of the Optimizer which modifies alpha each run until area reaches a maximum 2 at the angle 45 degrees as expected 2 Typical Example With Constraints Using the Optimizer Wizard The following problem is taken from Basic Programs for Production and Operation Management by Pantumsinchai Hassan and Gupta Prentice Hall 1983 Butane Virgin Naphtha and Catalytic Cracked Gasoline are to be blended into Super and Regular unleaded gasoline The goal is to blend the components so as to produce the two types of gasoline with a maximum profit There are six unknowns corresponding with the daily consumption of components used for each kind of gasoline This can be represented by two equations with TS and TR representing the total units of Super and Regular produced For example the variable x11 represents the units of Butane2. P mati 1658040 A Amortization cost of pipe mate 4 2 ee Total annual cost of operation 7 38 Chapter 7 Solution Optimizer In order to store the optimum inputs for flows within a range 1 We associate lists with the relevant variables Q L D PlandT 2 For handling ease we collect these lists into a table 3 We run the Optimizer with the current set of values 4 We save the results in the relevant lists by using Command gt Put Values to Lists 5 We repeatedly do the same with the other sets of values After repeating the process for all values of Q we have the following table of results mt 0 0 P 10 000 000 35 6563 37 0316 190 74 9 7848860 o 9 000 000 37 4406 35 2849 200066 9 383 160 8 000 000 39 5446 334185 211 266 8 915 380 It is frequently more informative to present list results in plot form 2 Notice that we are NOT going to List Solve F10 7 39 Part 2 Equation Solving PLOT TotalCost Minimum Annual Cost vs Flow Rate 10000000 5000000 3000000 2000000 r c i R m 1000000 1000000 2000000 5000000 5000000 10000000 Volumetric Flow Rate using log log scales Eli PLOT Interval Compressor separation vs Flow Rate Minimum Cost Condition 90 80 20 BU s0 40 30 1000000 2000000 5000000 5000000 10000000 Volumetric Flow Ra
3. E G0t rr P The variance of the returns is then PROCEDURE port OO computed using another loop within the 7 er procedure For each scenario the overall 7575 expected return is subtracted from each 2 of the expected scenario returns That T L MEH re difference is squared and the result iah a C Expected ret f h i multiplied by the associated probability ch 1750106 eee ae The results are then summed The m m Z required statements are added to the o for i 1 to m port procedure function Here is the tli doteF sip E complete function Overall expected return E dot r P C variance f r i 1 t m vey PLi 2222 lt ll fol lal E The local variables E and v are declared as Output Variables in the function and the rule is modified to accept the change El Rules Miil Ei Rule call porte E w Solving F9 the Variable Sheet displays the results for the collection of inputs in the tables Variables OR 0 Erm Expectedretwn OOOO 7 1 e 0 A common goal in investing is to obtain a certain expected level of return with a minimum variance In other words there may be many ways of obtaining a particular level of return but those that are mote likely to result in greater fluctuations from the expectation are sometimes less desirable In this case we are interested in determining the fraction to invest in each o
4. 2724187157 341064825 Here are the last few elements Y lz A 58 59 149 432402013 438849371 150 425722512 43139187 419126792 4235585155 We can easily create a TK Line Chart to plot y and z vs x 10 5 2 y 2 y FAT AA 4 32 x ee x x and z eee eee ae a KA js ae ee oj azi an nir arg AA AA 1 1 1 1 r 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 eee 7 31 Part 2 Equation Solving If these solutions are to be used in other parts of the model they can be applied in TK List Functions Here is the definition of the y function S LIST FUNCTION Fy The x list is used as the domain of the function The y list is used as the range list Using linear interpolation the function will be continuous over the entire domain of x The function will be undefined for values of x outside the domain Comment Wx Domain List x Mapping Linear Range List v b Here are a few elements of the values used in the function bho hb o Y 15 5 469336142769403 The fz function is set up in the same way We can now reference these functions anywhere else in the model Here are two rules on the rule sheet amp Rules OR x Status Rule Unsati y yix Unsati 2 2 If we enter a value of 0 2135 fo
5. Solution Optimizer Constraints can look exactly the same as bounds but constraints are repeatedly updated and checked and they can be temporarily violated during the course of the optimization process For example if you have a variable p whose result should not be more than 80 of pcrit you can set a constraint p lt 0 8 pcrit and the Optimizer will use that as a guide At times during the process p might exceed the limit but the Optimizer will not finish until the constraint is satisfied When all the constraints are true and the Optimizer fails to improve on the target condition the Optimizer is finished and the results are posted Adding Constraints Clicking the Add button opens the Add Edit Constraint dialog for entry of constraints Any valid TK expressions are allowed This dialog will also appear if you double click on any existing constraints in the Optimization Parameters window In that case the constraint will be displayed and the contents can be edited An alternative to double clicking on a constraint is to highlight the constraint and click the Change button To remove a constraint highlight it and click the Delete button Add Edit Constraint N Cancel Add After the expressions have been entered in the Add Edit Constraint dialog click the Add button to add another constraint or the OK button to close the dialog and return to the Optimization Parameters window The Cancel button cancels an
6. 0100 Tolerance o co0000 m Constraint tolerance Convergence Reset to Default Values Search Derivatives Estimates Help f Tangent Forward Newton Cancel C Quadratic Central Conjugate Maximum time allowed defaults to 100 seconds You may want to extend this for problems that you knovv vvill take a long time to solve The limit is 32000 Maximum iterations allowed defaults to 1000 You may want to extend this for problems that you know require many iterations The limit is 32000 Precision controls the precision of solutions by using the number you enter to determine whether the value of a constraint cell meets a target or satisfies a lower or upper bound Precision must be indicated with a fractional number between 0 and 1 Higher precision is indicated when the number you enter has more decimal places for example 0 0001 is higher precision than 0 01 The default is 0 0001 Tolerance is used to compare values with the closest integer when integer variable bounds are used Constraint Tolerance Convergence When the relative change in the target cell value is less than the number in the Convergence box for the last five iterations TK stops Convergence applies only to nonlinear problems and must be indicated by a fractional number between 0 zero and 1 A smaller convergence is indicated when the number you enter has more decimal places for exampl
7. Inc All rights reserved Macromedia and Flash are trademarks of Macromedia Inc Contains Macromedia Shockwave Player software by Macromedia Inc Copyright 1995 1999 Macromedia Inc All rights reserved Macromedia and Shockwave are trademarks of Macromedia Inc Trademarks Following is a list of Trademarks or Registered Trademarks and the companies by which they are owned TK Solver for Windows TK Solver TK TK Solver Plus TK Plus TK Solver Roark amp Young Demo MiniTK Roark amp Young on TK Heat Transfer on TK Dynamics amp Vibration Analysis Gear Analysis Software DesignLink Queueing Analyses Tolerance Analysis TK CAD Link and MathLook by Universal Technical Systems Inc Windows Word Excel and Visual Basic by Microsoft Corporation Limitation of Liability Universal Technical Systems Inc UTS does not warrant that the software program will function properly in every hardware software environment The software may not work in combination with modified versions of the operating system with certain print spooling or file facility programs or with certain printers supplied by independent manufacturers Although UTS has tested the software and reviewed the documentation UTS makes no warranty or representation either expressed or implied with respect to this software or documentation including their quality performance merchantability or fitness for a particular purpose As a result this software and
8. Part 2 Equation Solving Setting Up the Optimizer The Optimizer is based on the widely used Solver technology from Frontline Systems If you are familiar with Microsoft Excel you might know that it includes a version of Frontline s Solver as an add in and you may have become comfortable with that interface to Solver TK includes two different approaches to setting up the Optimizer One is very similar to the Excel Solver approach so Excel users feel comfortable right away We also include a second approach via a TK Solver Wizard which we hope ts easier to use for those of you who might be new to the concept The TK Solver Commands menu includes two selections related to the Optimizer Commands J Display Plot F7 Display Interactive Table Fa ek Solve E List Solve F10 Block Solve Setup Optimizer Run Optimizer Fil h Examine CtrH Fa 2 List Fill Ctrl L Put Values to Lists Get Values From Lists 20 Display Solution Time Ctrl M Setup Optimizer This starts the configuration process which can culminate with running the Optimizer or saving the settings for later The interface is similar to that used for configuring the Excel Solver Chapter 7 Solution Optimizer Run Optimizer This runs the Optimizer with the values last selected using current Variable Sheet inputs as initial guesses It is equivalent to using Setup Optimizer without changing
9. The Optimizer can also be used to backsolve the differential equations for an initial condition that results in a desired solution at a particular value of x To do that a rule is created which maps the initial conditions to variables The PLACE function performs this task placing the values of yi and zi into the first elements of the associated lists 7 33 Part 2 Equation Solving Rules Status Rule place y 1 yi place z 11 2 These rules must be inserted above the ODE_RK4 function call Otherwise they will not be applied until after the integration has been completed using the prior values The initial conditions are input on the Variable Sheet El Variables OF x 1 b k OOo ooo p emes 1 8 m m Try changing zi from 1 to 1 2 The effect can be seen on the plot The z function is raised 107 axpi4 5 x 2 y zZ x 1fy a a a cr 3 eee eee ee he x x and z EEEE WAA ee ee eee oe ees ee ee ee ee The Optimizer can be used to determine the value of zi which forces the y and z curves to intersect atx 1 The target variable is set as d with a target value of 0 again This time zi is used as the decision variable with bounds of 1 and 1 5 Here is the solution 7 34 Solution Optimizer Chapter 7 ur A 22 d m comment 966057876 2 yi Mame Status 2565057576 21 1 370631733 The plo
10. anything The keystroke shortcut is F11 Optimization Parameters q Set Target Variable IT 4 Equal Ta Maximize Minimize Value of Cancel By Changing Variables L D P1 m Bounds Subject to the Constraints pi Add Optimize r gt 1 01 Options Change 2 Reset All Delete FRE Help 1228 5 m The Optimizer works with several different types of variables and it is important to note how it interacts with them The Target Variable is the TK variable that you want to maximize minimize or set equal to a value Any TK variable that does not have an input value can be set as the target variable A target variable is sometimes referred to as a dependent variable or an objective variable The target can be any valid TK expression that produces a numeric value If the target value in a TK model is actually the element of a list you can create a rule equating the list element with a variable before using the Optimizer For example if you want to use the 100 element of the list named temperature as your target value you could create the following rule on the Rule Sheet and use variable tn as your target variable tn temperature 100 Change Variables are those variables that you want the Optimizer to try changing in order to achieve your goal for the target variable A change variable must be an input variable in the TK model It cannot have a guess status Change var
11. way to design components or allocate scarce resources The resources may be raw materials machine time or people time money or anything else in limited supply The best or optimal solution may be maximizing profits minimizing costs or achieving the best possible quality An almost infinite variety of problems can be tackled this way The Optimizer significantly broadens the domain of problems you can solve with TK Students and professionals in all areas of science engineering finance operations research and management will find this to be a valuable time saving tool Here are a few problems TK can now tackle with ease Finance and Investment Working capital management involves allocating cash to different purposes accounts receivable inventory etc across multiple time periods to maximize interest earnings Capital budgeting involves allocating funds to projects that initially consume cash but later generate cash to maximize a firm s return on capital Portfolio optimization creating efficient portfolios involves allocating funds to stocks or bonds to maximize return for a given level of risk or to minimize risk for a target rate of return Manufacturing Job shop scheduling involves allocating time for work orders on different types of production equipment to minimize delivery time or maximize equipment utilization Blending of petroleum products ores animal feed etc involves allocating and combin
12. with lower bounds set as 0 25 and 50 Upper bounds are set as 100 60 and 100 These bounds are based on the assumption that Weight is 100 Here are the constraints Weight 100 TS gt 12 E gt 16 H gt 25 H lt 35 The Optimizer F11 returns the following solution El Variables Z x lk 7 L This example model is located in the Optimization section of the TK Library in the Optimizer Examples group 7 23 Part 2 Equation Solving Portfolio Optimization The Scenario Approach Given a collection of scenarios and the probabilities associated with them we can determine the configuration resulting in the optimal outcome This approach can be used in a variety of situations including investment portfolio management This TK table summarizes seven scenarios for a portfolio of three investments The values ee e are the expected decimal fraction returns over a 071 144 4588 the next 12 months For example scenario 3 predicts that the second investment will return 32 1 over the next year 54 1065 305 32 s o ss om s ns m ns LIST P Oy xl The probabilities associated with each of these scenarios are contained in the following list The expected return from each of these scenarios can be computed if we input the fraction we invest in each of the three equities We can expand our input table to include these fractions
13. 1 Since the default precision and comparison tolerance values are around 1E 6 it can be helpful to try to keep absolute values between 1E 6 and 1E6 but closer to 1 is always better If the target variables the change variables or any of the variables involved in the constraints are relatively small or relatively large it might be causing the Optimizer to work harder to reach a solution On rare occasions the Optimizer might even overlook the solution altogether Using the pressure units example above it s possible that your equations are given in units of Pa The resulting pcrit value would be 60000000 Pa By converting to MPa the pcrit value becomes 60 and the Optimizer can work with it more efficiently For more on computational accuracy see the Technical Time Out starting on page 5 13 7 19 Part 2 Equation Solving Optimization vvith Lists and Tables TK Solver includes a flexible programming language for processing lists and matrices Values in lists can be used as inputs or outputs in optimization problems by mapping them with variables on the Variable Sheet Change Variables A change variable used in an optimization process can be placed into a list using the PLACE function For example the following rule can be used to place the value of the variable named throttle into the 10 element of the list named xyz PLACE xyz 10 throttle Target Variables A target variable can receive a value from a list direct
14. 2 Equation Solving These equations can be entered into TK Solver to establish the objective function for the optimization The TK Solver Variable Sheet is shown next El Variables Status o x utput Comment 5 Blending Gasoline Butane vapor Pressure igin Haphta vapor Pressure CYC Gasoline Vapor Pressure Butane Octane 7 Airgin Haphta Octane CC Gasoline Octane 7 Profit per unit of Super IN Butane in Super 7 Argin Haphta in Super Gasoline in Super Profit per unit of Regular 575 Butane in Regular 577 Argin Haphta in Regular 7 Gasoline in Regular Units of Butane 7 Units of Hain Haphta Units of CC Gasoline Reid vapor Pressure of Super 9 fo Reid vapor Pressure of Regular wo Octane of Super Total Units of Super 5 Total Units of Regular Total Profit With inputs provided for each component and with values of 1 input for each of the six sample quantities we see the results at the bottom of the sheet Our goal is to determine the values of these six quantities that produce a maximum value for F and which satisfy the vapor pressure and octane rating requirements for the blends To accomplish this we can use the TK Optimizer 7 8 Chapter 7 Solution Optimizer We launch the Optimizer Wizard and define the target variable and constraints e Total profit F is selected as the objective variable the same as a target variable Quant
15. Conjugate requires less memory than the Newton method but typically needs more iterations to reach a particular level of accuracy Use this option when you have a large problem and memory usage is a concern or when stepping through iterations reveals slow progress Units and the Optimizer The Optimizer works with variables in their calculation units This is done to be consistent with the rules on the rule sheet and makes TK based reports easier to understand It also means that users can subsequently change the display units for any variables without worrying about affecting any Optimizer settings You may want to set all the variables directly involved in the optimization to their calculation units before initially setting the optimization constraints and bounds This will make it easier to make your optimization entries and to compare values with the bounds and constraints Use the Properties dialog to quickly view the calculation units for each variable of interest Users of existing TK models should be careful in preparing new Optimization settings just as they must be careful when adding new rules to a model In both situations it is vital that the calculations are done in consistent units 7 18 Chapter 7 Solution Optimizer Units and Improved Performance and Accuracy You can use units and conversions to improve the performance of the Optimizer Try to use display units that cause values of variables to be relatively close to
16. This excerpt is taken from the TK Solver 5 0 User Manual to give you a better idea of how the Solution Optimizer can help you accomplish more in less time The solution Optimizer included in the TK Solver 5 0 Premium Edition is a valuable tool that gives you advanced control over constraints and bounds to help you achieve optimal design parameters no programming required You ll like the easy to use VVizard driven system that readily walks you through the process and empowers you to take your design to a whole new level See what the Solution Optimizer can do for you today TK Solver 5 0 User s Guide Universal Technical Systems Inc Credits Page Copyright The UTS documentation and the software are copyrighted with all rights reserved Under the copyright laws neither the documentation nor the software may be copied photocopied reproduced translated or reduced to any electronic or machine readable form in whole or part without the prior written consent of UTS Copyright 1983 2013 Universal Technical Systems Inc Universal Technical Systems Inc 4053 North Perryville Road Loves Park Illinois 61111 Phone 815 963 2220 Fax 815 963 8884 All rights reserved First edition printed in August 1994 in the U S A Revised 2004 Contains images by Corel Corp 1600 Carling Ave Ottawa Ontario Canada K1Z 8R7 613 728 3733 Contains Macromedia Flash Player software by Macromedia Inc Copyright 1995 1999 Macromedia
17. d output values differ by several orders of magnitude Errors occurred in the TK model You may need to add bounds to avoid such error conditions You used guess status variables in the TK model Use default first guesses on the variable subsheet Optimizer reports that it is converging too slowly Try reducing the required precision settings One or more expressions result in 0 0 divisions which TK interprets as undefined When this happens TK skips the rule This can cause the optimization process to lose continuity and to behave erratically Add constraints to the model to force the Optimizer away from such conditions More Examples We conclude this section on the Optimizer with four case studies The first example demonstrates how to minimize cost The second solves a portfolio optimization problem involving the use of data tables The third solves a relatively difficult problem involving the solution of two simultaneous differential equations The fourth is a simpler problem showing how to store solutions in tables The Optimization folder in the Mathematics section of the TK Library includes models associated with these examples and more 7 21 Part 2 Equation Solving Minimize Cost Example A company produces rubber used for tires by combining three ingredients rubber oil and carbon black The cost of rubber is 0 04 per pound The cost of oil is 0 01 per pound The cost of carbon black is 0 07 per pound Here are the eq
18. documentation are licensed as is and you the licensee are assuming the entire risk as to their quality and performance In no event will UTS be held liable for direct indirect special incidental or consequential damages arising out of the use or inability to use the software or documentation even if advised of the possibility of such damages In particular UTS is not responsible for any costs including but not limited to those incurred as a result of lost profits or revenue loss of use of the computer program loss of data the cost of substitute programs claims by third parties or for other similar costs TK5 090904 Solution Optimizer 7 Solution Optimizer is included with the Premium version of TK Solver If you do not have this feature and would like to order it please call UTS at 815 963 2220 The purpose of the Optimizer is to find the set of input values that would provide the preferred solution to the current model The user can set preference criteria and limitations to be imposed in the seatch The Optimizer can help solve linear nonlinear and integer programming problems It can also be used as an alternative to TK Solver s built in Iterative Solver for solving systems of equations In this use the Optimizer offers more control over constraints and bounds on variables The Optimizer acts like a shell around a TK Solver model It calls the model repeatedly starting with the current values until the constraints are sa
19. e 0 0001 is less relative change than 0 01 The smaller the convergence value the more time TK takes to reach a solution Part 2 Equation Solving Assume Non Negative is an option you may want to use frequently It saves you the trouble of setting bounds and or constraints on all variables that you want to assume ate greater than or equal to zero Use Automatic Scaling Select this to use automatic scaling when inputs and outputs have large differences in magnitude Estimates This specifies the approach used to obtain initial estimates of the basic variables in each one dimensional search The Tangent option uses linear extrapolation from a tangent vector Quadratic uses quadratic extrapolation which can improve the results for highly nonlinear problems Derivatives This specifies the differencing used to estimate partial derivatives of the objective and constraint functions Use the Forward option for most problems in which the constraint values change relatively slowly Central should be used for problems in which the constraints change rapidly especially near the limits Although this option requires more calculations it might help when TK returns a message that it could not improve the solution Search This specifies the algorithm that is used at each iteration to determine the direction to search Newton uses a quasi Newton method that typically requires more memory but fewer iterations than the Conjugate gradient method
20. ed We assume that the fractions must total 1 and that the minimum return is 15 Bounds are also placed on the three change variables forcing them to be equal to or greater than zero That is short selling is not considered Bounded ariables x EH Change Variable al KI 0 Continuous Continuous Continuous Cancel Help After a few iterations the solution is displayed on the Variable Sheet You might investigate what happens as the minimum expected return is changed as well as what happens if the bounds are removed El Variables OF x a Unt Comment 583633987 F 057 Fraction of portfolio in equality 1 ABB2BB 3 F2 Fraction of portfolio in equality 2 61068 3 Fraction of portfolio in equality 3 00 EE 1 Expected return am Variance 7 28 Chapter 7 Solution Optimizer Differential Equations and the Optimizer This is a very rigorous exercising of many different TK Solver features starting with the solution of differential equations and concluding with use of the Optimizer to backsolve the problem Suppose one needs to solve the following differential equations for x 0 1 0 2 1 6 with initial conditions y 0 1 0 2 and z 0 1 1 0 2 5 7x y 10 e 2 Y 1 Z x Y TK Solver includes several built in functions for solving differential equations numerically They are all used similarly For this example the ODE_RK4 function is used R
21. er CIC Gasoline in Super Profit per unit of Regular Butane in Regar Virgin Nephts in Regular CIC Gasoline n Regular os 1139377339999 Butane Vapor Pressure irqin Manhhis Pres Units oF Virgin Maphita Cit Gasoline vapor Pres Units of C C Gasoline Butane Ottana Reid vapor Pressure virnin Octane Reid vapor Pressure cf Cit Gasoline Octane 15 Octane oF Super Profit per unit of Super Octane of Ragular mi m ly ir iF La r r r r Ww kil 1E4214X21 Reach a value az Maximize Al laz Minimize 7 10 Clicking the Finished button produces the solution quickly displayed on the TK Variable Sheet Status El Variables Chapter 7 Solution Optimizer o x utput Comment SS Butane vapor Pressure igin Haphta vapor Pressure CYC Gasoline Vapor Pressure 77 Butane Octane 7 Airgin Haphta Octane CYC Gasoline Octane 7 Profit per unit of Super IN Butane in Super 7 Argin Haphta in Super Gasoline in Super Profit per unit of Regular 575 Butane in Regular 577 Argin Haphta in Regular 7 Gasoline in Regular Units of Butane 7 Units of Hain Haphta Units of CC Gasoline Reid vapor Pressure of Super 706 Reid vapor Pressure of Regular Blending Gasoline th m om 7 11
22. f the equities such that the return is at least 15 with minimum variance 7 26 Chapter 7 Solution Optimizer The target value for this problem is the variance v The change values are the three fractions in the list F The Optimizer requires that the change values be declared as variables so additional rules are required The PLACE function is used to place values from the Variable Sheet into list elements El Rules OF x a Status Rule place F 1 F1 place F 2 F2 placet F 3 F3 call porti Ew It is important to note the sequence of the rules The port function assumes that the values have already been placed into the F list so the function call should follow the rules that perform that task The Variable Sheet is updated as shown below Yariables Miil E Unt Comment 12 ENT l ee Fraction of portfolio in equality 1 ee ee ae Fraction of portfolio in equality 2 BU dl Fraction of portfolio in equality 3 o EE rreri Expected return y Vqoastsi variance The Optimizer can now be set up to solve the problem Optimization Parameters x Set Target Variable lv Equal Tu Masimize Minimize Value of pone By Changing Variables F Fe Fa EA Bounds Subject to the Constraints Fl Fe Fi 1 E gt 2 015 Options Change Reset All Delete Help m 7 27 Part 2 Equation Solving Constraints are add
23. iables are sometimes called decision variables or independent variables 7 13 Part 2 Equation Solving If a change value of interest in a TK model is actually a list element you must create a rule placing the value of a variable from the Variable Sheet into the list element before using the Optimizer For example if the 49 element of the list named depth should be changed in the optimization process the following rule assigns the value of the variable named D to the list element and the Optimizer uses D as a change variable Place depth 4 It is very important that you not use a TK variable with a guess status for a target value when you launch the Optimizer A guess value will only be used by the Optimizer on its first pass through the TK model and will not be available on subsequent passes If a guess value is required in the model use a first guess on the variable subsheet that guess will be triggered automatically on each pass Such a variable could be used as the target variable by the Optimizer Available Variables Whenever you see the ellipsis button Eli you may click on it for a list of variables to add to the current field The variables listed will meet the required criteria For example if you click on the ellipsis button for the Target Variable you will be shown a list of variables with output status on the TK Variable Sheet Likewise the ellipsis button for the change variables produces a list of variables that ha
24. in Super TS x11 x21 31 TR x12 x22 x32 For any blend the octane rating can be computed for Super and Regular using the following equations with OCT1 OCT2 and OCT3 the octane rating values of the three components OCT1 x11 OCT2 x21 OCT2 x31 OCTS TS OCTR OCT1 x12 ore OCT3 gt x82 The minimum octane numbers for Super and Regular are 92 and 87 respectively Another important criterion of the blends is the Reid Vapor Pressure RVP The following equations are used to compute the RVP for Super and Regular with RVP1 RVP2 and RVP3 the values from each of the components RVP x11 RVP2 x21 RVP3 31 RVPS TS RVPR RVP1 x12 RVP3 x32 1 If you do not see the target you had in mind or if you cannot select the variables you want to adjust re check in the Variable Sheet to be certain the target variable has an output value and the variables to be adjusted have input values 7 6 Chapter 7 Solution Optimizer The maximum for both RVPS and RVPR is 8 There are limitations on the availability of each of the components The equations for the total of each component used are as follows A1 x11 x12 A2 x21 x22 31 4282 This will allow us to compare Al A2 and with the available amounts The last equation defines the total profit from both blends where p1 and p2 are the profits per unit of production of Super and Regular respectively F p1 TS p2 TR Part
25. ing raw materials of different types and grades to meet demand while minimizing costs Cutting stock for lumber paper etc involves allocating space on large sheets or timbers to be cut into smaller pieces to meet demand while minimizing waste 7 2 Chapter 7 Solution Optimizer Distribution and Netvvorks Routing of goods natural gas electricity digital data etc involves allocating something to different paths through which it can move to various destinations to minimize costs or maximize throughput Loading of trucks rail cars etc involves allocating space in vehicles to items of different sizes so as to minimize wasted or unused space Scheduling of everything from workers to vehicles and meeting rooms involves allocating capacity to various tasks in order to meet demand while minimizing overall costs Anything you can think of modeling with TK Solver is now fair game for the Optimizer Just enter some equations and you re ready to gol Here are two examples followed by a summary of the commands and features and three case studies for additional practice 7 3 Part 2 Equation Solving 1 Trivial Example A simple example with known solution illustrates the basics of this tool We will find the rectangular quadrilateral of largest area that can be inscribed in a circle of fixed radius We begin by modeling the situation for a sample radius of unit value A model of this situation might be the fo
26. ities x11 x21 x31 x12 x22 and x32 are identified as decision variables with minimal values set to 0 Constraints are placed on the octane rating of Super and Regular variables OCTS and OCTR to be a minimum of 92 and 87 respectively e Constraints are placed on the resulting vapor pressures of both the Super and Octane blends In this case the maximums are 8 for both e Constraints are placed on the available units of each of the three components 25 units of Butane 40 units of Virgin Naphtha and 100 units of C C Gasoline e Constraints are added to indicate that at least some Super and some Regular must be produced That is x11 x21 x31 gt 0 1 and xl PX x gt 0 1 The Optimization Wizard screens are shown on the following pages along with the optimal solution Optimization Wizard Select your objective variable Unts of Butane Units of irgin Ma ibba Unts of Gasoline Faid Vapor Pressure of Super Fel Vapor Pressure of Regular Cactane of Super of Regular Total Units of Super Total Units of Regular Total Profit 2171711717 31 77171 1 1 Maxime Minimize C Raath a vals of 7 9 Part 2 Equation Solving Fr Optimization Wizard 12 Butane Pressure 0 1 S Viegin Naphte vapor Pressure CIC Gesoling Vapor Pressure Butane Oclane Virgin Haphta Octane CC Gasoline Octane Profit per unit of Super Butane in Super Virgin Haphta in Sup
27. l Cmain Clabor f O0 008 7 36 Chapter 7 Solution Optimizer After testing the model let us seek the ideal input values for a flow of 1 405 790 SCF hr proceeding as in the previous examples But this time we have constraints to record them we begin by clicking on the Add button This brings up a form for setting those relationships Enter a series of constraints by using the Add button in the form itself The final constraints set is 7 37 Part 2 Equation Solving Optimization Parameters x Set Target Yarable T DK Equal Tu Masimize Minimize Value of Cancel By Changing PO ey Bounds Subject to the Constraints Add Optimize Options Change xo Reset All il H Delete Help Optimizing we find the best parameters for that flow volume El Variables Z x a Unit Comment _ Piping Stations 35 K 1 052900 SCFihr Volumetric tow 1886206 L Distance between compressor 7 ess 0 Pine diameter 1 22 in Pine wall thickness aaa P ea Compressor discharge pressul 00 re ser rin Hp 1 055 47 hp Povver of each compressor 0 five 3828550 A Annual fixed cost of all compre 2937 sive Annual fuel cost 0 Cman 3557 Pyr Annual maintenance cost 0 cew 1184279 Mr Annual labor cost 7 14105340 gar Pipe installation amortization
28. llowing Rules Miil x Status Rule Simplest tkw Simplest model tor Optimizer demonstration Rectangle of maximum area inscribed in circle of given radius base 27 R cosdalphal height 2 R sindfalpha area base height ay Hi Yariahles L x Status Name Output Unit Comment 1 30 1887 a lam KAA these pram imi 10 sa irs 12 VVe can see that assigning different input values to the angle alpha gives different solutions for the variable area Our goal is to find the value of alpha that results in the maximum value of area The Optimizer first searches the Variable Sheet for outputs and provides a list of variables that can be declared as the target variabk If the model has not been solved and no outputs are displayed the Optimizer will not display any target variable options In this example the target variable is area 7 4 Chapter 7 Solution Optimizer We open the Optimizer Clicking Commands gt Setup Optimizer Commands Display Plot F7 Display Interactive Table Fa ek Solve F aR List Solve F10 Block Solve Setup Optimizer Run Optimizer Fill GY Examine Ctrl F9 2 List Fill Ctrl L Put Values to Lists Get Values From Lists 201 Display Solution Time The parameters for the optimization can then be set in the panel that opens Optimization Parameters x Set Target Variable farea m UK
29. ly on the Rule Sheet The rule below sets the value of the variable T equal to the value of the n element of the list abc T abc n It is important to note the difference between referencing list elements in this way and using the PLACE function Both are one way processes but in opposite directions Constraints Constraints can reference list elements directly For example you can add a constraint that specifies that the 4 element of the list r must be greater than or equal to 0 3 Add Edit Constraint xX ri gt 103 Cancel Add 7 20 Chapter 7 Solution Optimizer Troubleshooting The Optimizer might stop before reaching a solution for any of the following reasons 1 Z J 10 You interrupted the solution process by pressing Ctrl Break The maximum time or number of iterations was reached The Target Variable or Objective Function you specified is increasing or decreasing without limit For problems with integer constraints you need to decrease the Tolerance setting in the Options dialog box so that Optimizer can find a better integer solution For nonlinear problems you need to decrease the Convergence setting in the Options dialog box so that Optimizer can keep searching for a solution when the target variable value is changing slowly You need to select the Use Automatic Scaling check box in the Options dialog box because some input values are several orders of magnitude apart or input an
30. m jon aa pe s2 106 ao 1 5 108 s oa sh 7 24 Chapter 7 Solution Optimizer For each of the scenarios the expected return is computed as the dot product of the returns list and the list of the fraction invested In TK terms this is represented as r DOT s1 F Since we need to do this for each of the seven scenarios we can use a procedure function We create the function and call it port Here are the required statements Comment Portfolio evaluation 4 b statement Expected returns for each scenario m lengtht s for iz 1 tom m TE dotUF 0 next i These statements assume that a TK matrix s has been created The list s is simply a list of list names Element Value st 22 s3 4 Ta s5 Now we can call the port function from the Rule Sheet El Rules Miil E dh call port The resulting list of r values can be included in a table with the corresponding probabilities These are the probabilities and returns associated with each of the seven scenarios Note that the sum of the probabilities is 1 INT TABLE Probabilities Mi X Element e oo a ho oe e ls am 7 25 Part 2 Equation Solving It is now possible to compute the overall expected return multiplying the expected returns by the associated probabilities and summing the results This is another job for the DOT function The required statement is added to the port procedure function
31. r x on the Variable Sheet we can solve for y and 2 EF Ri m Biel E Unit Comment R KA x 0807 00 n 00 r rem We can also input a value for y and solve for x EA m 6 m un Biel Ei Ooo x lam Ty 00 7 32 Chapter 7 Solution Optimizer It is important to note that TK List Functions backsolve directly and always return the first solution they find In this case there are two solutions as can be seen from the solution plot It looks like there should be another solution somewhere near x 0 95 El Variables OF x unit Comment O s bk D T yY Vk 0 s AI TK Optimizer can be used to find this solution The Optimizer is configured with the Target Variable defined as y set equal to 1 and with the Change Variable set as x Using bounds of 0 8 and 1 for x the Optimizer displays the following solution El Variables OF x I lk The Optimizer can also be used to determine an intersection point for the y and z curves We enter a rule d y z to define the difference between the two functions We set the objective function as d with a target value of 0 Bounds are set on x of 0 1 and 0 3 and the Optimizer returns the following solution Variables OF x F ih Comment The second intersection point can be found by setting the bounds on x of 1 4 to 1 6 El Variables OF x 497261 936 491 34001 4
32. s for y and z The List Fill Command is used to fill the x list with values from 0 1 to 1 6 in steps of 0 1 The next step is to create the TK procedure function test2 which defines the set of simultaneous differential equations Here are the statements required statement 2 PE 10 expi 2 5 x 259011 yYR x 1411 The variable x is the independent variable The variable y represents the derivative with respect to the independent variable x The variable y represents the function with respect to x The expression y i represents the derivative of the 1th function with respect to x Likewise the expression y i represents the value of the ith function at the value of x The variables are mapped to the calling ODE_RK4 function as follows PROCEDURE test OF x Comment Simultaneous Differential Equations Parameter ariahlez Input Variables WoW Output variables b The input variables must be defined in the proper sequence in order for the built in ODE RK4 function to work properly The sequence must represent the derivative function and independent variable values respectively Because these are passed into the function they are declared as Input Variables 7 30 Chapter 7 Solution Optimizer The program is now ready to solve the problem Click the solve icon and the solution values are displayed in the table Here are the first 6 elements of the table Ga INT TABLE Solution
33. t confirms the solution 10 expb4 a x 27 3 I b b a a Sa xi m la r 7r 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ee ee ee 7 5 med 7 35 Part 2 Equation Solving Repeated Optimization Subject to Constraints This example is taken from Engineering Optimization Methods and Applications by G V Reklaitis A Ravindran and K M Ragsdell John Wiley amp Sons 1983 A natural gas pipeline transmission system is required to pump 100 1012 cu ft day 100 MMSCF day of natural gas over a distance of 600 miles Compressor stations are to be placed at equal distances The design variables are the pipe diameter zD the compressor discharge pressure P1 and the distance between stations L The optimum design should be such that the total annual cost of amortization and running of the pipeline T is minimized If this minimum cost depends on the flow it should be evaluated for a given set of values We create a TK model with the following equations MathLook OFX R 1 987 g P2 0 219 r 1 hp 27160 X 4 SEG R OSES L L Cfuel hi F Cfizegd om main 52 ik 5 54E6 7 1 L L A labor Pinstall 73600 DPA 2000 2 Pmatl 15 58 P T Pmatl 9E5 Pinstall Cfized Ctue
34. te SFr n im a n n or log linear scales
35. tisfied or the process fails It is important to distinguish the Optimizer from TK functions You cannot include a rule on the Rule Sheet to activate the optimizer TK does not call the Optimizer The reverse is actually true the entire TK model acts like a subroutine called by the Optimizer The Optimizer requires that the underlying TK model be solvable without error conditions over the domain of the variables in the optimization If a TK error is triggered the Optimizer is halted and the error message is displayed Additional bounds can be added to prevent such errors After such an error the Optimizer must be restarted It does not continue on from the point of the error It is recommended that a model be tested before running the Optimizer on it Such testing would automatically cause the outputs or error messages to appear Any TK Solver model can be used with the Optimizer There is no special programming required The Optimizer works with the existing variables and lists This means that all the TK models from the many TK based applications can be optimized Here are the steps required for using the Optimizer 1 Build a TK model that solves over the required domain 2 Set up the optimization constraints and bounds using either the Set Up Optimizer Command or the Optimizer Wizard 3 Press F11 to launch the Optimizer 7 1 Part 2 Equation Solving Optimize V hat An optimizer is a software tool that help users find the best
36. uations relating the composition with its characteristics Rules Oe x T5 125 0 0il 0 001 ol FU 17 0 55 Rubber 0 04 Oil 0 002 Rubber Lu He 34 0 1 Rubber 0 06 Oil 0 5 Carbon 0 001 Rubber Oil 0 005 Oil 2 0 001 Carbon The hardness must be between 25 and 35 the elasticity must be at least 16 The tensile strength must be at least 12 Here is the equation for the cost of the mixture 00 1 0 11 SRubber Rubber SCarbon Carbon Cost Another equation is added to compute the total pounds in the mixture Weight Oil Rubber Carbon Here is the Variable Sheet with some sample inputs and outputs El Variables OF x a Unit Comment ot Sil in the ix 40 Rubber Rubber in the mix 40 Carbon b Carhtninfhe mix ie 100 Total mix 3 3 1l fi m a s oltcostper pound 7 o rubber 5 uhber cast per pound o Carbon o O Carbon black cost per pound feast 45 5 Costipound of mixture Oo o S S o m Tensile strength E KU 27 Huastety 1316 H urdness Using the equations and variables above the model will now solve for the cost of the material for a given mixture The Optimizer can be set up to solve for the mixture producing the required qualities at the lowest cost 7 22 Chapter 7 Solution Optimizer The target variable is Cost and it should be minimized The change variables are Oil Rubber and Carbon
37. ules OF x call ODE RK 4aftest2 x The ODE_RK4 function requires three inputs The first represents the differential equations If there are simultaneous differential equations a TK procedure function must be used to define the set For this example a function called test2 is used to hold the equations The second input to ODE_RK4 is the reference to the list containing the values of the independent variable These are the values at which the differential equations are to be integrated The list x is used here The last input to ODE_RK4 is the name of the list containing the solution list names In this case we will have two solution lists because we have two differential equations The list Y will be used to store the two solution list names The next step is to create the list Y on the List Sheet and fill it with the solution list names The names y and z are used for the two solution lists ii Lists OR o master list for unknown functions Here are the contents of the list Y Comment master list for unknown functions b A table can be used to store the independent variable and solution values Tahles Iof Solution to test2 differential equations 7 29 Part 2 Equation Solving TABLE Solution OF x Here are the definitions of the content lists for the table Solution ea Heading The table is used to supply the values of the independent variable and the initial condition
38. ve inputs Bounds and Constraints Bounds are limits beyond which the Optimizer should not search Constraints are limitations set on the end result but the Optimizer may search beyond them in finding the optimal solution In other words it may be necessary for the Optimizer to get a bigger picture of the problem to better solve it Because of this important difference bounds and constraints are handled separately in the interface It is possible to set bounds for variables that may cause an error condition For example if you know that the model fails if the value of T exceeds 100 and T is one of your change variables you can set an upper bound on T You can also set bounds that are functions of existing variables and expressions but it is very important to note that these bounds are set based on the initial conditions of the model and do not dynamically update as the model is repeatedly solved Bounds should be numeric values or expressions involving only input variables or list elements that will not change in the solution process For example you might set a bound that Di must be less than Do restricting an inner diameter to be less than an outer diameter This allows you to change the input for Do and rerun the optimization without having to use the commands menu to change the bounds Optimizer can violate the bounds but will only do so as a last resort if it is unable to find its way to a solution within the bounds 7 14 Chapter 7
39. y edits and returns to the main window 7 15 Part 2 Equation Solving Adding Bounds To add bounds on any change variables click the Bounds button This opens the Bounded Variables dialog By default all the change variables are listed without bounds and declared as continuous type variables Other Type options are Integer and Binary you can change the type to either of those by entering an 1 or a b in that column Integer variables can only take on integer values Binary variables can only be 0 or 1 Bounds can be expressions based on other inputs in the model If t must be greater than Tcrit then Tcrit can be assigned as the lower bound You should not use change variables as bounds for other change variables The bound values are only evaluated at the beginning of the optimization process Bounded Variables xX m Change Variable L Type al 350 p 50000 55000 Continuous Cancel Help Optimize The Optimize button launches the Optimizer This can also be done using function key F11 It may be useful to return to the Variable Sheet before launching the Optimizer so you can watch the action 7 16 Chapter 7 Solution Optimizer Options Clicking the Options button opens the Options screen Optimizer Options xX Masimum time alowed li Seconds Assume Non Megative iterations allowed foo oo Use Automatic Scaling Precision b
Download Pdf Manuals
Related Search