xxii 16.16 16.17 16.18 16.19 16.20 16.21 16.22 17.1 17.2 17.3 17.4 17.5 17.6 17.7 17.8 18.1 18.2 18.3 18.4 18.5 18.6 18.7 18.8 Operations Research: Concepts, Problems & Solutions Crashing to minimum project éurati Project cost variation with crashing: Example 16.9 Crashing of activity A Crashing of activity C Float values and different time values of activity .. 2... , Network diagram of Example 16.12... 02+ +2 2.00.0. Single channel single phase system . . - Multi-channel single phase system Single channel multi-phase system. . Multi channel! multi phase system . . . - Multi-channel single queue system. . - Queuing system cost variation... . . « M/M/I system with limited capacity Multiple server queuing system model Steps in a simulation study Comparison of area of square and a sector Comparison of area of square and a sector Plot of the given polynomial function . . Markov Chain model of the random walk Queuing system simulation logic Single server queuing system ..... Queue and system size during the simulation . + 524 + 527 + 528 + 530 - 331 + 534 + 535 » 548 » 548 . 549 549 549 559 563 565 587 594 596. 597 598 603 604 6091 Operations Research Basics Operations Research (OR) is a collection of scientific mathematical methods to analyse systems and to arrive at solutions that enable proper and optimal decision making. The OR methods follow systematic and scientific approach to study and solve the problems. These methods are also called OR techniques. A decision making problem in real life is transformed to an equivalent mathematical form or mode] and then the model is used to solve the real life problem. The mathematical formulation of a problem in real life is thus the first and the foremost step in solving it by OR techniques. The solution to a problem depends largely on the mathematical model used for solving it. Because the techniques provide solution to the mathematical model of the real problem. The trueness of the mathematical model to represent the real life problem defines the fitness of the solution to the real situation. The OR techniques discussed in this text are based on the appropriate mathematical models to solve different decision making problems. 1.1 Emergence of OR The World War II created lot of pressure to have suitable solutions to provide supplies to the war front to maintain strategic presence in the war. The British forces started to realise the need for suitable solutions to manage the difficulties they faced while trying to maintain their supplies to the war front. It included food, water, clothes, medicine, arms and ammunition, fuel, man power substitutes, and even war strategies. Most of them involved much uncertainties and the impact of the implementation of solutions were also not known. The problems faced in the war front included the quantities that are economical to move from place to place, the timing and mode of shipments, the source to initiate the movements, the routes suitable for shipments, combined shipment quantities and returns, and so on, Expert mathematicians and scientists were called to study the problem situations and to suggest solutions. While solving the problems, they classified the problems in to different categories and the methods suitable to solve them were developed. After the World War II, the techniquesOperations Research: Concepts, Problems & Solutions 4 Jes the analysis of the effects of all available actions E “ional sub-system of an organization can effects on the performance of Systems approach: OR f on the system's outcome. Each functional su be modelled and synthesised to investigate the system asa whole. vo a Interdisciplinary nature: OR has the capability to integrate all disciplines tq. eather in the mathematical model and each diverse disciplines can be sepa. rately modelled to solve the problems. a Uncovering new problems: A solution by the OR techniques can lead to the dis. covery of the possibility of another problem to be solved. In the situations where the solutions seems unrealistic can lead to such conclusions and dis- covery, These characteristics make OR a unique problem solving methodology applica- ble to many decision making situations. 1.3. Problem Modelling The given real life problems will be of abstract nature that need to be converted to more realisable mathematical forms that are solvable by applying the suitable OR technique. This conversion process will define variable notations to various unknown parameters of the problem. The techniques applied to these mathematical models will provide the solutions to these models. The solutions are the set of values to the variables present in the problem that optimises the defined objectives and satisfies the constraint restrictions. Based on the solutions, one will be able to conclude what optimal conditions exist for the mathematical model. However, these conditions and conclusions have to be extended to the real world problem to say what are the meaning and applicability of the solutions to the problem situation. It is likely that a solution may become unrealistic representing the inadequacy of the developed model. The process of modelling will be revisited and model will be refined to go on to the next round of solutions. Fig. 1.1 gives the pictorial representation of these modelling and solution processes. Solid = Chaotic & abstract : {| Realword [s._Formulationof J cn problems 5 i Model s Validation of - Results Deduction of Real world Interpretin, conclusions [= J model conclusions Figure 1.1: Problem modelling steps in OROperations Research Basics 5 1.3.1 Modelling Principles Modelling of a mathematical form is an exercise that can misguide the analyst sion maker. A problem can be misinterpreted and modelled to a really ale mathematical form. The solution to this model will never give a suitable realisa ic solution. Some problems need to be modelled to its mathematical form of higher levels of complexity and to solve. At the same time there need not be any such model required to solve a problem. There are ten important principles to be followed while formulation of models in OR context. They are: 1. Do not build a complicated model when a simple one will suffice. 2. Beware of moulding a problem to fit the technique. 3. The deduction phase of modelling should be conducted rigorously. 4, Models should be validated prior to implementation. 5, A model should never be taken too literally. 6. A model should neither be pressed to do, nor for which it was never intended. 7, Beware of over selling a model. 8. Some of the primary benefits of modelling are associated with the process of developing the model. 9. A model cannot be any better than the information goes into it. 10. Models cannot replace decision makers. zed for failing to do, that These principles help analysts and decision makers to prepare suitable models to solve real life problems. 1.4 Some OR Models Some OR Models of interest to the industry are: 1. Linear Programming Problem (LPP): the most basic constrained model that helps to get resource utilization to achieve an optimum solution to a linear objective function. 2. Non-linear Programming Problem (NLP): problems with non-linear objective function or constraints or both comes under this category that combines dif- ferent techniques to solve the problems. Both constrained and unconstrained problems are part of it. 3, Integer Programming Problem (IPP): another form of the linear programming problem that have the restriction that the solution can only take integer values. 4. Transportation Problem (TP): the basic model to address the problem of shipping of same type of items from many sources to many destinations having different transportation costs along the available routes. 5. Assignment Problem (AP): assigning resources to tasks such that the best allocation that minimizes effort or maximizes returns is achieved. . Dynamic Programming (DP): method of solving problems that can be divided into stages. Suitable for problems of all kind and that fit into other models. a6 Operations Research: Concepts, Problems & Solutions, 7, Travelling Salesman’s Problem (TSP): getting the routing by connecting. different nodes and coming back to the original (starting) node at the least possible effort, time/cost. ot Bee 8. Network Problems: scheduling of activities in a network of activities that completes a project. ; - 9. Queuing Theory: solving the queuing system problems that seeks the answers to the questions related to the system capacities required to manage the service requirements. : ; : 10. Decision Theory: helping a decision maker in decision making under different states of the nature, ; ; 11. Game Theory: solving multi player games with the assumption of zero-sum, 12. Replacement Analysis: defining the optimal time period for economic use of equipment and deciding the optimal mode of replacement. There are some other models in Operations Research. This textbook will consider some of the above models and some varied models to solve problems with certain conditions and some models not listed above. 1.4.1 Applications of OR The applications of OR is not limited to industrial problems. The span of application extends to any system where the problem can be modelled into any one of the OR model. The area of application extends from strategic decision making to medical research and analysis. Some of the dominant fields of application and the corresponding models of OR classified under different fields are discussed here, Government Sector The governments have the focus on the welfare and well-being of the society and they have to have proper application of scientific methods on the functions mentioned in the following non-exhaustive list. # budgeting and budgetary control system design ‘* economic planning * housing and distribution system ‘* mining and natural resources problems ‘© power and fuel supply © defence systems and safety mea- * social security planning and sures health systems © water resources and supply sys- * town planning and transportation tems Production Management : The production systems have another focus based on the utilization of the costs incurred while the manufacturing activities are c: ae arried out. Some major focus of application of OR techniques are as follows,operations Research Basics 7 facilities and layout problems research and development e aggregate planning © project scheduling and manage- ¢ logistics and transportation net- ment work design * maintenance « sourcing « scheduling and sequencing ¢ personnel management Enterprise Management As an enterprise, the focus of the management functions are not limited to production alone. They need to focus on higher level decision making not limited to the following problem areas. site selection and product mix * man power planning © CVP analysis © project planning and management SWOT analysis, CBA and SCBA © quality control finance and fund management management information systems © auditing and accounting © stakeholder interests and dividend market research and competition policies management ‘© mergers, acquisitions, and foreign © product diversification investments ¢ purchasing and replacements local laws and tax rules 1.4.2 Software Solutions to OR Models The OR models can be solved by hand for relative less complex problems. As the number of variables and constraint situations increase the problems become complex and solution procedure by hand becomes tedious. Many spreadsheet applications ad software packages are available for easily solving such complex problems. The mathematical models can be prepared using any of the following general formats: AMPL, ECLiPSe-CLP, GEKKO, GAMS, LINDO, OPL, Mathematica, etc. The solution to the models can be achieved by the corresponding software systems such as: CPLEX, LINDO, Mathematica, CLP, Lp_solve, SCIP, MINOS, NPSOL, etc. Spreadsheet applications have inbuilt solvers to solve the OR models.2 Linear Programming Problems enue The learning objectives of this chapter are to: @ introduce the concepts of linear programming problems, their struc- ture, and requirements © provides the practical methods to model a linear programming prob- Jem @ introduce the steps to solve linear programming problems by graphical method and the interpretation of the results # providing the solutions to real life problems from the results Learning Outcomes The learners will be able to: demonstrate and prepare liner programming problems from the de- scription of a real life problem * solve linear programming problems by the graphical method © develop conclusions and put forward the conditions that can provide optimum solution to the real problems An optimization problem in general has an Objective Function, a set of Con- straints that define the utilization of the resources, and the decision variable decla- rations as seen in Section 1.2, The objective function is a mathematical expression which gives the overall outcome that the sets of decision variable values can provide. ‘The set of con: ints define the usage of the resources which when utilised at an optimal level will provide the optimum value of the objective function. The decision variables represent the level of different activities to be followed without violating the resource restrictions or requirements. LPP focuses on the10 Operations Research: Concepts, Problems & Solutions optimum allocation of the resources to different activities such that an overall objective from these activities is optimized. 2.1 General Structure A Linear Programming Problem (LPP) is one of the models in the broad class of optimization problems. Similar to every optimization problem, the LPP includes an objective function, a set of resource restrictions/requirements, and the variable type declarations. Each of the above will have: © Decision variables: a set of unknown quantities of interest in the real world problems, the values of which are being determined that help to achieve a well-defined objective. © Objective function coefficients: unit contributions offered by each decision variable into the objective. They define the functional relationship of variables in the objective function. © Technological coefficients: the set of coefficients of variables in the con- straints taken together in the form of a matrix. They define the functional relationship of variables in the constraints. © Resource availability/requirements: They provide the actual physical quan- tities that the activities will consume or the minimum level of resources or targets that the activities are supposed to maintain, 2.1.1 Requirements by the Problem A linear programming problem has to satisfy the following conditions to account it in that category. (i) The decision variables involved in the problem are non-negative. (ii) The objective function can be expressed as a linear function on these decision variables. iii) The operating rules governing the process that utilize the resources called constraints are also linear functions of the decision variables. The last two conditions give the name'linear programming problems to this class of problems. 2.1.2 Assumptions The major assumptions in a linear programming problem are: Proportionality This is a major requirement for every linear relation. Additivity The linear relations use the concept of additivity. The returns from the activities are assumed to be additive in nature, Similarly, the use of resources for different activities are also assumed to be additive.Linear Programming Problems 4 Continuity The decision variables are assumed to take any real value that satisfy the constraint relations. Certainty A solution, if exists, there is no uncertainty associated with it. Irrespec- tive of time the solution to a problem will remain same as long as there are no changes in any of the parameters of the problem. Finite Choices A solution to the problem if exists can come from a finite set of choices. This set can be either finite with only a few choices or with a larger choice set. Infinite choices for the solution leads to a situation of unboundedness of solutions where the modeller can continue with still better solutions and continuously improved objective function value. The given problem descriptions will be unstructured and most of the time in the form of a verbal model. This unstructured problem has to be first converted in to a structured mathematical form. Every such problem descriptions thus goes through the stage of modelling which converts the verbal models in to the appropriate mathematical model. Modelling stage of a problem defines the variables, the constraint expressions. the objective function expression(s), and the variable types. The following section discusses these steps in detail with examples to demonstrate the modelling process for the linear programming problems. 2.2. Steps in LPP Modelling ‘The modelling of a LPP involves identifying the variables, constraint relations, and the objective function expression. These are generally carried out through the following steps. Step I Identify the unknown quantities called decision variables to be determined that are associated with the problem and represent them in terms of algebraic symbols. Decision variables in most of the OR models are represented by: Xp Joly Qo M Step II Identify all the restrictions or requirements in the problem and express them as equations or inequalities called constraints that are linear functions on the decision variables. Step III Identify the objective or criterion to be achieved in the problem and represent it as a linear function of the decision variables. + Example 2.1 — LP Modelling Example, The following data pertains to the manufac- turing and selling of three models; A, B, and C, of kitchen accessory by M/s. XYZ Co. Each unit of model A requires 7 hours of labour and 4 kilograms of materials. Model B requires 3 hours of labour and 4 kilograms of materials per unit. Model C requires 6 hours of labour and 5 kilograms of materials per unit. The profits out of the sale of these models are %4, €2 and %3 respectively. The total supply of raw materials is restricted to 200 kg per day and the daily labour availability is only L50hrs. Formulate a LPP to maximize the profit by utilizing the resources to manufacture and sell the products A, B, and Cc .2 Operations Research: Concepts, Problems & Solutions «Solution 21 The first step of solution to the problem is to identify and define the decision variables, The problem here is associated with determining of the daily production of Model A, Model B. and Model C of kitchen accessories by M/s. xyz Co. Since the production quantity of these models are unknown and the question specifies the model should focus in finding out these unknown quantities. Therefore, the decision variable should correspond to the quantities of the models A, B, & C to be produced. Since there are three unknown quantities to be determined, there needs three decision variables. Since they have to be in the form; 7; j= 1, 2, ..., n, and have n= 3. The decision variables for the model become: 41, 2, and x3. Since the decision variables are identified, they have to be explicitly written as follows. Quantity of Model A produced per day:- x1 Quantity of Model B produced per da} Quantity of Model C produced per da} In the second step. the resources have to be identified. In the given problem, there are two resources, viz., labour hours and materials. The activities in the company can use these available resources to produce the three products. 2 3, such that x1, x2, 43 > 0 The level of activity are the decision variables. It means that the quantity of Model A to be produced represented by 1 says the level of production of Model A by using the resources. Since the resources are limited in quantity the available quantity in the company is the upper limit. For example, labour availability is specified as 150 hours. This available labour hours can be used to produce x; quantity of Model A. xp quantity of Model B, and x3 quantity of Model C. When the resources have upper limit, the usage of the resources should follow an empirical rule: Usage of resources < Availability By introducing this restriction in the given problem, it gives: Labour hours required for Model A for x; units = 7 x x, Labour hours required for Model B for x unit Labour hours required for Model C for x3 unit The total required labour hours become: 7x, + 3x) + 6x3 , which can never be greater than maximum availability, 150 hours, Mathematically, this can be expressed as: 7x; + 3x9 + 6x3 < 150 Similarly, for the material resource, the availability is 200 kg. This needs to be utilized in the following way. Materials required for Model A for x, units of Model A= 4 x x, Materials required for Model B for x2 units of Model B= Materials required for Model C for 13 units of Model C. The total required material quantity become, 4x; + 4x) +533 , which can never be greater than maximum availability, 200 kg. This gives the mathematical expression: 4x) + 4x + 5x5 < 200 Since the probl lem has only two resources to be utilized, we have a set of twoLinear Programming Problems 13 expressions in the constraint relation set as: Txy + 3x2 + 6x3 < 150 (2.1) 4x) + 4x2 + 5x3 < 200 (2.2) In the third step, the objective function has to be identified and expressed in terms of the decision variables. The problem specified the profit as the return out of the activities, It is specified as one unit of Model A gives %4 as the profit, Model B gives %2, and Model C gives 3 as the profit. Since we represent the unknown quantities of these models as x), x2, and x3, the profits from these models become: xp, 2x, and 3x3 respectively. ‘The total profit becomes: 4.x; + 2x7 +33. Since the objective is profit out of the activities, the sense of optimization of the objective function has to be maximization. ‘Therefore, the objective function can be expressed as: Maximize Z = 4x1 +2x7+3x3 Now to compete the model, we need to specify the type of the variable values. ‘The level activity on manufacturing the three models can be set to zero to represent there is no production of the specific model. However, there cannot be a situation of negative value to the decision variable. A physical interpretation to negative value to a decision variable cannot be given in certain situations. Here it arises. Either we can produce some quantity of a model or we can produce nothing of a model. Therefore, the value of the decision variable can either be positive or zero. This declaration has to be provided to complete the model definition. That is, M1, %2, 320. ‘The complete linear programming model of the problem becomes as follows. Maximize Z = 4x, + 2x2 + 3x3 subject to. 7.x; + 3x2 + 6x3 < 150 4x + 4x + 5x3 < 200 x. 320 Where, x):- the quantity of Model A produced, x2:- the quantity of Model B produced, and x3:- the quantity of Model C produced. . A linear programming model of a problem will be represented with the objective function first, then the constraint expressions, and then the type declaration of the variables. It should accompany the definition of the variables also. » Example 2.2 — LP Modelling Example. A manufacturer produces two types of assembled products Model P and Model Q. One unit of Model P requires 4 hours of grinding and 2 hours of finishing. Model Q requires 2 hours of grinding and 5 hours of finishing. There are two grinding machines available: each machine working for 40 hours a week. The finishing works are provided by three workers; each worker working for 40 hours a week. The profit from Model P is estimated as 2450 and that from Model Q is $500. All products manufactured can be sold. in the market in the same week. Prepare a linear programming problem model to maximize the weekly profit of the manufacturer from the models. Clearly mention all the assumptions used. .14 Operations Research: Concepts, Problems & Solutions «Solution 2.2 The problem here is to determine the quanti es of the two models of assemblies to be manufactured such that the total profit will be maximum, The solution to the problem can be obtained as follows. From the given information, the number of decision variables is two: say, x and x2. Let xy:- number of units of Model P produced per week and x2:- number of units of Model Q produced per week ‘The activities required for the manufacturing of these models are grinding and finishing. The available resource for grinding is two machines, each working for 40. hours a week. Therefore, total grinding hours available per week = 80 hours, For the finishing work. there are three workers working for 40 hours a week, Therefore, the total finishing hours available per week = 120 hours. The requirement of these resources to different models are different. Grinding hours required for the models are 4 hours and 2 hours, respectively. Total grinding hours for the two models together cannot be greater than the available 80 hours, Therefore, we can write the first resource constraint as: 4x, + 2x2 < 80 The finishing hours required for the models are 2 hours and 5 hours, respectively. The available finishing hours for both the models together is 120 hours. The corresponding resource constraint becomes: 2x1 + 5x2 < 120 Coming to the objective function, we have two unit profit values from the models as %450 and &500, respectively. Since the unit profit values from the models are specified, the sense of optimization becomes maximization. Therefore, the objective function can be written as: Maximize Z = 450x, + 500x2 The complete mathematical form of the problem becomes: Max. Z = 450x; + 500x2 Subject to. 4x, + 2x2 < 80 2x; + 5x2 < 120 x1,%2 20 where, x; represents the number of units of Model P to be produced and x2 repre- sents the number of units of Model Q to be produced. . + Example 2.3 M/s. Metallic Engineering Works have obtained a large contract for the supply of an alloy steel. The alloy needs three metals X, Y and Z. The minimum requirement of the metals per week would be: 15 units of X, 12 units of Y, and 20 units of Z. The metals are available from dealers who supply them in standardised boxes numbered B1, B2, and B3 respectively. Box B1 contains | unit of X, 2 units of Y and | unit of Z; Box B2 contains 3 units of X, 2 units of Y and 1 unit of Z, whereas box B3 contains 1 unit each of X and Y, and 5 units of Z. The cost of one box of type B1, B2 and B3 is respectively, 21000, 850, and 1400. Draft it as a LPP to determine the number of boxes of each kind to be bought per week so that the amount required to buy them is minimum. .Linear Programming Problems 15 The problem is related to buying of three different boxes of metals. The number decision variables in this problem is three. The resources are the different metals, which are needed to meet certain requirements. These requirements specify the minimum limit of the resource. The relation to be satisfied in this situation becomes: total utilization should be > the requirement. Per unit of the different boxes, the cost is given. Therefore, the objective is to find the quantity of different boxes to be bought that minimizes the total cost. «Solution 2.3 Decision variables be x), x2, & x3 that represent the number of boxes of different kind to be purchased. xj: number of boxes of type Bl, x2! number of boxes of type B2, and ‘x3- number of boxes of type B3 Three metals X, Y, and Z are required at minimum levels of 15, 12 and 20 units. These metals are available from three boxes, B1, B2, and B3. The combination of the metals in these boxes provide the total availability of metal X as: x1 + 3x + x3. Similarly the availability of metal ¥ is 2x; + 2xz + x3 and metal Z is xy + x2 + 5x3. Since the minimum requirements are specified as 15, 12, and 20, the constraint expressions have to be written as: M1 + 3x2 +43 2 15 2x, + 2xq + x3 2 12 xy + X2 + 5x3 2 20 ‘The cost of purchase of box B1 = 1000x1 The cost of purchase of box B2 = 850x2 The cost of purchase of box B3 = 1400x3 Therefore, the objective is to Minimize the total cost of purchase. That is, the objective function is: Minimize Z = 1000x + 850x2 + 1400x3 Since the decision variables represent the number of boxes, they cannot be negative. The complete model becomes: Minimize Z = 1000x; + 850x2 + 1400x3 x1 + 3x2 +.x3 215 2x; +2xy +43 2 12 x + x2 + 5x3 = 20 21, %2, x3 20 where, x:- number of boxes of type B1, x9:- number of boxes of type B2, and x3:- number of boxes of type B3. . «Example 2.4 A small company can take up three different types of small projects: P1, P2, and P3. Each project requires three resources to complete them. These resources as denoted as: RI - man power (hours), R2- machinery (hours), and R3 - Materials (ton). The resource requirements for the different projects differ. PI requires 160 hours of R1, 32 hours of R2, and 3 ton of R3. For P2, the requirement are: 240 hours, 40 hours and 2 ton respectively. For P3 they are16 Operations Research: Concepts, Problems & Solutions respectively. 120 hours. 80 hours and 5 ton. The resource availability of the company are: R1 - 4000 hours, R2 - 400 hours and 20 ton. The expected profits out of the projects are respectively %20,000. %25.000. and 22,000. Prepare a linear programming model that determines the number of Projects of each type to be selected which maximizes the total profit from all the projects together. «Solution 2.4 The given information can be summarised as in the following table. olution 2.4 The given information can be Se Resource Pl P2 P3__available RI (hrs.) 160 240 120 4000 R2 (hrs.) 32 40 80 400 R3 (ton) 3 2 5. 20 Profit from project ®) 20,000 25.000 22,000 The rows in the table are arranged to represent the constraint coefficients and the resource availability. The resource requirement for different projects can be read from the columns. The profit contribution from each project are available as the last row. The unknown quantities are the number of projects of each type to be selected by the company. Let x; be the number of projects of each type to be selected by the company; j= 1,2.3. The constraint inequalities become: 160x; + 240x2 + 120x3 < 4000 => RI 32x1 + 40x2 + 80x3 < 400 => R2 3x + 2x + 5x3 20 =R3 The objective is to maximize the profit form all the projects together. The objective function becomes: Maximize Z = 200002; + 25000x2 + 22000x3 ‘The complete mathematical representation of the problem becomes: Max. Z = 20000x; + 25000x2 + 22000x3 Subject to. 160) + 240x2 + 120x3 < 4000 32x + 40x, + 80x3 < 400 3x1 + 2xz + 5x3 < 20 My 22, x3 20 where, x; - the number of projects of type P1, x2 - the number of projects of type P2, and x3 - the number of projects of type P3. 2.3. Forms of LPP We have seen that a LPP has an objective function and a set of constraints that are expressed in terms of a set of unknown variables called decision variables. This can be represented in a compact matrix form as follows,Linear Programming Problems 7 Let us define the following matrices first. x — the set of 1 decision variables, xj; j= 1,2,...,n, as a column vector b — the column vector of m elements representing the RHS constants Z — the objective function with n terms A — the coefficient matrix (constraint coefficients: mxn); [aij], i = 1,2. ¢ — the row vector representing the objective function coefficients, [¢;]” 2.3.1 General Form ‘The general form of the problem can be represented as: 1. Objective function: Z = c1x) + c2x2 + +++ + Cnn 2. Constraints: (linear expressions either < type or > type or combinations of these types) QyyXy + 2X2 +++ + AInXn = By + danXn 2 b2 aX) + a72X2 + Gin X) + Bn2X2 + +++ + Ann = Bry 20;f=1,2,....0 2.3.2. Canonical Form LPP written in the form: 1, Objective function: Maximize Z = cyx) + Cox) +-+* + CnXn (z= ye 7 2. Constraints: ax + aaxr ts Faint SB; (Dairy < bis Vis 1,2,5+\m 4720; f=l2, If the problem contains inequalities in the > form, they have to be converted into < form by multiplication by a —ive sign on both sides. An objective function in the form “Minimize Z” has to be changed to “Maximize by the multiplication by a —ive sign throughout, wn 2.3.3 Matrix Form This is the generally used form of LPP for analysis purposes. Matrix notation of a LPP is: 1, Maximize/Maximize Z = ex (Obj. Fn.) 2. Subject to: Ax{, <,=}b (constraints) 3. andx>0 (non-negative restriction)18 Operations Research: Concepts, Problems & Solutions 2.3.4 Standard Form Solution methods of LPP other than the graphical method require this form of a linear programming problem. All inequalities in the constraint set will be converted to equations by introducing required slack/surplus variabl LPP in the standard form will be written in the form: 1, Objective function: Maximize/Minimize Z = )) ep) 2. Subject to: Vi=l.2.--- mj an The number of variables in the standard form will be more than the actual number of original decision variables. To convert inequalities into equations. add slack variables to left side of < constraints and subtract surplus variables from left side of the = constraints. E.g.. (a) 2x + 4x9 — x3 < 4 can be changed into = 2x + 4x) —13 + 51 =4 where, 5; is a slack variable (b) x; — 2x3 + 5x3 > 13 can be changed into => x) — 2x2 + 5x3 - 5) = 13 where. s; is a surplus variable The coefficients of these slack/surplus variables in the objective function will be zero. The following examples illustrate the different form of LPP and their equivalent other forms. = Example 2.5 Write down the different forms of the following LPP. Max. Z = 3x) + 4x2 + 63 Subject to: 2x; + x2 + 2x3 2 6, 3x; + 2x3 <8, 7x1 — 3x2 +523 29x), 22, 13208 Solution 2.5 The given problem is: Maximize. Z = 3x) + 4x2 + 623 subject to. 2x) + x2 + 2x3 >6 3x, + 2x3 <8 Tx, — 3x2 +5x3>9 My X2, 4320 It has different types of inequalities in the constraint set. Therefore, this gen- eral form can be converted to the canonical form and the standard form. The corresponding matrix forms can also be prepared. From the given details we have: ¢ = [3, 4,6] with the sense of optimization as maximization. In the canonical form, all the constraints have to be in the < type.Linoar Programming Problems 19 The first and the third constraints are given as > type. Therefore, they have to be converted to < type by multiplication by a -ive sign. ‘The problem gets converted to: Maximize Z = 3x) + 4x) + 6x3 subject to. - 2x) — x) — 2x3 < -6 3x, + 2x3 <8 7x) + 3x7 - 5x3 < -9 My X2,.%3 20 -2 -1 -2 In this form, the coefficient matrixA =| 3 0 2 |, decision variable vector, 7 3 -5 1 -6 x = |x|, right hand side column, b = | 8 |, m = 3, andn = 3. x3 -9 The corresponding matrix notation of the canonical form is: Maximize Z = ex subject to. Ax type constraints and one < type constraint. Therefore, two surplus variables and one slack variable has to be added to the problem. Let them be 1, 5, and 53 respectively, in the first, second, and the third constraints. The contribution of slack/surplus variables in the objective function has to be kept as zero. The standard form of the given problem with these additional variables becomes: Maximize. Z = 3x, + 4x2 + 6x5 + 051 + 05 + 053 subject to. 2x +2 + 2x5 — 51 + 0s + 053 = 6 3x) + Tx — 3x2 + 5x3 +05; +052 — 53 =9 My 2, HB SL, , 53 20. 2 + 2x3 +051 +57 +053 = 8 In the matrix form of this standard form of the problem becomes: Maximize Z = ex subjectto. Ax=b x20Operations Research: Concepts, Problems & Solutions 212-100 . 0 20 1 o|, 7 7-35 0 0 -1 20 : 7 right hand side column. b= [6 8 9} am «Example 2.6 A linear programming problem is defi 3. and n = 6. a ined as follows. Maximize Z = 2x) — 42 + 5¥3 subject to. x1 — 2X2 + 4x3 > 12 Byte — 211 ay #2 + 5x3 S17 7 Write down the standard form of the problem. Separate the matrices in the standard form of the problem. : «Solution 2.6 The given problem has three decision variables and three constraint inequalities and one constraint equation. We need to introduce slack/surplus vari- ables to the three inequality constraints only. Thus, a total of three additional variables required to form the standard form. Let them be; sj, 52, and s in the first three constraints, respectively. Introducing these variables into the problem, the total number of variables become six. The coefficients of these variables in the objective function have to be kept as zero. 5 and 5» are surplus variables and 53 is a slack variable. The fourth constraint does not require additional variable since it is an equation. The standard form of the problem becomes: Maximize Z = 2x; — x2 + 5x3 +05; +052 +055 subject to. xy — 2x9 + 4x3 — 1 + 05 +083 = 12 3x, +22 ay +05) — 52 +053 = 11 2 + x2 + 5x3 +05; + 052 + 53 = 17 Oxy +x, +45 +05) + 05 + 053 =7 u X1, %2, 33, $1, 52, 5320 In the standard form of the problem, we have the following matrices. 1-2 4 -1 0 0 3.1 -1 0 -1 0 11 5 0 0 3f 01 1 000 decision variable vector, x = [x 2 3 Sy 82 ss] right hand side column, b=[12 1 17 J] ,m=4,andn=6. ¢ = [2,—1,5,0,0, 0], the coefficient matrix A =Linear Programming Problems 24 When all the constraints are in the form of equations, there are no additional variables required to prepare the standard form of the problem. . . Example 2.7 Prepare the standard form of the following LPP Minimize Z = 5x, + 2x2 subjectto. 3x, +42 27 x +5x2 210 x42 s5 a1 220 ‘Separate the matrices in the standard form of the problem. . «Solution 2.7 The given problem has a minimization objective function. It needs to be transformed to maximization type by multiplying it by a -ive sign throughout. The problem has two variables and three inequality constraints. Two surplus variables and one slack variable are required in the problem to prepare the standard form. The standard form becomes: Maximize —Z = -Sxy - 2x) + 05, + 052 + 053 subject to. 3x txp— 5) +052 +053 =7 10 x +32 +05; +02 4+53=5 Xy + 5x2 + 05; — 52 + 053 Xt, Xr, Sis 52, 9320 The different matrices in the standard form of the problem are: b= (7, 10, 5),.e=65,-2.0,0,0,x=[n © a 2 31-1 0 0 15 0 -1 0 . 110 01 and A = 2.4 Solution of LPP In the general form of the LPP, we have the decision variable set represented by x which is an n-dimensional vector. A solution to the LPP is a set of values to the n-variables that satisfies one or more of the constraints. A solution needs only the constraints to be satisfied. However, there are different types of solutions to a problem. A solution in turn has certain properties which qualifies it to be acceptable or not. The following are the major types of solutions. 2.4.1 Feasible and Infeasible Solutions A solution has the property that it represents a vector in the n-dimensional space. If this vector is contained well within the space defined by the constraint planes, all the constraints can be satisfied by the solution. Such solutions that satisfy all thea, | 20 Operations Research: Concepts, Problems & Solutions constraints are called feasible solutions, It includes the non-negativity requirements also. As an example, assume the constraints in the example problem, 27. 3x 44) 27, 2y #52 2 10,.xy +ay < 5,and ay, 12 2 0. Let the constraints be named as C,, G and C3, respectively. | There are two variables in the problem. Therefore, a solution vector can be represented in the Cartesian coordinate system by the corresponding coordinates, Assume the X-axis represents the variable x and the Y-axis represents the variable We need to take the positive value region of these variables in the coordinate system. The first quadrant is the area where the non-negativity of both the variables get satisfied Any point in the first quadrant can be a solution to the problem. Let’s take the point (0. 0). the origin. Since the non-negativity is satisfied, we need to check other constraints at this point. It is clear that only the third constraint of this problem is satisfied at this point. The solution at this point is: x; = 0, x2 = 0 and all the constraints have the left sides as zeros. The first two constraints are violated by this solution while the third one is satisfied. When a solution violates any of the constraints, it is called an infeasible solution. Thus, even though the solution is non-negative. it is not a feasible solution. The solution represented by the origin is thus not qualified to be a feasible solution as some of the constraint inequalities are violated. By taking points from the first quadrant of the Cartesian coordinate system, we can identify which are feasible and which are infeasible solutions. Assume that we have solutions; (1, 1), (1,2), (2, 1), and (2,2) as some example points in the first quadrant. If we substitute the variable values at these points in to the constraints. we get the relations reduced to: 4 > 7, 6 > 10, and 2 < 5 at the first point, (1, 1). The first two constraints are violated. Therefore, (1, 1) is infeasible. At the second point. (1,2), we get the constraints reduced to: 5 > 7, 11 > 10, and 3 < 5. This point violates the first constraint and thus an infeasible solution. At point (2.1), we get: 7 > 7,7 2 10, and 3 < 5 which violates the second constraint. The point (2,2) gives the values of the relations as: 8 > 7, 12 > 10, and4 <5 which satisfies all the constraints. Thus (2, 2) qualifies to be a feasible solution to the constraint set and thus to the problem. Assume that we plot the constraints as equations and the half space defined by the inequality are marked. Then in a graph, the above discussion points can be plotted as in 2.1. The constraints are named as C1, C2, and C3. The feasible region is shown by the triangle ABC. All points coming inside this region will satisfy all the constraints. Therefore, the triangle ABC and all points inside it together represent the feasible region for the given problem. 2.4.2 Basic Solutions and Basic Variables Consider the example problem 2.7 and the standard form of the problem. Fig- ure 2.1 represents the feasible region defined by the constraints. We will consider the feasible point, (2,2), out of the four points considered in the figure and the discussion made in Section 2.4.1, page no. 21.Linear Programming Problems ss 3 8 Feasible half space of C: X Figure 2.1: Feasible solutions Considering the point (2,2), the constraint evaluations provide the inequalities 827,12 = 10, and 4 <5. All these conditions are true and we have defined the solution is feasible. However, if we examine the constraint Cl; 3x1 +.x2 > 7, it can be observed that it needs a surplus variable to keep the left- and right hand side values equal. In the standard form of the constraint, 3x; + x2 — 51 + 052 +053 = 7, the surplus variable s; has to assume a value 5. If we consider the other constraints also, we get the conditions on slack/surplus variables as follows. C2 => x + 5x2 + 0s; — sz + 0s; = 10, which follows that 59 = Qand C3 => xy + x2 +05; + 0s2 + 53 = 5 leading to 53 = 1. Now we can check the standard form of the constraints together. We have: 3xy + x2 — 5, +052 +053 =7 21 + 5x2 +05 — 52 +053 = 10 xy +X) +05) +052 +53 =5 My X2, Sty $2, 3 20 For a linearly independent set of three equations in five variables, we are able to solve for three variables. The remaining two variables need not be in the solution. A solution to the variable set that has the property that exactly m number of variables have non-zero values and the remaining n — m number of variables are not present in the solution or at value equal to zero is called a basic solution. In the selected example, the considered points in this discussion are not basic solutions. Consider a point, (1.5, 2.5) on the constraint Cl. The constraint becomes an equation without the help of a surplus variable leading to s; = 0. This is true for24 Operations Research: Concepts, Problems & Solutions any constraint, Every point on the constraint line will eliminate the s ack/surplus variable used in the constraint. Thus, if a solution lying on any of the con: ‘ints is selected, we can eliminate one of the additional variables. Here, we have five variables and three equations. To get a basic solution two variables have to be eliminated. It is possible on the point of intersection of two constraints that both, the constraints become equations and the corresponding two additional variables can be set to zero. The corner points A. B, and C formed by the constraints thus qualify to be basic solutions. At the corner A. we get the additional variables in the constraints C1 and C2 become zero. Leaving the other variables x), 2. and 53 to define the solution, At the comer B, the additional variables in the constraints C2 and C3 becomes zero and the solution set will contain x, x2, and s1. Similarly, at the corner point C, the additional variables in the constraints C1 and C3 becomes zero. The solution set will contain x1, x2. and sz, These three corner points represent basic solutions defined by the constraints of the problem. The non-zero variables that define the basic solution are called the basic vari- ables. The constraint equations can be rewritten in terms of the basic variables alone by eliminating the non-basic variables. This will not alter the basic solution and the properties of the basic solution. 2.4.3 Basic Feasible Solution Assume that we have a set of constraints as: x) + 72 2 2 and x) + 2x2 <5, The plot of these constraints gives a triangle shaped feasible region PQR as shown in Figure 2.2. Any point inside this triangular region will satisfy both the constraints. A corner point defined by the intersection of the constraints, R will give a basic solution, The slack/surplus variable required in the constraints become zero and the variables x; and x2 will provide solution. However, the solution defined at this corner point is (~1,3). This solution is basic not feasible. A solution which is both basic and feasible is called a basic feasible solution. Then the feasible region defined by the constraints have to be the quadrilateral PQR,R2 which is limited by the Y- axis to avoid x1 < 0, Now we can check whether the corner points of this feasible region show basic solutions. Assume that the additional variables required in the constraints be sy and 52. Let's take the point P. This point is on the X-axis (x-axis). Therefore, the variable 42 gets a value 0. Since this solution is on the second constraint, the additional variable required in the second constraint becomes zero; s = 0. Therefore, the point P represents a situation that two variables are at zero level and the other two variables exists: a basic feasible solution, Similarly, the corner point Q is on the first constraint and consequently s; = 0. Since it is on the X-axis (x-axis), x2 = 0. Therefore, Q also represents a basic feasible solution. On the corner points R; and R» we get the conditions that the variable x; = 0 and one of the additional variables is also zero. On the corner Ry, we get x1 = 0 and $1 = 0, and on the comer point Ro, we get x = 0 and s; = 0. Therefore, we have seen that the comer points on the feasible region represent basic feasible solutions. Alll other solutions are feasible solutions but not basic. All corner points definedLinear Programming Problems 25 AL i T T Figure 2.2: Basic Feasible solution (BFS) by the constraints are not feasible solutions. That is why some of the corner point (basic) solutions are not coming under the basic feasible category as illustrated by the corner R in Figure 2.2. In Figure 2.1, the corner points A, B, and C thus represent the basic feasible solutions. The X and Y axes restrict the corner points with negative x2 and x; values, respectively, becoming basic feasible solutions. « Example 2.8 Check whether the following constraint set can provide a basic feasible solution. x; + x2 <2,x1 +22 25, and 41,» 20 . =Solution 2.8 We can introduce slack/surplus variables into the system of constraints and make them equations. The constraints become: xy $2 + 51 +05, =2 x +2x7 +05) - 2 =5 My X2y Sty 2220 Now, n = 4 and m = 2. Two variables need to be there in the solution and the other two to be zero. Into this standard form, we can assume different pairs of variables to be in the solution. Let’s take x = [0, 0, si, 2)". That is, we consider only the additional variables in the solution. This will provide: 51 +05, =2 Os - 2 =5 S1, 5220 The solution becomes: s; = 2, 52 = —5. This is a basic but an infeasible solution.is Research: Concepts, Probl 26 Operation: PI lems & Solutions Now take: x = [w. a2. 0, 01". This will lead (0 the reduced form of tt m7 ~ he constraint equations as: mee +2 M2 r equations become: The solution to this system of linear equa me: x1 = ~Land ay = “This ig also nota basic feasible solution. A quick observation is that the constraings do not meet in the first quadrant. Comer point basic solution is there but not feasible. Tak xy. 0, 81. 0)", The constraint equations become: x5 =2 x +05, =5 x, 5120 The solution i 5, and 5) = ~3: This is also not a basic feasible solution, Take: x = [x1, 0, 0. 52)”. The constraint equations become: 2 mo =5 x1 +05; x, 5220 This provides: x; = 2, s2 = —3, which is also an infeasible basic solution. Now take: x = [0, x2, 51, 0]". This gives the constraint equations reduced to: yt =2 2x +051 =5 X, 1 20 The solution that we get from this setting is: x = 5/2 and s = -1/2. This basic solution is also not feasible. The final setting of the variables [0, x2, 0, s2]7. The constraint equation set becomes: 22 +0 =2 2x - 9 =5 3, 220 ‘The corresponding solution becomes: x) = 2 and sp = —1. This point is alse not qualified to be a basic feasible point. No comer point defined by these constraints qualifies to be a basic feasible solution. This shows that this constraint set cannot define a feasible region where @ basic solution can be obtained, The summary of this analysis can be tabulated 3 follows in Table 2.1, a?rr Linear Programming Problems 27 Table 2.1: Feasibility analysis of basic solutions 2 Basic variables Variables. |*" [2 | x fi x2 x2 2 si bn s 2 a I - 5 2 - me 3 - : : 5p 2 sy - 2 - 1/2 - S : 3 - a Remarks infeasible infeasible infeasible infeasible infeasible _ infeasible If the solution is feasible, we will say there exists a basic feasible solution defined by the constraint set. Else, the problem does not have a basic feasible solution. . Example 2.9 Check if the following constraint set can provide a feasible basic solution. 2x) + x2 + 3x3 <5, x1 — 2x2 + 2x3 <6, 41, 42, *3 20 . Solution 2.9 The constraint set in the standard form will have five variables. The fequired number of basic variables is two and the remaining three will be non-basic variables, The standard form of the constraint set is: 2x, + x2 + 3x3 + 5, +052 =5 xy — 2x2 + 243 +05) + 52 = 6 2, XZ, XB, St, 9220 ing different combinations of two basic variables, we get the follow- By consider ing summary information. Some variable combinations as the basic variables the s have negative entries. They represent infeasible solutions. While some solution: solutions. Table 2.2 shows the combinations of variables represent feasible basic summary of the solutions and their properties. ‘Table 2.2: Feasibility analysis of basic solutions 2(¥ for yes and N for No) [:] 1 Basic variable combinations (fl td EI al [ Variables al [* eo) [as 1 x 195 8 52 - - - ” 2s - + + HBF a a s 7 - 7: - 8 - 4 - 5 5 - oo. RB N Y N Y Y Feasible? N NN ¥ ON A quick inspection of the basic feasible variable reveal that the constraintSa 28 Operations Research: Concepts, Problems & Solutions intersections does not belong to the feasible space. 1; 2 0. However, in combination With the second slack variable, they provide basic feasible solutions. This means that the second constraint do not take part in providing feasible solutions at the meeting points with the axes. The constraints that become part of a basic feasible solution are called active constraints. Tn the given constraint set, the first constraint is active and the second constraint is not an active constraint. The origin remains as a feasible solution. 2.5 Solving a LPP Solving a linear programming problem is the process of getting a solution that helps to achieve the objective. In the generic form, the LPP has an objective function which needs to be optimized. The optimization can be either maximizing or minimizing the quantity represented by the objective function. The restrictions in selecting a solution is the set of constraints. We have seen in the preceding sections that the set of constraints will define a space where the solutions will be feasible. By satisfying the constraints, we have a feasible set of solutions. Within the set of these feasible solutions we can move.freely from point to point without violating any of the constraint restrictions. The selection of the solution in the feasible region can however influence the value of the objective function that the solution can provide. The task of the analyst or modeller is to select that movement within the feasible space which will help to achieve the objective. ‘The plane corresponding to the objective function has to be placed within the feasible space such that the value of the function is at its optimum. Solving a LPP focuses on the proper placement of the objective function within the feasible space such that it is at its optimum value. The linear programming problems can be solved by different methods. The forth coming sections and chapters will discuss the Graphical Method and the Simplex Method of solving LPP and the variants of the Simplex Method. In the preceding sections we have seen the feasibility and infeasibility of solu- tions, the basic feasibility, and basic and non-basic variables. Any solution which | is feasible has to come within the convex space defined by the constraints and the non-negativity conditions. A solution to the problem represents a point through which the objective function plane can pass through. Placing the objective function | ! at a point that optimizes the value of the objective function is the process of solving a linear programming problem. 2.6 Graphical Solution Graphical solution method is applicable to a two variable problem where the visualization of the constraints and the objective function are easier in a two axes | coordinates system. The two variables of the problem are represented by the two | axes of the Cartesian coordinate system. Since a linear programming problem has aLinear Programming Problems 29 the non-negativily restrictions on the decision variables, the positive portion of are considered. Therefore, the first quadrant of the Cartesian coordinate the ax system is used for the graphical solution. However, for illustration purposes and discussion on infeasibility, we may use more quadrants of the Cartesian coordinate system. The constraints of LPP arc in the form of inequalities. In the graph, a constraint can be plotted as an equation by a line. This line will divide the feasible space into two halves. One of these half-spaces will provide the area that satisfy the inequality. In Figure 2.1 the hatching provided along the constraints show the feasible directions that satisfy the inequality. All the three constraints C1, C2, and C3 have separate feasible half spaces. The area formed by the intersection of the feasible half-spaces represented by triangle ABC is a closed convex space. All points in this area represent feasible solutions that satisfy the constraints. The objective function in two variables can be plotted as another line by as- suming a value. The direction of movement of this line that optimize the value of this line can be determined and the line can be moved in that direction. Each step moved in that direction will improve the value of the objective function. However, the movement of the objective function cannot continue indefinitely. It is limited to be within the feasible area. Revisit the example problem 2.7 given in page no. 21. Let the objective function of the problem be: Minimize Z = 8x) + 422. The value of the function is unknown at the beginning. However, we can place the function inside the first quadrant by selecting a value for the objective function. From the given form of the objective function, Z = 8x; + 4x, the slope can be separated and the expression can be modified as x) = ~ —2xy. The slope of the line becomes -2. Therefore, any line with slope -2 will represent the objective function having a specific value. We are restricted to select the value of the function such that the line passes through the feasible convex space of Figure 2.1, ie., within the triangle ABC. In addition to the information in Figure 2.1, the objective function to the problem is added to the graph and reproduced as Figure 2.3. In this figure, three values are assumed for the objective function: Z = 16, Z = 24, and Z = 32, The corresponding plots have different properties with respect to the position in the feasible area. The positions of these lines are such that the value of the objective function reduces as its position nears the origin. The line corresponding to Z = 16 can be seen not passing through any point inside the feasible area. Out of the other two lines, the line Z = 24 is more towards the optimality since it is less than 32. Maintaining the slope of the line the objective function can be further moved towards the origin. This process will end up once the line just leaves the feasible area through the corner point A. As any further movement towards the origin will make the line move away from the feasible area. At the corner A, the position of the objective function is such that it is within the feasible area and possesses the minimum possible. The corner point variable values will satisfy all the constraints. This is the optimum position of the objective function and the corresponding coordinates of the corner are the optimum solution to the problem.a i Operations Research: Concepts, Problems & Solutions Figure 2.3: Graphical Representation of LPP: different objective function values Thus, the optimal solution to a linear programming problem will be well within the feasible region, satisfy all the constraints, and provide the optimum value to the objective function. We can also observe that this point represents a basic feasible solution. 2.6.1 Solution by Graphical Method To get the solution to the problem, we need to identify the corner point through which the objective function leaves the feasible area. If the objective was to maximize Z. the movement of the objective function has to be opposite to the direction that minimize the objective function. If we assume the objective was to Maximize Z = 8x, + 4x9, then the last point through the line leaves the feasible area will be the corner B. Once the corner is identified, the solution has to be obtained by solving the corresponding constraint equations. In the current problem, we have the corner A giving the most minimum value for the objective function. Here, the corner A is formed by the intersection of constraints C1 and C2. They are respectively, 3x1 + x2 2 7, and x + 5x2 = 10. Solving these constraints as equations, we get 21 = 23/14 and x2 = 25/14. The corresponding objective function value is equal to 292/14 = 20.86. If we consider the maximization objective in this problem with the same ob- Jective function, the optimum solution corresponds to the intersection point of the constraints C2 and C3. That is the comer point can be obtained by solving the Constraint equations corresponding to C2 and C3: x; + 5x2 > 10 and x; +22 $5- Solving them we get the solution x, = 15/4 and x) = 5/4. The objective function value at this corner is equal to 35,Linear Programming Problems at = Example 2.10 Solve the following LPP by graphical method. Maximize Z = Say + 3x2; subject to. xy + x2 2, x) + 2xy <5, and x, 1 >0 . Solution 2.10 The constraint set is similar to that in the example represented by Figure 2.2. In the feasible area in the figure, we need to introduce the objective function and the corner point through which the objective function leaves the feasible area has to be identified. The slope of the objective function is -5/3. A line that passes through the feasible region is introduced in Figure 2.4 to represent the objective function at Z = 15. A portion of the objective function is well within the feasible region. Therefore, the objective function can be moved further towards the corner point B to improve its value. The optimum solution to the problem can be obtained at this corner point. The solution to the decision variables can be obtained by solving the corresponding constraints. 3 % 2.5 Figure 2.4: Graphical solution: Optimal solution - Example 2.10 The corner point B is on the X-axis and the constraint equation x; + 2x2 = 5. This gives the solution to the system of equations as: xj = 5 and x3 = 0, leading to the optimum value of the objective function as: Z* = 25. If the objective function was to minimize the objective function, the movement of the function plot of Z has to move in the direction opposite to that used for maximization. With such a minimization objective, we have the corner D, which is the inter- section point of the first constraint and the Y-axis. Therefore, the solution becomes: x} = 0, x} = 2 providing Z* = 6. . + Example 2.11 Solve the following LPP graphically and comment on the solutionOperations Research: Concepts, Problems & Solutions 32 obtained. Max, 9 Z=3m1+5%2 si. = 3x, +422 S12 xsd x2 22 2xy + 3x 2 12 » Solution 2.11 The solution to the problem can be as follows in Fig. 25. The four constraints of the problem are plotted along with the objective function. The feasible area defined by the constraint set C1, C2, C3, and C4 is shown by the quadrilateral PQRS. a Unique + Optimal Solution Maximizing direction of Z cl “ x Figure 2.5: Graphical solution: Optimal solution - Example 2.11 The last corner point in the feasible region in the optimizing direction of Z is corer S. The solution at this corner point corresponds to the solution of the simultaneous equations representing the constraints C1 and C2 Solving the constraints: ~3x, + 4x. < 12 and x, < 4 we get the optimal solution Bs x, 7 4 and x; = 6 providing the maximum value of the objective functionLinear Programming Problems 33 2.7 Types of Solutions In the preceding section we have seen that the linear programming problems have a single solution which is optimal. Such solutions are called unique optimal solutions. However, there are other different situations to a problem. There may be a situation that the convex space defined by the constraint set has the property that the objective function is able to leave the space parallel to one of the constraints, leading to many points through which it can leave the feasible space. In certain cases, the constraint set does not define a unique convex area so that we will not be able to define a unique feasible area. By certain set of constraints we may be able to define a convex space but the objective function may not have a unique corner point or a side of the feasible space through which it can leave this space in its optimizing direction. In certain prob- lems, there may be an optimal corner point defined by more than two constraints. There the number of non-zero variables will become less than m. These situations are common in LPP because of improper data gathering and logic modelling or improper definitions of variables and constraint formation. Even if all of them are taken care of, some problems will lead to these situations because of its structure. The following sections will discuss all these situations with suitable examples. 2.7.1 Multiple Optimal Solutions As mentioned earlier, it may not be possible to define a unique point through which the objective function will leave the feasible region. This occurs when one of the constraints is having the same slope as that of the objective function and when this constraint is an active constraint. The following problem will explain the situation of multiple optimal solutions. « Example 2.12 A linear programming problem modelled to maximize profit from two products has the constraints: —x; + 2x» < 4, 4x, + 3x2 < 18, and x; <3. The variables x; and x2 represent the quantities of each products to be produced. Each of these products can provide unit profits of 20 and 15 respectively. Determine the optimum quantities of the products to be manufactured that maximizes the profits. « «Solution 2.12 The constraint set of the problem are reproduced here: =x $ 2x2 <4 4x) + 3x2 < 18 ms3 XM, 220 From the description of the problem, we can construct the objective function as: Maximize Z = 20x, + 15x. The constraint set is plotted as in Figure 2.6. The objective function is plotted for Z = 50. In the direction of maximizing the objective function, we can observe that there is no single corner point through which the objective function line can38 Operations Research: Concepts, Problems & Solutions leave the feasible region. Instead, there are many points through which Z can leave this region in the direction of optimizing the value of 2. The slope of the second constraint is equal to -4/3. The slope of the objective function is also -4/3. In this case. the objective function can meet the corners C and D of the feasible region together. Any further movement in the optimizing direction of Z will cause it to move outside the feasible region. When it coincides the segment CD. it is equivalent to say that the objective function is optimal over many points along the side CD of the feasible region. C and D are basic feasible points. All other points in between them along the line segment are feasible points, A Figure 2.6: Graphical solution: Multiple optimal solutions On the basic feasible points, the solution can be obtained by solving the con- straints. On the corner C, we have x, = 3. By substituting this value in to the second constraint, we get: 4 x 3 + 3x2 = 18 => x7 = 2. Thus (3, 2) is the corner C, giving the objective function value Z = 20 3 + 15x 2 = 90 Again, on the comer point D, the solution can be obtained by solving C1 and C2. The solution is: (24/11, 34/11). The value of the objective function; Z = 90. This value of the objective function can be achieved on all other points on the edge CD of the feasible region. Thus, there are infinite number of solutions that can give the same optimal value of the objective function. If the objective was to minimize the objective, this phenomenon would not happen. Because, the constraint parallel to the objective function will not restrict the movement of the objective function in its optimizing direction. . 2.7.2 Infeasible Solution This is the case of a certain type of problems that it is not possible to get @ feasible solution to the problem. In this situation the constraint set is unable toae Linear Programming Problems 35 define a unique convex space to search for an optimum. solution? Let’s assume the following problem to explain the graphical interpretation of this situation. Example 2.13 Solve the following LPP graphically and comment on the solution. Max. = 3x, - 2x) sat. xtmsl 4 nt zm24 x, 220 «Solution 2.13 The constraints and the objective function are plotted on the Carte- sian coordinate system as shown in Figure 2.7. The first constraint forms a feasible region shown by triangle ABC in the first quadrant. The second constraint forms another feasible region confined between the X; axis, line segment DE, and the X2- axis which extends towards infinity. Figure 2.7: Graphical solution: Infeasible solution space Both these constraints are forming separate feasible spaces where these con- straints will get satisfied individually. There is not a single point in the graph that can satisfy both the constraints together. Two independent non-intersecting regions are formed by the constraints. If we check the feasibility of the constraint set as discussed in Section 2.4.3 it can be observed that the selection of m = 2 number of basic variables in the standard form of this constraint set does not provide feasible solutions at any of the corners. This shows that this constraint set cannot form a convex region as the feasible region for the problem. .36 Operations Research: Concepts, Problems & Solutions 2.7.3. Unbounded Solution In certain problems. the constraint set may be able to define a fe However, the movement of an objective function in some direction may not be bounded. That is. the movement within the area in some direction will not get restricted by a constraint or corners. In such problems. the objective function improves in the direction of this unboundedness. Given any solution, it is possible to further improve the objective function without any limit. Such problems, in Which the improvement of the objective function value is not bounded are called problems with unbounded solutions. The same feasible region may however provide bounded solutions to other objective functions. « Example 2.14 Solve the following problem. | Maximize Z = x) +4%2 subjectto 4m -3x2 4 3x; + 5x2 215 <3 x1,%2 20 ‘ ‘What are the properties of the problem? . «Solution 2.14 The graphical representation of the problem is shown in Figure 2.8, | The constraints define a feasible unique region. The corners A, B, and C are basic feasible solutions to the problem. cee The movement of the objective function in the direction of improving its value is however, not restricted by a constraint? That is the feasible region is unbounded in the direction of improving the objective function. Let's assume the objective function was: Minimize Z = x, + 4x2. The direction of the objective function will become opposite to that with the maximizing objective. Figure 2.9 shows the graphical representation of the problem with the minimizing objective function. The objective function in this situation will approach the corner B as the last point before leaving the feasible region. Thus, the unbounded feasible region here provides an optimal solution because the direction of optimizing the objective function value is being restricted by the feasible region within its bounds. Constrains C1 and C2 define the optimal solution to the problem with this minimizing objective. i The optimal solution to the problem becomes: x; = 65/29 and x} = 48/29. The value of the objective function becomes: Z* = 257/29. . 2.7.4 Degenerate Solution A basic solution has the property that there are m number of basic variables with non-zero solution and (1 — m) number of variables having solution as Zero. There is a special situation where less than m number of non-zero variables define the solution in place of m number of basic variables. That is, the number of zero valued variables will be more than (n—)m. An example is discussed below. — |Linear Programming Problems 7 A Xa Figure 2.8: Graphical solution: Unbounded region with unbounded solution + Example 2.15 Solve the following problem: Maximize: Z = x; + 6x2 Subject to: «Solution 2.15 The given constraint set and the objective function are plotted in Figure 2.10. The feasible region defined by the constraint set is unique and marked by the tetrahedron ABCD. In this problem, there are three constraints; m = 3 and in the standard form, there are five variables; n = 5. Solving the constraint equation pairs, we get the38 Operations Research: Concepts, Problems & Solutions Xe / =_ 6 / : Un bounded region L , Optimal solution Figure 2.9: Graphical solution: Unbounded region with optimum solution following basic solutions. Taking CI and C2 together, the solution obtained i which is also a feasible solution. Taking C1 and C3 also the solution becomes (2, 2). Now with Cl and C3 also the basic solution becomes (2, 2), which thus satisfies all the constraints and becomes a basic feasible solution. Cl and X> axis gives the solution (0, 3). C2 and X2 axis gives the basic solution (0,1). C3 being horizontal, there is only one axis it meets. The corresponding solution is (0, 2). C1 and Xj axis provides the solution (6, 0) which is basic feasible. Out of these solutions, the corners represented by B(6, 0), C(2,2), and D(0, 1) along with the origin are basic feasible solutions. The other solutions obtained are not feasible. Except at the corner C, it can be observed that the other three basic solutions have exactly m = 3 non-zero variables and n — m = 2 non-basic variables. At the corer point C, it can be observed that three constraints become equations and therefore, the three additional variables become zero._ Linear Programming Problems 39 x Optimum solution; intersection point of three constraints" Figure 2.10: Graphical solution: Degenerate Solution, ‘The slope of the objective function and its optimizing direction are such that the feasible region defines an optimum solution at the comer C. Thus, the solution at corner C, (2,2) is basic, feasible, and optimal but only m — 1 = 2 variables are non-zero. Such solutions where the non-zero basic variables are less than m are called degenerate solutions. In the set of constraints, even if one of the constraints meeting at the corner C is removed, the problem will have the optimal solution same as that obtained now. In other words, one of the constraints is redundant in the given problem. . Summary This chapter discussed and demonstrated the following aspects by suitable examples. © The structure of an optimization problem in general: An optimization problem in general has an objective function, constraints and the variable type declaration. The objective function and the constraints are represented as linear functions of the decision variables. The decision variables are the unknown quantities in a problem that need to be determined without violating the constraints and by achieving the optimum value for the objective function. * LPP modelling steps from a problem situation: Define the decision variables by looking into the unknown quantities in the problem. Represent the objective function and the constraints as linear functions of the decision variables. Declare the type of variables involved in the problem.40 Operations Research: Concepts, Problems & Solutions © Different forms of LPP: | A LPP can be represented in its general form, canonical form, matrix form, and the standard form. © Solutions to LPP and the concept of solution vector: A problem having many decision variables. say 7, can assume many sets of values to the variables together. A value set to the variables represent a vector in n—dimensions. A solution set that satisfy all the constraint is feasible solution. If a solution forces two or more constraints to be equations, it becomes a basic feasible solution. © Graphical method of solving LPP and different types of solutions and their meaning: Prepare the half spaces defined by the constraint inequalities and check for a convex area defined by them. If such a unique space exists either bounded or unbounded, feasible solutions to the problem can exist. If no unique convex space defined by the constraints. no feasible solution exists. The corner points of the feasible space are basic feasible points. The point through which the objective function plot leaves the feasible space represents the optimum, solution to the problem. If the objective function leaves the feasibility through many points, it has multiple optimal solutions. If the movement of the objective function in the optimizing direction is not bounded by the. feasible space, it is said to have unbounded solutions. If the optimum point is defined by more than two constraints, it is a degenerate solution. In an n—variable problem, a degenerate solution represents a corner point defined by 1 + 1 or more number of constraints. If the problem has exactly n—number of constraints, there is no chance of a degenerate solution.Linear Programming Problems a Exercise problem 2-1: A tool Manufacturer produces two types of tools, say; A & B. Each unit of A gives %75/= as the profit while B gives 860/= as profit. The manufacturer wants to maximize the total profit out of the sale of these two tools. The question to be answered is how many of A and B to be produced. To produce each unit of A, 3 hours of cutting time and 5 hours of finishing time is required. Tool B requires 3 hour of cutting and 2 hours of finishing time per piece. The manufacturer has a total of 150 hours of cutting and 90 hours of finishing. Model the problem to find out the approximate number of tool A and tool B such that the total profit is maximized. Problem 2.2: A textile company manufacturers three types of clothes: A, B, & C. These clothes are prepared from three types of wool: Red, Green and Blue. One unit length of type A cloth needs 3 meters of Red wool and 4 meters of Blue wool. One unit length of type B cloth needs 2 meters of Red wool. 3 meters of Green wool, and 1 meter of Blue wool, One unit length of type C cloth needs 6 meters of, Green wool and 3 meters of Blue wool. The firm has a stock of only 10 meters of Red wool, 13 meters of Green wool and 18 meters of blue wool. It is assumed that the income obtained from one unit of type A cloth is 8400, of type B cloth is 2600 and type C cloth is 2500. Formulate a linear programming problem that determine how the firm should use the available material so as to maximize the income from the finished cloth. Problem 2.3: A manufacturer has the quantities a, and ay available of a certain commodity at two warehouses. He wants to send these quantities to his three retail stores where the requirements are bj, bs, and bs, respectively. It is assumed that ay + ay = by +b + bs. He knows that it costs ¢ij; (i = 1, 2; = 1. 2, 3) to transport one unit from i warehouse to ** store. He wants to to know the number of units to be transported from each warehouse to the different retail stores. Formulate a LPP to solve the problem. Problem 2.4: A company has two grades of inspectors, I and II, who are to be assigned for quality control inspection. It is required that at least 2000 pieces be inspected per 8 hour day. Grade I inspectors can check pieces at a rate of 50 per hour with an accuracy of 97%. Grade II inspectors can check pieces at a rate of 40 Per hour with an accuracy 95%. __ The wage rate of grade I inspectors is €150 per hour and that of grade II inspectors is $125 per hour. Each time an error is made by an inspector, the cost to the company is ten rupee. The company has available for the inspection job 10 grade I and 5 grade I] inspectors. Formulate the problem to minimize the total cost of inspection, Problem 2.5: A company has two plants, each of which produces and supplies two Products: A and B, The plants can each work up to 16 hours a day. In plant 1, it42 Operations Research: Concepts, Problems & Solutions takes three hours to prepare and pack 1,000 units of A and one hour to prepare ang pack one ton of B. In plant 2, it takes two hours to prepare and pack 1,000 units of A and 1.5 hours to prepare and pack one ton of B. In plant 1. it costs 815,000 to prepare and pack 1,000 units of A and 28.000 to prepare and pack one ton of B, whereas in plant 2 these costs are %18,000 and €26,000. respectively. The company is obliged to produce daily at least 10 thousand units of A and 8 tons of B Formulate this problem as an LP model to find out as to how the company should organize its production so that the required amounts of the two products be obtained at the minimum cost. Problem 2.6: An electronic company is engaged in the production of two compo- nents C1 and C2 that are used in radio sets. Each unit of C1 costs the company %5 in wages and &5 in material, while each of C2 costs the company %25 in wages and 815 in material. The company sells both products on one- period credit terms, but the company’s labour and material expenses must be paid in cash. The selling price of C1 is €30 per unit and of C2 it is €70 per unit. Because of the company’s strong monopoly in these components, it is assumed that the company can sell, at the prevailing prices, as many units as it produces. The company’s production capacity is, however, limited by two considerations. First, at the beginning of period 1, the company has an initial balance of %4,000 (cash plus bank credit plus collections from past credit sales). Second, the company has, in each period, 2,000 hours of machine time and 1,400 hours of assembly time. The production of each Cl requires 3 hours of machine time and 2 hours of assembly time, whereas the production of each C2 requires 2 hours of machine time and 3 hours of assembly time. Formulate this problem as an LP model so as to maximize the total profit to the company. Problem 2.7: A company engaged in producing tinned food has 300 trained employees on its rolls, each of whom can produce one can of food in a week. Due to the developing taste of public for this kind of food, the company plans to add to the existing labour force, by employing 150 people, in a phased manner, over the next five weeks. The newcomers would have to undergo a two-week training programme before being put to work. The training is to be given by employees from among the existing ones and it is a known fact that one employee can train three trainees. Assume that there would be no production from the trainers and the trainees during training period, as the training is off-the-job. However, the trainees would be remunerated at the rate of $300 per week, the same rate would apply as for the trainers. The company has booked the following orders to supply during the next five weeks: Week = 1 2 3 4 5 No. of cans: 280 298 305 360 400 Assume that the production in any week would not be more than the number of cans ordered for, so that every delivery of the food would be ‘fresh’. _ Formulate this problem as an LP model to develop a training schedule that minimizes the labour cost over the five-week period.Linear Programming Problems 43 problem 2.8: A patient requires two vitamins V1 and V2 that are available from two food items. The daily requirement of V1 is at least 60 units and of V2 is 80 units. ‘One unit of the food item Fl contains 3 units of V1 and 4 units of V2 while F2 contains 5 units of VI and 3 units of V2. The unit cost of Fl is €13/= and that of F2 is T15/= per unit. ‘The objective is to find out the combination of F1 and F2 that will provide the daily requirement of vitamins V1 and V2 at minimum possible cost. Formulate a LPP to solve the problem. Problem 2.9: Old hens can be bought for 125 each but young ones cost €245 each. The old hens lay 3 eggs per week and the young ones 6 eggs per week, each being worth %5. A hen costs Re. 18 per week to feed. If there is only €900 for hens, how many of each kind should be bought to give a profit more than 600 per week, assuming that there can be a maximum of 20 hens? Formulate a LPP model of this problem. Problem 2.10: A hospital has the following daily requirement of nurses. Period Clock Time Min. no, of nurses 1 6am- 10am 7 2 10am-2pm 8 3. 2pm-6 pm 16 4 6pm-10pm ul 5 10pm-2am 7 6 2am-6am 5 Nurses report to the hospital at the beginning of each period and work for 8 consecutive hours. The hospital wants to determine the minimum number of nurses available for each period. Formulate this as a LPP by setting up appropriate constraints and objective function. Problem 2.11: A person requires 10, 12, and 12 units of chemicals A, B, and C, respectively for his garden. A liquid product contains 5, 2, and 1 units of A, B, C per jar. A dry product contains 1, 2, and 4 units of A, B, and C per carton. The liquid product is sold for %3 per jar and the dry product is sold for %2 per carton. The person has problem of determining how many units of each product should be purchased, in order to minimise the cost and meet the requirement. Model the problem into the standard linear programming problem format. Problem 2.12: A paper mill produces two grades of paper, say, X and Y. Owing to raw material restrictions, the company cannot produce more than 400 tons of grade X and 300 tons of grade Y in a week. There are 10 production hours in a week. It requires 0.2 and 0.4 hours to produce a ton of products X and Y, respectively. They provide profits of £200 and %500 per ton, respectively. Formulate the above problem in the standard liner programming problem format which will find the optimum product mix that will maximize the profit.