Chapter 2 The Linear Programming Model Section 2.1 The Blending Model Problem: to determine how much of each of the ingredients to blend together so that the total cost of the mixture is minimized while the composition of the mixture satisfies specified requirements. Example 1 To feed his stock a farmer can purchase two kinds of feed. His requirement is (per day) 60 units of nutritional element A, 84 units of nutritional element B, 72 units of nutritional element C. The contents and cost of a pound of the two feeds are given in the following table. x y Feed 1 Feed 2 Nutritional Elements (units/lb) A B C 3 7 3 2 2 6 Cost (cent/lb) 10 4 The farmer wants to determine the least expensive way of providing an adequate diet by combining the two feeds. Formulate the mathematical model. Variables: x the number of pounds of Feed 1, y the number of pounds of Feed 2. Cost function f ( x, y ) = 10 x + 4 y Constraints: 3x + 2 y ³ 60 --nutritional elements A requirement 7 x + 2 y ³ 84 --nutritional elements B requirement 3x + 6 y ³ 72 --nutritional elements C requirement x, y ³ 0 . Þ Mathematical model: Minimize f ( x, y ) = 10 x + 4 y Subject to 3x + 2 y ³ 60 7 x + 2 y ³ 84 3x + 6 y ³ 72 x, y ³ 0 . Geometric solution The hatched region is the set of all diets that satisfy all nutritional requirements. The problem ® to determine the minimum of f ( x, y ) in the hatched region. Iso-cost lines: 10 x + 4 y = C C is any positive constant All iso-cost lines have 5 the same slope = - < 0 . 2 ¯ As C , the line moves to the left. The problem ® to seek the line farthest to the left and still intersects the hatched region. The line through the point (6, 21) is the line that we seeked. So, the minimal diet is: 6 lb of Feed 1, and 21 lb of Feed 2. The minimal cost is = 10 × 6 + 4 × 21 = 144 cents Example 2 A landscaper has on hand two grass seed blends. Blend I contains 60% bluegrass seed, 10% fescue, and costs 80$/lb; Blend II contains 20% bluegrass seed, 50% fescue, and costs 60$/lb. (Each also contains other types of seeds and inert materials). The field about to be sowed requires a composition of seeds consisting of at least 30% bluegrass and 26% fescue. What is the least expensive combination of the two blends that meets these requirements? Ambiguities can arise because the problem involves percentages. To avoid these, fix an amount of the final product. For example, let the number of pounds of the required composition be 100. Variables: x the number of pounds of Blend I , y the number of pounds of Blend II . Þ Mathematical model: Minimize the cost function C ( x, y ) = 0.8 x + 0.6 y Subject to 30% bluegrass requirement 26% fescue x < 1500 0.10 x + 0.50 y ³ 26 Þ x + 5 y ³ 260 x + y = 100 x, y ³ 0 . the direction for C to reduce The set of all points that satisfy the constraints is given graphically by the line segment between (25, 75) and (60, 40). The isocost line 0.8 x + 0.6 y = C has the slope - 4 < 0. 3 As C¯, the line moves to the left. Thus minimum of the cost function is (25, 75). That is, the minimum cost prescription is to use 25% Blend I and 75% Blend II. Section 2.2 The Production Model Problem: To determine a way of operating a system (operation or production system) that maximizes profit using limited resources or minimizes costs while meeting special production requirements. Example 1a Suppose a boat manufacturer produces two types of boats: a family rowboat and a sport canoe. The boats are molded from aluminum by means of a large pressing machine and are finished by hand labor. A rowboat requires: 50 lb of aluminum, 6 min of machine time, and 3 hr of finishing labor. A canoe requires: 30 lb of aluminum, 5 min of machine time, and 5 hr of finishing labor. For the next 3 months the company can commit up to 2000 lb of aluminum, 5 hr of machine time and 200 hr of labor for the manufacture of the two types of boats. The company realizes $50 profit on the sale of a rowboat and $60 profit on the sale of a canoe. Assuming that all boats made can be sold, how many of each type should be manufactured in the next 3 months in order to maximize profits? Variables: R -- the number of rowboats, C -- the number of canoes. Profit function: 50R + 60C. Þ Mathematical model: Maximize 50 R + 60C Subject to 50 R + 30C £ 2000 (aluminum limited) 6 R + 5C £ 300 (machine time limited) 3R + 5C £ 200 (finishing labor limited) R, C ³ 0 . Example 1b In the above example, the $50 and $60 profit estimates would be determined by subtracting production and delivery costs from the selling price of each of the two types of boats (that is, profit = revenue-cost). Suppose now that the cost to the manufacturer of the 1 ton of aluminum is not fixed. In particular, assume that the price per pound of the last 500 lb of aluminum is 20¢/lb more than the price of the first 1500 lb, and that the price of the first 1500 lb was the cost used in determining the $50 and $60 profit estimates. With this increment in cost of the last 500 lb of aluminum, what is the optimal production schedule? Variables: R the number of row boats, C the number of canoes, x the amount (in lb) of aluminum used over 1500 lb. Þ Mathematical model: Maximize 50 R + 60C - 0.2 x Subject to 50 R + 30C £ 1500 + x 6 R + 5C £ 300 3R + 5C £ 200 x £ 500 R, C , x ³ 0. Example 1C If the price of the last 500 lb of aluminum is 20¢ less than the price of the first 1500 lb, what is the optimal production schedule? Let R and C be as before, and let x ~ lb of aluminum bought at regular price; y ~ lb of aluminum bought at discount price (less $0.2/lb) Profit function: 50R + 60C + 0.2y (to be maximized) Constraints: 50R + 30C = x + y (aluminum used) 6R + 5C £ 300 3R + 5C £ 200 x £ 1500, y £ 500, R,C, x, y ³ 0. Also, y > 0 only if x = 1500. (1) How to realize the request (1)? We may introduce variable z asking that: if x = 1500 ì1 z=í (*) if x < 1500 î0 and y £ 500 z . But how to make z satisfy (*)? We may add two constraints: x z£ , and z is a 0−1 variable. 1500 Then (*) can be satisfied when we maximize the profit function (consider why it is true). Therefore, the model is: Max 50R + 60C + 0.2y Subject to 50R + 30C = x + y 6R + 5C £ 300 3R + 5C £ 200 x £ 1500 y £ 500z x z £ 1500 R, C, x, y ³ 0; 0 £ z £1 & integral. Example 2 Consider the operation of a division in a large plant. The division is responsible for manufacturing two parts of the plant’s final product. The division manager has available four different processes to produce these two parts; each process uses various amounts of labor and two raw materials. The inputs and outputs for 1 hr of each of the four processes are given in the following table. Run per hour of Process 1 2 3 4 Input Labor Raw Material A (man-hour) (lb) 20 160 30 100 10 200 25 75 1000 (+200) 8000 Raw Material B (lb) 30 35 60 80 4000 Output Units of Units of Part 1 Part 2 35 55 45 42 70 0 0 90 The division is responsible for producing each week 2100 units of Part 1, and 1800 units of Part 2. The division manager has at her disposal each week 4 tons (i.e.8000 lbs)of Raw Material A, 2 tons (i. e. 4000 lbs) of Raw Material B, 1000 man-hours of labor. One pound of Raw Material A costs the firm $3, and One pound of Raw Material B costs the firm $7. Because of labor contracts, the plant must pay its employees a full week’s salary, regardless of whether or not the employees are used that week, and so the cost of the 1000 man-hours of labor is fixed. However, the division manager can request her workers to work up to an extra 200 man-hours per week in overtime, at a cost of $8/hr to the firm. The plant vice-president in charge of production wants to know: 1. if the division can meet its weekly production requirements with the material on hand without using overtime and, if so, the minimal cost of this operation. 2. And, because the decision to allow overtime must be made at the plant level, the vice-president also wants to have some estimate on how much money, if any, the division can save by using overtime. First model: (overtime is not used) Variables: xi the number of hours a week that Process i is used, i = 1, 2, 3, 4. Þ The first model: lb of R.M. A Minimize the cost function f ( x1 , x 2 , x3 , x 4 ) = 3(160 x1 + 100 x 2 + 200 x3 + 75 x 4 ) + 7(30 x1 + 35 x 2 + 60 x3 + 80 x 4 ) lb of R.M. B = 690 x + 545 x + 1020 x + 785 x 1 2 3 4 Subject to 20 x1 + 30 x2 + 10 x3 + 25 x4 £ 1000 (labor) 160 x1 + 100 x2 + 200 x3 + 75 x4 £ 8000 (R.M .A) 30 x1 + 35 x2 + 60 x3 + 80 x4 £ 4000 (R.M .B) 35 x1 + 45 x2 + 70 x3 ³ 2100 (Pa rt 1) + 90 x4 ³ 1800 (Pa rt 2) 55 x1 + 42 x2 x1 , x2 , x3 , x4 ³ 0. Second model: (overtime can be used) Introduce a new variable: x5 -- the number of hours of overtime used Þ the second model Minimize g ( x1 , x2 , x3 , x4 , x5 ) = 8 x5 + f ( x1 , x2 , x3 , x4 ) = 8 x5 + 690 x1 + 545 x 2 + 1020 x3 + 785 x 4 Subject to 20 x1 + 30 x 2 + 10 x3 + 25 x 4 £ 1000 + x5 160 x1 + 100 x 2 + 200 x3 + 75 x 4 £ 8000 30 x1 + 35 x 2 + 60 x3 + 80 x 4 £ 4000 35 x1 + 45 x 2 + 70 x3 55 x1 + 42 x 2 x1 , x 2 , x3 , x 4 ³ 0, ³ 2100 + 90 x 4 ³ 1800 x5 £ 200. Section 2.3 The Transportation Model Problem To determine a shipping schedule that minimizes transportation costs. Assume that the costs of shipping goods from a source to a destination are directly proportional to the amount of goods shipped. Example A paper manufacturer having two mills must supply weekly three printing plants with newsprint. Mill 1 produces 350 tons of newsprint a week and Mill 2, 550 tons. Plant 1 requires 300 tons/week, Plant 2, 400 tons, and Plant 3, 200 tons. The shipping cost, in dollars per ton, are given in the following table. Paper Mills 1 2 1 17 18 Plants 2 22 16 3 15 12 The problem is to determine how many tons each mill should ship to each plant so that the total transportation cost is minimal. Variables: xij -- the amount (in tons) to be shipped (weekly) from Mill i to Plant j , i = 1, 2, j = 1, 2, 3. Þ Mathematical model: Minimize the shipping cost 17 x11 + 22 x12 + 15 x13 + 18 x 21 + 16 x 22 + 12 x 23 Subject to x11 + x12 + x13 = 350 (product of Mill 1) x21 + x22 + x23 = 550 (product of Mill 2) x11 + x21 = 300 (requirement of Plant 1) x12 + x22 = 400 (requirement of Plant 2) x13 + x23 = 200 (requirement of Plant 3) xij ³ 0, i = 1, 2, j = 1, 2, 3 . Unbalanced Transportation Problem In the above problem, Total Supply = Total Demand (= 900 Tons). Such problem is called balanced transportation problem. If now Mill 1 produces 450 tons of newsprint a week, and other supply and demand have no change, then Total Supply = 1000 (Ton) > 900 (Ton) = Total Demand. We call such problem an unbalanced problem. An unbalanced transportation problem can be transferred into a balanced one by introducing a dummy plant (demand). Here we introduce the plant 4, which is an imaginary one, and let its demand be: Dummy Demand = Total Supply – Total Actual Demand (=100 Ton). Let the shipping cost from the two mills to the dummy plant be 0. That is, the table above is extended to: The problem now becomes a balanced transportation problem. If we use two additional variables xi 4 , i = 1,2, then the model is: Minimize the shipping cost 17 x11 + 22 x12 + 15 x13 + 18 x 21 + 16 x 22 + 12 x 23 Subject to x11 + x12 + x13 + x14 = 450 (product of Mill 1) x21 + x22 + x23 + x24 = 550 (product of Mill 2) x11 + x21 = 300 (requirement of Plant 1) x12 + x22 = 400 (requirement of Plant 2) x13 + x23 = 200 (requirement of Plant 3) x14 + x24 = 100 (requirement of Dummy Plant) xij ³ 0, i = 1, 2, j = 1, 2, 3 ,4. When the problem is solved and suppose in the solution, x14 =60 and x24 =40, then Mill 1 should keep 60 tons of newsprint unshipped, and Mill 2 should keep 40 tons of its product unshipped. Note that in the above model we let the shipping cost from the two mills to the dummy demand be 0. If the storage cost in Mill 1 is $5/ton per week and Mill 2 has no storage space, then we should change the shipping costs from the two mills to the dummy plant to 5 and ∞ (or a very large number N), respectively. Section 2.4 The Dynamic Planning Model The operation of many processes can be divided into distinct time periods. Decisions for one period affect not only that period, but also subsequent periods. Example 1 Consider the operation of a dealer of home heating oil. Maximum storage capacity ~ 10,000 gal (initially, has 3,000 gal stored) In each month, the dealer can sell up to 8,000 gal; purchase up to 5,000 gal. In the next 3 months; the purchase prices are, respectively, 80¢, 82¢, 85¢; (per gal) and the sell prices are, respectively, 92¢, 95¢, 97¢. (per gal) After the third month, any oil left has the value of 78¢. Problem: How much oil should the dealer purchase, sell, and store during each month in order to maximize profit? For i = 1, 2, 3, let Pi ~ # of gallons of oil purchased; Di ~ # of gallons of oil sold (distributed); Si ~ # of gallons of oil stored. Obvious constraints: for i = 1, 2, 3, Si £ 10,000; Pi £ 5,000; Di £ 8,000 . Consider interrelations between two successive months: for each month, Oil purchased during that month + Oil sold during the month = Oil stored from the previous month So, Month 1 : Month 2 : Month 3 : + Oil stored during the month P1 + 3000 = D1 +S1 P2 + S1 = D2 +S2 P3 + S2 = D3 +S3 Total Profit = (92 D1 + 95 D2 + 97 D3 + 78 S3) − (80 P1 + 82 P2 + 85 P3). Mathematical Model for the Problem Max 92 D1 + 95 D2 + 97 D3 + 78 S3 − 80 P1 − 82 P2 − 85 P3 Subject to: P1 + 3000 = D1 + S1 P2 + S1 = D2 + S2 P3 + S2 = D3 + S3 0 £ Pi £ 5,000, i = 1, 2, 3 0 £ Di £ 8,000, i = 1, 2, 3 0 £ Si £ 10,000, i = 1, 2, 3. Example 2 Under a short-term contract, a firm should produce 3700 units of some commodity over a period of 4 weeks with the delivery schedule below: Week 1 # of units 700 2 1200 3 1000 4 800 For late delivery, there is a $5 / unit / week of penalty. No penalty or storage charge for any units delivered early. The firm must deliver the 3700 units by the end of the fourth week. Working force Stable (under long-term contracts) workers: 35. New workers can be hired, but require a week of training with 1 experienced worker capable of training 5 new workers each week. Each worker (excluding the trainees & trainers) can produce 25 units / week. All workers receive $350 / week. All new workers hired during the 4 weeks must be laid off by the end of the fourth week, if not sooner, at a cost of $125 / worker. Raw material One raw material is required for the production. Each unit of the product needs 2 lb of the raw material. Each week the firm can obtain 2000 lb free material, but can be used only during the week. Extra amount of the raw material can be purchased from the market at a cost of $3 / lb. Problem: Set a production plan to meet the contract, meanwhile to minimize the total cost. Introducing the following variables: for week i, i = 1, 2, 3, 4, Hi ~ new workers to be hired Fi ~ workers to be laid off Ti ~ workers to train & to be trained Ii ~ workers to be idle Pi ~ workers to produce the commodity Mi ~ # of lb of raw material purchased from the outside source Di ~ # of units produced (delivered) Ui ~ the accumulated number of units required, but not delivered. Week 1 constraints: 35 + H 1 = T1 + I 1 + P1 H H 1 + 1 = T1 5 ( F1 = 0) cost: 25P1 = D1 2000 + M 1 ³ 2 D1 (why not 2000 + M1 = 2D1 ?) D1 + U 1 ³ 700 (why not D1 + U1 = 700 ?) 350 (35 + H 1 ) + 3M 1 + 5U 1 . Week 2 constraints: 35 + H 1 + H 2 - F2 = T2 + I 2 + P2 H H 2 + 2 = T2 5 F2 £ H 1 25P2 = D2 2000 + M 2 ³ 2 D2 D2 + U 2 ³ 1200 + 700 - D1 cost: 350 (35 + H 1 + H 2 - F2 ) + 125F2 + 3M 2 + 5U 2 . Week 3 Write by yourself & explain the constraints / costs. Week 4 Finally, let F5 ~ the workers hired for the project who are still in the plant in week 4. constraint: F5 = H 1 + H 2 + H 3 + H 4 - F2 - F3 - F4 . cost: 125 F5 . 4 Model: Min å (cost of week i ) + 125 F5 i =1 Subject to: constraints of week i for i = 1, 2, 3, 4 all variables ³ 0 From the above examples we see that the basic problem of linear programming is to optimize (maximize or minimize) a linear function subject to a system of linear constraints (linear equations or linear inequalities).