Linear programming: introduction and examples

Transcription

1 Linear programming: introduction and examples G. Ferrari Trecate Dipartimento di Ingegneria Industriale e dell Informazione Università degli Studi di Pavia Industrial Automation Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 1 / 42

2 Linear Programming (LP) Widely used optimization technique in management science optimal allocation of limited resources for maximizing revenues or minimizing costs Basic problem A Linear Programming (LP) problem is (1) with f (x) = c T x (linear cost) g i (x) = a T i x b i (affine constraints) An LP problem is a convex optimization problem min g i (x) 0 f (x), x R n (1) i=1,2,...,m Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 2 / 42

3 Linear Programming (LP) Canonical form An LP problem is in canonical form if it is written as or min a T i x b i, i=1,2,...,m x j 0, j=1,2,...,n max a T i x b i, i=1,2,...,m x j 0, j=1,2,...,n c T x c T x constraints and positivity constraints on all variables Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 3 / 42

4 PL - matrix notation Vector inequalities x 0 means x 1 0 x x n 0 Constraints a1 T x b 1 a2 Tx b 2... amx T b m a T 1 a T Ax b, A = 2., b = b 2. b m a T m b 1 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 4 / 42

5 PL - matrix notation LP problem in generic form min Ax b ct x LP problem in canonical form (LP-C) min c T x Ax b x 0 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 5 / 42

6 LP and management science Typical decision problems in industry Product mix Diet problems Blending problems Transport problems Product mix with resource allocation Multiperiod production planning Portfolio optimization... Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 6 / 42

7 Product mix A company manifactures two radio models (low-cost and high-end) and produces two components Department A: antennas (max. 120h hours of production per day) 1h of work for a low-cost antenna 2h of work for a high-end antenna Department B: case (max. 90h hours of production per day) 1h of work for a low-cost case 1h of work for a high-end case Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 7 / 42

8 Product mix The company has two assembly lines (1 radio=1 antenna + 1 case) Line 1: production of low-cost models. No more than 70 units/day Line 2: production of high-end models. No more than 50 units/day Profits: 20 Euros for a low-cost radio and 30 Euros for a high-end radio. Assuming the company will sell all radios, which is the optimal number of units, for each model, that must be produced daily for maximizing the revenue? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 8 / 42

9 Product mix Choice of variables Usually this is the most difficult step in representing decision problems as optimization problems! Guideline: cost and constraints must be a function of optimization variables only. Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 9 / 42

10 Product mix Choice of variables - product mix x 1 : number of produced low-cost radios per day x 2 : number of produced high-end radios per day Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 10 / 42

11 Product mix Choice of variables - product mix x 1 : number of produced low-cost radios per day x 2 : number of produced high-end radios per day Cost and type of problem Cost: 20x x 2, to be maximized Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 10 / 42

12 Product mix Constraints Capacity constraints of assembly lines x 1 70 (line 1) x 2 50 (line 2) Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 11 / 42

13 Product mix Constraints Capacity constraints of assembly lines x 1 70 (line 1) x 2 50 (line 2) Capacity constraints of production departments x 1 + 2x (department 1) x 1 + x 2 90 (department 2) Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 11 / 42

14 Product mix Constraints Capacity constraints of assembly lines x 1 70 (line 1) x 2 50 (line 2) Capacity constraints of production departments x 1 + 2x (department 1) x 1 + x 2 90 (department 2) Positivity constraints: x 1 0, x 2 0 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 11 / 42

15 Product mix LP problem max x1,x 2 20x x 2 x 1 70 x 2 50 x 1 + 2x x 1 + x 2 90 x 1 0 x 2 0 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 12 / 42

16 Product mix LP problem max x1,x 2 20x x 2 x 1 70 x 2 50 x 1 + 2x x 1 + x 2 90 x 1 0 x 2 0 LP problem - matrix notation min c T x Ax b x 0 [ x1 ] [ ] 20 x =, c = x A = , b = Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 13 / 42

17 Product mix revised A company manifactures two models of cases for telephones (model 1 and model 2). The production cycle comprises three phases with bounded resources that must be allocated to the two products Each phase is modeled through the maximal availability of men-hours per day and men-hours required to process a single unit Profits: 30 Euros for a model 1 unit and 20 Euros for a model 2 unit Assuming all cases will be sold, which is the optimal number of units, for each model, that must be produced daily for maximizing the revenue? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 14 / 42

18 Product mix revised Choice of variables x 1 : n. of model 1 units x 2 : n. of model 2 units Cost and type of problem Cost: 30x x 2, to be maximized Constraints Capacity constraints in each phase Positivity constraints: x 1 0, x 2 0 8x 1 + 4x (assembly) 4x 1 + 6x (finishing) x 1 + x (quality control) Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 15 / 42

19 Product mix revised LP problem max x1,x 2 30x x 2 8x 1 + 4x x 1 + 6x x 1 + x x 1 0 x 2 0 LP problem - matrix notation max c T x Ax b x 0 [ ] [ ] x1 30 x =, c = x 2 20 A = , b = Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 16 / 42

20 Diet problem A company producing chicken food can use two types of additives with different vitamin contents. The product is obtained by mixing additives in a suitable quantity Vitamin contents and additive costs are given in the picture 1 box of chicken food must contain at least 2 gr. of vitamin A, at least 12 gr. of vitamin B, at least 36 gr. of vitamin C and at least 4 gr. of vitamin D Which is the quantity of additives that allows the company to produce one box of chicken food while minimizing the costs? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 17 / 42

21 Diet problem Choice of variables x 1 : n. of boxes of Additive 1 x 2 : n. of boxes of Additive 2 Cost and type of problem Cost: 20x x 2, to be minimized Constraints Minimal quantities of vitamins Positivity constraints: x 1 0, x 2 0 x 1 2 (vitamin A) 2x 1 + x 2 12 (vitamin B) 2x 1 + 5x 2 36 (vitamin C) x 2 4 (vitamin D) Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 18 / 42

22 Diet problem LP problem min x1,x 2 20x x 2 x 1 2 2x 1 + x x 1 + 5x 2 36 x 2 4 x 1, x 2 0 LP problem - matrix notation min c T x Ax b x 0 [ ] [ ] x1 20 x =, c = x A = , b = Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 19 / 42

23 Blending problem A company produces 3 types of gasoline (leaded, unleaded, high-octane) blending 4 types of oils in suitable proportions Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 20 / 42

24 Blending problem Problem data The maximal number of available barrels and the buying costs are given in the picture Each gasoline type must fulfill the following composition requirements GASOLINE % of Oil 1 % of Oil 2 % of Oil 3 % of Oil 4 Leaded at least 40% no more than 20% at least 30% Unleaded at least 40% High octane at least 10% no more than 50% Gasoline selling prices are: 12 Euros/barrel for leaded, 18 Euros/barrel for unleaded and 10 Euros/barrel for high-octane Which are the optimal quantities of the 4 components that have to be bought in order to maximize the profits (profits = revenues - costs)? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 21 / 42

25 Blending problem Choice of variables x ij : n. of barrels of oil i used for gasoline of type j, i = 1, 2, 3, 4, j {L, U, H}, L= Leaded, U= Unleaded, H= High-octane Cost and type of problem Cost to be maximized 12 4 x il + 18 i=1 4 x iu + 10 i=1 4 i=1 x ih }{{} revenues 9 x 1j 7 j {U,L,H} j {U,L,H} x 2j 12 j {U,L,H} x 3j 6 j {U,L,H} } {{ } costs x 4j Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 22 / 42

26 Blending problem Constraints Maximal availability of barrels x 1L + x 1U + x 1H 5000 (Oil 1) x 2L + x 2U + x 2H 2400 (Oil 2) x 3L + x 3U + x 3H 4000 (Oil 3) x 4L + x 4U + x 4H 1500 (Oil 4) Composition requirements x 1L x 1L +x 2L +x 3L +x 4L 0.4 in affine form 0.6x 1L 0.4x 2L 0.4x 3L 0.4x 4L 0 x 2L x 1L +x 2L +x 3L +x 4L 0.2 in affine form 0.2x 1L + 0.8x 2L 0.2x 3L 0.2x 4L 0 x 3L x 1L +x 2L +x 3L +x 4L 0.3 in affine form 0.3x 1L 0.3x 2L + 0.7x 3L 0.3x 4L 0 x 3U x 1U +x 2U +x 3U +x 4U 0.4 in affine form 0.4x 1U 0.4x 2U + 0.6x 3U 0.4x 4U 0 x 1H x 1H +x 2H +x 3H +x 4H 0.1 in affine form 0.9x 1H 0.1x 2H 0.1x 3H 0.1x 4H 0 x 2H x 1H +x 2H +x 3H +x 4H 0.5 in affine form 0.5x 1H + 0.5x 2H 0.5x 3H 0.5x 4H 0 Positivity constraints: x ij 0, i = 1, 2, 3, 4, j {U, L, H} Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 23 / 42

27 Blending problem LP problem - matrix notation with max c T x Ax b x 0 A = x = [ x 1L x 2L x 3L x 4L x 1U x 2U x 3U x 4U x 1H x 2H x 3H x 4H ] T c = [ ] T b = [ ] T Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 24 / 42

28 @ I Transport I I " A company owns n fuel storage points (i.e. n supply points) and it must supply m gas stations (i.e. m demand points). Every gas station can be reached from all storage points. Data and requirements Every storage point has a maximal availability of a i 0 i = 1,..., n liters of fuel Every gas station specifies a demand of b j 0, j = 1,..., m liters of fuel t ij 0 is the cost for shipping 1 liter of fuel from storage point i to gas station j Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 25 / 42

29 @ I Transport I I " Which are the optimal quantities of fuel to be shipped from each storage point to each gas station in order to minimize transport costs while fulfilling exactly the demands? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 26 / 42

30 Transport problem Choice of variables x ij, i = 1,..., n, j = 1,..., m, n. of liters of fuel shipped from storage point i to gas station j Cost and type of problem Cost to be minimized n i=1 m t ij x ij j=1 Constraints Maximal availability of fuel at storage points m j=1 x ij a i, i = 1,..., n Demand constraints n i=1 x ij = b j, j = 1,..., m Positivity constraints: x ij 0, i = 1,..., n, j = 1,..., m Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 27 / 42

31 Transport problem LP problem min n i=1 n j=1 t ijx ij m j=1 x ij a i, i = 1,..., n n i=1 x ij = b j, j = 1,..., m (A) x ij 0, i = 1,..., n, j = 1,..., m Remarks Constraints (A) can be replaced by n i=1 x ij b j, j = 1,..., m. Why? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 28 / 42

32 Transport problem LP problem min n i=1 n j=1 t ijx ij m j=1 x ij a i, i = 1,..., n n i=1 x ij = b j, j = 1,..., m (A) x ij 0, i = 1,..., n, j = 1,..., m Remarks Constraints (A) can be replaced by n i=1 x ij b j, j = 1,..., m. Why? The LP problem is always feasible? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 28 / 42

33 Transport problem LP problem min n i=1 n j=1 t ijx ij m j=1 x ij a i, i = 1,..., n n i=1 x ij = b j, j = 1,..., m (A) x ij 0, i = 1,..., n, j = 1,..., m Remarks Constraints (A) can be replaced by n i=1 x ij b j, j = 1,..., m. Why? The LP problem is always feasible? NO! A necessary and sufficient condition for feasibility is n a i i=1 m j=1 b j Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 28 / 42

34 Product mix with resource allocation In standard product mix problems, capacity constraints are fixed. We now assume to have resources that can be allocated to different production phases A food company produces 4 products P i, i = 1, 2, 3, 4. Each product must go through 3 production lines L j, j = 1, 2, 3. Each line is modeled through the units of work (u.o.w.) required to process a single unit of product Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 29 / 42

35 Product mix with resource allocation Requirements At least 1000 units of P 2 and no more than 500 units of P 1 must be produced U.o.w are men-hours that fall in 7 categories T i, i = 1,..., 7 according to the versatility of workers Category Destination line Max. availability T 1 Line T 2 Line T 3 Line T 4 Lines 1 and T 5 Lines 1 and T 6 Lines 2 and T 7 Lines 1,2 and Profits per unit are 10 Euros for P 1, 12 Euros for P 2, 13 Euros for P 3 and 14 Euros for P 4 Assuming all products will be sold, which is a joint production and resource allocation plan that maximizes the profits? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 30 / 42

36 Product mix with resource allocation Choice of variables Two categories of variables are needed P i, i = 1,..., 4: units of product i T 4 L 1, T 4 L 2, T 5 L 1, T 5 L 3, T 6 L 2, T 6 L 3, T 7 L 1, T 7 L 2, T 7 L 3. For instance T 4 L 1 is the number of men-hours of category T 4 allocated to line 1. Cost and type of problem Cost to be maximized 10P P P P 4 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 31 / 42

37 Product mix with resource allocation Constraints Capacity constraints of production lines 2P 1 + 2P 2 + 3P 3 + 3P T 4 L 1 + T 5 L 1 + T 7 L 1 (line 1) 2P 1 + 3P 2 + P 3 + 2P T 4 L 2 + T 6 L 2 + T 7 L 2 (line 2) P 1 + 2P 2 + 2P 3 + 3P T 5 L 3 + T 6 L 3 + T 7 L 3 (line 3) Bounds on the number of products P P Maximal availability of u.o.w. for versatile workers T 4 L 1 + T 4 L T 5 L 1 + T 5 L T 6 L 2 + T 6 L T 7 L 1 + T 7 L 2 + T 7 L Positivity contraints: P 1, P 2, P 3, P 4, T 4 L 1, T 4 L 2, T 5 L 1, T 5 L 3, T 6 L 2, T 6 L 3, T 7 L 1, T 7 L 2, T 7 L 3 0 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 32 / 42

38 Product mix with resource allocation LP problem max 10P P P P 4 2P 1 + 2P 2 + 3P 3 + 3P T 4 L 1 + T 5 L 1 + T 7 L 1 2P 1 + 3P 2 + P 3 + 2P T 4 L 2 + T 6 L 2 + T 7 L 2 P 1 + 2P 2 + 2P 3 + 3P T 5 L 3 + T 6 L 3 + T 7 L 3 P P T 4 L 1 + T 4 L T 5 L 1 + T 5 L T 6 L 2 + T 6 L T 7 L 1 + T 7 L 2 + T 7 L P 1, P 2, P 3, P 4, T 4 L 1, T 4 L 2, T 5 L 1, T 5 L 3 0 T 6 L 2, T 6 L 3, T 7 L 1, T 7 L 2, T 7 L 3 0 Exercise Write the LP problem in matrix notation Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 33 / 42

39 Multiperiod production planning A company must determine how many washing machines should be produced during each of the next 5 months. Washing machines production can be internal or subcontracted to another company. Production capacities and costs are given in the figure Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 34 / 42

40 Multiperiod production planning Data and Requirements Inventory cost: 2 Euros/unit for a whole month Inventory at the beginning of the first month: 300 units Inventory at the end of the fifth month: 300 units Demand during each of the next five months (the company must meet demands on time) Month Demand How many washing machines must be produced (internally and by the subcontractor), each month, in order to meet demands on time and minimize the costs? Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 35 / 42

41 Multiperiod production planning Choice of variables Three categories of variables are needed P i, i = 1,..., 5: units produced internally during month i S i, i = 1,..., 5: units produced by the subcontrator during month i I i, i = 1,..., 4: inventory at the end of month i Cost and type of problem Cost to be minimized 10 5 P i + 15 i=1 5 S i + 2 i=1 4 i=1 I i Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 36 / 42

42 Multiperiod production planning Constraints Maximal production P i 2000, i = 1,..., 5 (internal production) S i 600, i = 1,..., 5 (subcontracting) Balance constraints: for month i, Inventory at the end of month i = Inventory at the end of month i month i production month i demand I 1 = P 1 + S I 2 = +I 1 + P 2 + S I 3 = +I 2 + P 2 + S I 4 = +I 3 + P 3 + S = +I 4 + P 4 + S Positivity contraints: P i, S i 0, i = 1..., 5, I j 0, j = 1..., 4 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 37 / 42

43 Multiperiod production planning LP problem min 10 5 i=1 P i i=1 S i P i 2000, i = 1,..., 5 S i 600, i = 1,..., 5 I 1 P 1 S 1 = I 2 I 1 P 2 S 2 = 2100 I 3 I 2 P 2 S 2 = 2400 I 4 I 3 P 3 S 3 = 3000 I 4 P 4 S 4 = P i, S i 0, i = 1..., 5 I j 0, j = 1..., 4 i=1 I i Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 38 / 42

44 Portfolio optimization A manager must decide how to invest Euros in different financial products. His goal is to maximize earnings while avoiding high risk exposure Financial products and expected return on investment Financial product Market Return % T 1 Cars - Germany 10.3 T 2 Cars - Japan 10.1 T 3 Computers - USA 11.8 T 4 Computers - USA 11.4 T 5 Household appliances - Europe 12.7 T 6 Household appliances - Asia 12.2 T 7 Insurance - Germany 9.5 T 8 Insurance - USA 9.9 T 9 BOT 3.6 T 10 CCT 4.2 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 39 / 42

45 Portfolio optimization Investment requirements 1 No more than Euros in the car options 2 No more than Euros in the computer options 3 No more than Euros in the appliance options 4 At least Euros in the insurance options 5 At least Euros in BOT or CCT 6 At least 40% of the money invested in CCT must be invested in BOT 7 No more than Euros must be invested in German options 8 No more than Euros must be invested in USA options Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 40 / 42

46 Portfolio optimization Choice of variables T i, i = 1,..., 10: thousands of Euros invested in the financial product i Cost and type of problem Cost to be maximized 0.103T T T T T T T T T T 10 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 41 / 42

47 Portfolio optimization Constraints Total investment: 10 i=1 T i = 500 Investment requirements: 1 T 1 + T T 3 + T T 5 + T T 7 + T T 9 + T T 9 0.4T T 1 + T T 3 + T 4 + T Positivity constraints: T i, 0, i = 1..., 10 Ferrari Trecate (DIS) Linear programming: introduction and examples Industrial Automation 42 / 42