IEOR 240: Midterm Exam November 19, 2019 5:00 PM - 8:00 PM Name: (please print) SID: • The use of any electronic device (such as smart phone, tablets, laptops, calculators, etc.) is strictly prohibited. Please turn off all such devices. • Any communication with other students during the exam (including showing, viewing or sharing any writing) is strictly prohibited. • Any violation of the above will result in a score of 0 points for the exam, and will be reported to the Fung Institution. • Clearly state all assumptions. • Answer each of the following questions in the space provided. If you need more space to show major computations you performed to obtain your answer for a particular problem, use the back of the preceding page. • You can quote and use any general result stated in class notes or in the main body of the textbook as well as well-known general mathematical results but no references to other sources (including homework and examples - both in textbook and class - are allowed). • Present your work in an organized and neat fashion. • Good luck! Problem Score 1 (25) 2 (25) 3 (25) 4 (25 +10) Total (100+5) 1. [25 points] Suppose you are a production engineer in a factory. Now, you need to decide how much to produce in each period for the upcoming 5 periods. 1 There is a demand at the end of each period, and it must be satisfied. 2 There are two costs involved: • The first cost is a holding cost: for every unit of product held in inventory from the end of period i to the beginning of period i + 1, you incur a holding cost hi . • The second cost is associated with changing production level: if you change the quantity of product produced in i − 1 to a different amount in i, you incur a cost per unit change ci (for example, if your order quantity from period 1 to period 2 increases or decreases by 5 units, the associated order change cost is 5 × c2 ). 3 Assume you start with zero inventory and a production level of zero. There is no restrictions on the level of inventory at the end of period 5. 4 The relevant data is given in following table: Demand (di ) Holding cost (hi ) Production change cost (ci ) Period 1 50 0.5 3 Period 2 80 1.5 2 Period 3 60 1 1.5 Period 4 110 1 2 Period 5 90 0.5 2 (a) formulate a linear programming model that minimizes cost. (b) Now, assume that you can choose to satisfy any portion (up to what specified in the table), but that for each unit you satisfy, you earn a net profit of $1 per unit. In addition, you are required to pay the cost associated with holding and production changes. To do that, you are provided with a budget of $100 which is available before the start of the first period. For each period, you can also use money earned from previous periods. At the end of each period, you can deposit whatever cash you have at hand to a bank which pays %1 interest per period (you can withdraw money from the accumulated deposits at the beginning of every period). Revise your formulation to model this new scenario to maximize profit at the end of period 5.. Clearly define any variables and parameters you use and briefly explain each of the constraints. In addition, clearly state any assumption you make. Solutions: (a) • zi : The change of production in period i • xi : Production level in period i • Ii : Inventory in period i N X (hi Ii + ci zi ) minimize i=1 subject to (b) Ii = Ii−1 + xi − di zi ≥ xi − xi−1 zi ≥ xi−1 − xi z1 ≥ x1 I1 = I1 − d1 Ii ≥ 0 xi ≥ 0 i = 2, . . . , N i = 2, ..., N i = 2, ..., N i = 1, ..., N i = 1, ..., N • zi : The change of production in period i • xi : Production level in period i • Ii : Inventory in period i • yi : Demand to meet in period i • ki : Money you have in period i (right after you make change on the production level) maximize N X (yi − hi Ii − ci zi ) + k5 i=1 subject to Ii = Ii−1 + xi − yi zi ≥ xi − xi−1 zi ≥ xi−1 − xi z1 ≥ x1 I1 = I1 − y1 k1 = 100 − zi ki = 1.01 ∗ ki−1 + yi−1 − zi Ii ≥ 0 xi ≥ 0 yi ≥ 0 yi ≤ di ki ≥ 0 i = 2, ..., N i = 2, ..., N i = 2, ..., N i = 2, ..., N i = 1, ..., N i = 1, ..., N i = 1, ..., N i = 1, ..., N i = 1, ..., N 2. [25 points] The NCAA is making plans for allocating tickets to the upcoming regional basketball championship of the South-East. Up to 10,000 available seats will be allocated to the media, students of the two competing universities, and the general public, subject to the following conditions: • Media people are admitted free, and at least 1000 tickets must be reserved for the media. • The NCAA receives $45 per ticket sold to students of the two competing universities and $100 per ticket from tickets sold to the general public. • At least half as many tickets should be allocated to students of the two competing universities as to the general public. With these restrictions, we formulate an LP so that the NCAA maximizes their revenue. To model this problem algebraically, we let x1 , x2 and x3 (in thousands) represent the numbers of tickets allocate to media, students and general public, respectively. The linear programming formulation is as follows: max 45x2 + 100x3 s.t. x1 + x2 + x3 ≤ 10, 1 x2 − x3 ≥ 0, 2 x1 ≥ 1, xi ≥ 0, i = 1, 2, 3. Solving the above model using the AMPL solver, we obtained the following sensitivity analysis report: Variables Name Media Tickets University Tickets Public Tickets Final Value 1 3 6 Reduced Objective Cost Coefficients 0 0 0 45 0 100 Upper Bound 81.667 100 1E+20 Lower Bound -1E+20 -200 45 Constraints Final Name Value Capacity Constraint 10 Univ./Public Constraint 0 Media Constraint 1 Shadow Price 81.667 -36.663 -81.667 Constraint Upper R.H.Side Bound 10 1E+20 0 9 1 10 Lower Bound 1 -4.5 0 Consider each of the following questions independently. Be sure to explain (using the sensitivity report) all your answers. (a) What is the optimal value of the decision variables, and what is the optimal revenue? (b) Suppose it is possible to reconfigure the arena to have a total of 20,000 seats (so that 20,000 tickets can be sold). How much additional revenue would be gained from the expanded seating? (c) In order to make the championship game more affordable for fans, it is proposed to reduce the price of the general public tickets to $50. Will the same solution be optimal if we implement this change? If the same solution (with the reduced price) is optimal, how much revenue for NCAA would be lost? (d) Suppose you don’t have the data to answer question (c), provide an upper bound to the possible revenue loss from the price reduction as in (c). You should try to get (and justify!) the lowest upper bound that you can. (e) To accommodate the high demand from the students in the participating universities, the NCAA is considering marketing a new scrunch seat that costs $15 each and consumes only 50% of a regular seat (that is, each physical seat can accommodate two students who purchase scrunch seat), but counts fully against the “university ≥ half public” rule. Specifically, let x4 indicate the number of scrunch seats (in thousands) allocated, then the problem can be reformulated as follows: max 45x2 + 100x3 + 30x4 s.t. x1 + x2 + x3 + x4 ≤ 10, 1 x2 − x3 + 2x4 ≥ 0, 2 x1 ≥ 1, xi ≥ 0, i = 1, 2, 3, 4. Would selling tickets for such seats be profitable? (f) Explain, in words, why the shadow price of the Media Constraint is the same as the shadow price of the Capacity Constraint, but with opposite sign. Solutions: (a) The optimal value of the decision variables are (x1 , x2 , x3 ) = (1, 3, 6). The optimal revenue is 735. (b) The increase of 10 of the right-hand side of the first constraint is within the allowable increase. So the shadow price remains the same and the additional revenue gained from the expanded seating is 816.67. (c) The decrease of 50 in the objective coefficient of x3 is within the allowable decrease. This implies that the optimal solution is still optimal and the lost revenue for NCAA will be 300. (d) The current optimal solution will be at least feasible after we change the objective coefficient of x3 to 50. It is clear that this feasible solution has the new objective value of 45∗3+50∗6 = 435. By the optimality condition, the new optimal revenue will be greater than or equal to this value. Therefore, the loss is not greater than 735 − 435 = 300. (e) Adding one more decision variable in the primal problem is equivalent to adding one more constraint to the dual problem. The feasible set of the dual problem becomes smaller, so its optimal objective value becomes larger since it is a minimization problem. By the optimality condition, the optimal value of the dual is the same as that of the primal problem, and the new optimal objective value is greater than or equal to the current one. Therefore, we conclude that selling tickets for such seats is profitable. (f) If we increase the right-hand side of the Media constraint by 1, then in order to maximize the objective value, the value of x1 will be set to 2 and x2 + x3 will be equal to 8 in the new optimal solution. On the other hand, when we decrease the right-hand side of Capacity constraint by one unit, again to maximize the objective function, x1 remains 1 and x2 + x3 will be equal to 8 in the new optimal solution. Therefore, increasing the right-hand side of the Media constraint will change the optimal objective value in the way decreasing the right-hand side of the Capacity constraint does. 3. [25 points] A factory works a 24 hour day, 7 day week in producing four products. Since only one product can be produced at a time the factory operates a system where, throughout one day, the same product is produced (and then the next day either the same product is produced or the factory produces a different product). The rate of production is: Product Number of units produced per hour worked 1 100 2 250 3 190 4 150 The only complication is that in changing from producing product 1 one day to producing product 2 the next day five working hours are lost (from the 24 hours available to produce product 2 that day) due to the necessity of cleaning certain oil tanks. To assist in planning the production for the next week the following data is available that details the current Stock and Demand (units) for each day of the week . Product 1 2 3 4 Current Stock(units) 5000 7000 9000 8000 1 1500 4000 2000 3000 2 1700 500 2000 2000 3 4 1900 1000 1000 3000 3000 2000 2000 1000 5 6 7 2000 500 500 500 1000 2000 2000 2000 500 1000 500 500 Product 3 was produced on day 0. The factory is not allowed to be idle (i.e. one of the four products must be produced each day). Stockouts are not allowed. At the end of day 7 there must be (for each product) at least 1750 units in stock. If the cost of holding stock is $1.50 per unit for products 1 and 2 but $2.50 per unit for products 3 and 4 (based on the stock held at the end of each day) formulate the problem of planning the production for the next week as an integer program in which all the constraints are linear. • Formulate the problem as a linear integer programming problem. Solutions: The solutions are as follows: • Approach 1: minimize 7 X hi Ii i=1 subject to x3,0 = 1 4 X xi,t t = 1, . . . , 7 i=1 Ii,7 ≥ 1750 Ii,0 = si Ii,t = Ii,t−1 + Pi,t−1 − di,t zt ≥ x2,t + x1,t−1 − 1 zt ≥ (x2,t +x2 1,t−1 ) Tt = 24 − 5zt Pi,t ≤ xi,t ∗ M Pi,t ≤ Tt θi Ii,t , Tt , Pi,t , zt ≥ 0 xi,t = {0, 1} i = 1, . . . , 4 i = 1, . . . , 4 i = 1, . . . , 4, t = 1, . . . , 7 t = 1, . . . , 7 t = 1, . . . , 7 t = 1, . . . , 7 i = 1, . . . , 4, t = 1, . . . , 7 i = 1, . . . , 4, t = 1, . . . , 7 • Approach 2: minimize 7 X hi Ii i=1 subject to x3,0 = 1 4 X xi,t t = 1, . . . , 7 i=1 Ii,7 ≥ 1750 Ii,0 = si Ii,t = Ii,t−1 + Pi,t−1 − di,t zt ≥ x2,t + x1,t−1 − 1 zt ≥ (x2,t +x2 1,t−1 ) Pi,t = θi ∗ (24 ∗ xi,t − 5 ∗ zt ) Ii,t , Tt , Pi,t , zt ≥ 0 xi,t = {0, 1} i = 1, . . . , 4 i = 1, . . . , 4 i = 1, . . . , 4, t = 1, . . . , 7 t = 1, . . . , 7 t = 1, . . . , 7 i = 1, . . . , 4, t = 1, . . . , 7 4. [25 points + 10 points bonus for question (f )] For each of the following claims, determine whether it is true or false. Justify your answer by providing either a short proof or a counterexample as appropriate. No explanation - no credit (a) For any linear program which is feasible, the dual program is also feasible. (b) An integer program must either have no optimal solution, a unique optimal solution, or infinitely many optimal solutions. (c) The optimal objective value of the linear programming max c1 x1 + c2 x2 + · · · + cn xn s.t. x1 + x2 + · · · + xn ≤ 1 xj ≥ 0, j = 1, 2, ..., n. where ci > 0 for i = 1, . . . , n, is max{c1 , c2 , ..., cn }. (d) Suppose the linear program P : min c> x s.t. Ax ≥ b, x ≥ 0. is unbounded. Then changing the values of the bi ’s will never result in this linear program having an optimal solution. (e) Suppose two minimization problems (A and B) with linear objective and linear constraints are identical, except that problem A requires to satisfy at least one of the first two constraints, while problem B requires to satisfy both the first two constraints. Given that problem A is feasible and is certified to be bounded with a lower bound of objective value α for every feasible solution, the optimal solution (if exists) of problem B has an objective function value of at least α. (f) Let x̄1 , . . . , x̄n ; ȳ1 , . . . , ȳm be feasible solutions to the primal and dual linear program (in symmetric form), respectively. Also, let s̄1 , . . . , s̄m ; r̄1 , . . . , r̄n be the slacks to the primal and dual feasible solutions above, respectively. Then, if Pm Pn Pm i=1 bi ȳi ≤ 0, then j=1 r̄j x̄j + i=1 s̄i ȳj is an upper bound on the optimal value of the dual. Solutions: The solutions are as follows, (a) False. The LP min x x≥1 s.t. −x ≥ 0 x≥0 is infeasible, since x cannot be ≥ 1 and ≤ 0, yet its dual, max s.t. y1 − 2y2 y1 − y2 ≤ 1 y1 , y2 ≤ 0 is feasible. Reverse the orders for a counterexample. (b) False. The IP max s.t. x+y x+y ≤1 x, y ∈ {0, 1} has exactly two optimal solutions, (1, 0) and (0, 1). (c) False. This is true only if ci ≥ 0, ∀i. If all ci < 0, for example, we can choose xi = 0, ∀i, giving us the optimal solution and value of 0. However, the optimal objective value of the above LP is equal to max{0, c1 , c2 , . . . , cn }. (d) True. We need to consider two cases. If P is still feasible after changing the values of the bi ’s, we find that c> z < 0, Az ≥ 0, z ≥ 0, still has a solution since the original P is unbounded. This implies that the certificate of unboundedness exists and P can not be optimal after changing the values of the bi ’s. Otherwise, P is infeasible and hence not optimal after changing the values of the bi ’s. (e) True. The solutions of B are all solutions of A since x or y ⇒ x and y. So the minimum is being taken over a (weakly) smaller set in B, so the min. of it will be at least min(A) = α. (f-1) True. Satisfying complementary slackness means being, in particular, feasible. Then complementary slackness says that they are optimal solutions for their respective systems, then weak duality gives the desired equation. (f-2) False. Choose c = 0, b = 0 so that all x, y satisfy this condition. Choose any A, x, y such that Ax 6≥ 0, AT y 6≤ 0, then these choices are not feasible, hence don’t satisfy complementary slackness conditions.