Sunteți pe pagina 1din 7

Sensitivity Analysis Sometimes Excel makes it easy for us to determine what will happen to the optimal solution for

a problem when a single data element is changed. In this handout, we look at two examples. Scenario: The Cement Mixers Problem Let G = number of bags granulated cement sold Let P = number of bags powdered cement sold max s.t. 6G + 3P G + P <= 1600 P >= 500 2G + P <= 2000 G, P >= 0 (max profit) (capacity) (demand) (time) (nonnegativity)

Part I: Objective Function Coefficients Suppose we vary the objective function coefficient for G. How does this affect the optimal solution(s)? A. The figure below is the sensitivity report when the objective function coefficient for G is equal to 4. (See worksheet Sensitivity (2var; c = 4)). We see that an optimal solution is G = 400 and P = 1200. This solution will remain optimal as long as the objective function coefficient for G stays between 3 (= 4 1) and 6 (= 4 + 2). Hence, the range of optimality for this coefficient associated with the solution (400,1200) is [3,6].

B. The figure below is the sensitivity report when the objective function coefficient for G is equal to 6. (See worksheet Sensitivity (2var; c = 6)). We see that an optimal solution is G = 750 and P = 500. This solution will remain optimal as long as the objective function coefficient for G stays between 6 (= 6 0) and ( 6 + 1030). Hence, the range of optimality for this coefficient associated with the solution (750,500) is [6, ).

Observe that the value of 6 is in both the range [3,6] and [6, ). Hence, when the O.F. coefficient for G is 6, we have alternate optima. C. The figure below is the sensitivity report when the objective function coefficient for G is equal to 8. (See worksheet Sensitivity (2var; c = 8)). We see that an optimal solution is G = 750 and P = 500. This solution will remain optimal as long as the objective function coefficient for G stays between 6 (= 8 2) and ( 8 + 1030). Hence, the range of optimality for this coefficient associated with the solution (750,500) is [6, ).1

We actually learned this fact when we selected the coefficient to be 6. That is, since 8 is in the range [6, ), (750,500) still must be an optimal solution.

D. The figure below is the sensitivity report when the objective function coefficient for G is equal to 2. (See worksheet Sensitivity (2var; c = 2)). We see that an optimal solution is G = 0 and P = 1600. This solution will remain optimal as long as the objective function coefficient for G stays between - ( 2 - 1030) and 3 (= 2 + 1). Hence, the range of optimality for this coefficient associated with the solution (0,1600) is (-, 3].

Observe that the value of 3 is in both the range [3,6] and (-, 3]. Hence, when the O.F. coefficient for G is 3, we have alternate optima.

NOTE: Excel determines the range of optimality by comparing the slope of the current objective function to the slope of the constraints that are binding at a solution. The book explains how to compute the range of optimality by hand, but you are not required to know how to do this. Part II: Right-Hand Sides A. What happens if we decrease the right hand side of the capacity constraint in the original problem to 1500 bags / day? The figure below is the sensitivity report for the ORIGINAL problem (with the RHS of the capacity constraint equal to 1600.) Excel reports an allow increase and decrease for RHS sides of constraints in the sensitivity report. We can use these numbers to calculate a range of feasibility associated with a solution. Definition: The range of feasibility associated with the RHS of a constraint is the range of values within which we may vary this RHS (ONLY!) without changing the set of constraints that are binding at the optimal solution. To determine the binding constraints at the current solution we compare the Final Value (LHS) of a constraint to its RHS. For the current solution, the DEMAND and TIME constraints are binding; the CAPACITY constraint is nonbinding (as are the nonnegativity constraints). From the report, we see that the allowable increase is infinite and the allowable decrease is 350. Thus, as long as the RHS of this constraint is between 1250 (= 1600 350) and ( 1600 + 1030), an optimal solution will be found at a point where the DEMAND and TIME constraints are binding; that is, at a point where these two constraints intersect.

B. What happens if we decrease the right hand side of the capacity constraint in the original problem to 1000 bags / day? Since 1000 is out of the range of feasibility for the RHS of this constraint in the ORIGINAL scenario, we expect different constraints to be binding at the optimal solution. The figure below is the sensitivity report for this new scenario. (See worksheet Sensitivity (2var; RHS)). From the report, we see that An optimal solution is (500,500). The binding constraints at this optimal solution are CAPACITY and DEMAND. For this RHS of the CAPACITY constraint, the allowable increase is 250 and the allowable decrease is 500. Thus, as long as the RHS of this constraint is between 500 (= 1000 500) and 1250 (= 1000 + 250), an optimal solution will be found at a point where the CAPACITY and DEMAND constraints are binding; that is, at a point where these two constraints intersect.

NOTE: When you change the RHS of a constraint, you change the shape of the feasible region. Thus the physical point where two binding constraints intersect may change even if the SET of constraints that are binding does not. Part III: Another type of Sensitivity Analysis Unlike changing an objective function coefficient or constraint RHS, answering questions that involve adding/removing a decision variable, adding/removing a constraint, or changing the LHS data of a constraint usually require resolving the model. Heres a simple example.

Suppose the cement mixer is considering introducing a new product 1 in cement cubes. It is estimated that the unit profit on the cubes is $2/bag. One bag of cubes requires twice as much time to produce as a bag of granulated cement. Let C = number of bags cement cubes sold max s.t. 6G + 3P + 2C G + P + C <= 1600 P >= 500 2G + P + 4C <= 2000 G, P, C >= 0 (max profit) (capacity) (demand) (time) (nonnegativity)

The figure below shows the sensitivity report for this scenario. (See worksheet Sensitivity (3var; original)). From the report, we see that no cubes are made. This makes sense since the profit per bag cubes is much lower than for any other product but the time per bag is much higher.

What if we increase the profit to $14/ bag cubes? The figure below shows the sensitivity report for this scenario. (See worksheet Sensitivity (3var; original)). From the report, we see that now cubes, but no granulated cement, are made.

S-ar putea să vă placă și