Controller2007
New Member
- Joined
- Sep 13, 2007
- Messages
- 7
I have a maximization problem in Solver which has a constraint I can't wrap my brain around. The constraint should be above a certain threshold, but it should also be the lowest value above that threshold. The problem is not linear - it's based on discrete values as in the following example.
I'd like to find the row with the maximum summed value. The constraint is that the first value must be over 3 (5, 7 or 10), but the lowest value, ie. 5 on row 3.
row 1: 1 10
row 2: 2 10
row 3: 5 10
row 4: 7 10
row 5: 10 10
Maximizing for the total value is easy with Solver, it gives row 5 with a value of 20. But if I introduce the constraint, the lowest value above 3 for the first column is 5 on row 3. That gives a total of 15, which is lower than the non-constrained value, but higher than what we could get by merely minimizing the first row as well.
Is it possible to get this done in Solver in one step? The only duct tape approach I've figured is a manual iterative approach: try different ranges for the value for the first row until you find the correct solution. But this gets unwieldy when there are hundreds of rows, or when the discrete values are bunched closely together.
I'd like to find the row with the maximum summed value. The constraint is that the first value must be over 3 (5, 7 or 10), but the lowest value, ie. 5 on row 3.
row 1: 1 10
row 2: 2 10
row 3: 5 10
row 4: 7 10
row 5: 10 10
Maximizing for the total value is easy with Solver, it gives row 5 with a value of 20. But if I introduce the constraint, the lowest value above 3 for the first column is 5 on row 3. That gives a total of 15, which is lower than the non-constrained value, but higher than what we could get by merely minimizing the first row as well.
Is it possible to get this done in Solver in one step? The only duct tape approach I've figured is a manual iterative approach: try different ranges for the value for the first row until you find the correct solution. But this gets unwieldy when there are hundreds of rows, or when the discrete values are bunched closely together.