SOLVER precision settings

Not sure if this is the right forum to post a SOLVER question, so feel free to redirect me if there's a better place.

I've created an "inputs" section of my worksheet specifically for Solver. It is a small table with two columns. The first column is the label. Second column is all integers. This is the "input" section. (where I need a real number in my formulas I simply multiply the # in the second column by .01 or other partial #).

I created the second column with all integers so that Solver would not need to go through millions of real numbers to find the optimal inputs. I was hoping this would significantly reduce the amount of time Solver takes to reach the best possible combination of numbers.

As an example: A constraint says find an integer with an upper bound of 300 and a lower bound of 0.

What I'm discovering is that Solver is iterating real numbers even though I've specified integers. Instead of running from 0 to 300 in integers e.g. 0, 1, 2, 3 etc. It is running numbers with 12 decimal places (e.g. 1.019853879534).

I'm using the Evolutionary method since that's the best approach for my workbook. I don't know what the two "precision" settings in Solver really do. I've reduced them from the default (about 8 decimal places?) to one. But that has no visible effect.

Usually, when Solver finishes its work, it does stop at an integer. But there are times when it stops at a number with 12 decimal places even though I've set it to run just integers.

Anyone have a suggestion? Thanks!

Solver work by using real numbers but if integers are specified end result should be an integer.

Still if Solver does not find a solution it may present a solution based on decimal numbers but it should specify that it has found no solution.

Can you see from the numbers which constraint that causes the problem i.e. stopping solver from finding a solution?

Perhaps you could upload a sample file with your model?

Alf

Thanks Alf. After posting this, I decided to try something new. I had been working on a large workbook for almost a year, saving it with a new version number each time I made a change. So as an experiment I did a copy/paste from the sluggish version to a fresh workbook, and this had an impressive effect on Solver. It runs very quickly now, and always stops at integers.

So the old workbook must have had some hidden or unused data. I wonder if Solver itself has the equivalent of a "cache."

I did note that with the old notebook, when Solver completed its run and left me with a number having 12 digits or so, I would manually change that number to an integer, which had no effect on the workbook's results. (But I can't remember if it also gave me the message that it couldn't improve the results).

And, I would usually be fine with uploading a workbook but I'm unable to do this with the one causing issues, as it contains sensitive data.. for which I've signed a nondisclosure agreement.

Maybe I'll leave this thread open another day or two in case someone else has a comment...?

Thanks again!

I'm cleaning up my old threads and realized I hadn't marked this one as solved. Kaper to your post, I looked into opensolver but since it was a bit too far out of my comfort zone, I realized that there was a quite simple method to constrain Solver's iterations.

When I know the general bounds of a variable involves binary numbers, I create two columns. Every row in Column A is always an integer (and defined as such in Solver setup). Rows in Column B are the actual "inputs" to be tested.

If an input is a percentage then I decide how many decimal places are going to be meaningful. If I want to see results to two decimal points, (e.g. .01 to .99) then Column B becomes .01 x Col A.

I make sure that Solver is configured to see Column A as an integer. When it runs, it will go through 99 iterations rather than thousands (at least, this is my perception of what its doing). I do note that it speeds up Solver significantly.