Originally posted by: Janisch

After my model (MILP) as far as what it should do, a new problem opens up. For my record, the calculation takes a very long time (so far CPLEX could never finish the calculation).

I have already made some sense and tried it with the tuning tool. Unfortunately that did not bring the desired improvement.

Maybe someone has more ideas / approaches to optimize the term significantly?

Enclosed you will find my model and the accompanying Excel file. In addition, you will find the mathematical formulation. Maybe it is possible to improve the implementation. Since I am only a beginner in programming so far, I miss possible optimization approaches.

regards

One problem seems to be that CPLEX has a hard time to find a first feasible solution. Can you easily guess a solution for your model? No matter how bad it is.

Originally posted by: Janisch

Hello Daniel,

In my calculation cplex finds the first possible solutions after about 1 min. Is that long? 

I can only give the model a first possible sequence (x). But I can not calculate the associated times. The effort is too big for that. Or do you mean something else with a first possible solution?

Is there any other possible options?

regards

Hi,

since your model seems to be a scheduling one, have you thought about CPO ?

https://es.slideshare.net/PhilippeLaborie/an-introduction-to-cp-optimizer

regards

Originally posted by: Janisch

Hi Alex,

thanks for the link. I did not know the cp optimizer yet. However, I miss a bit of the programming skills to reformulate my problem for the cp optimizer. 

Can you help me to reword this problem step by step? I would be infinitely grateful

I don't think you have to reformulate your model. I took what you posted in your initial post, changed all variables to integer and added 'using CP;' at the top of the file.

That was enough to get CP optimizer going on your model.
 

Hm, ok, so on your side you get a first feasible than I do here.

In case moving to CP does not help, can you post the engine log?

Originally posted by: Janisch

Hello Daniel,

i have attached my Engine Log. After two minutes the gap is still 100%.

If the dual bound is 0 then CPLEX always displays a gap of 100%, no matter what the best feasible solution is. So for this instance you should not look at the relative but the absolute gap.

Since you are using CP now, can you please take your problem to the CP forum: https://www.ibm.com/developerworks/community/forums/html/category?id=33333333-0000-0000-0000-000000000268

You will probably find better answers there.