Dear All,

I am solving a chemical problem where many batches are produces on a single machine. In this problem, two sets are defined: t *t /1*T/* maximum time and *b /1*N/* number of batches. Its MIP problem which has 11 binary variables (to represent 11 different actions in the batch) defined over both sets (e,g, binary1(t,b), …, binary11(t,b)) and many other continuous variables for masses. The variables are initializes with 0 since the feasible solution is not known in advance due to the chemical process description. Now when i use T=700 and N=1 (one batch), the solution give me an optimal solution in approx. 190 secs. However, changing N=2 and T=1400 (considering that one batch take maximum 700 times intervals), the solver does not give me an feasible solution even after 1000 secs.

In my model, i have some hard equality constraints for making some process actions duration fixed (f). E.g;

*equation (b)… sum(t,(binary1(t,b))=E=f;*

What i have tried?

1:I made the process dynamics faster hence reducing T to 800 (now one batch can take 400 times interval instead of 700) and use N=2 but still no solution in 1000 secs.

2: I checked different CPLEX options e.g, *lpmethod, predual, feasopt* to make sure that it find a feasible solution in reasonable time but it did not worked.

I believe that the hard equality constraints is the source of much computational time requirement (or may be I am wrong because for N=, its works fine) since they remains infeasible after 1000 secs for N=2 (but they are feasible for N=1). I increased the simulation time (IterLim= 3000) but still the solver does not return any solution and keep searching the node.

Can any body have any suggestion how i can reduce this computational time/or find feasible solution of this problem? Is there some tuning tools that can make the B&B algorithm search faster?

Thanks