MAiNGO
What can go wrong?

After starting the MAiNGO executable nothing happens

This is usually the case if some of the libraries (in most cases CPLEX) are not found. Make sure that after you installed CPLEX, the CPLEX .lib and .dll files can be found in your environment variables. If you cannot change your environment variables, simply copy the .dll files to the MAiNGO.exe folder.

MAiNGO reports a Problem to be infeasible although you expect it to be feasible

Double-check your variable bounds, model equations, and selection of constraints to make sure there are no mistakes. This is the cause in most cases. Check whether your problem might be badly scaled and consider scaling variables and objective (note that the constraints need to be properly scaled to ensure consistent application of tolerances, cf. Section Modeling in MAiNGO). If the problem persists, try disabling the settings BAB_alwaysSolveObbt, PRE_obbtMaxRounds and BAB_dbbt, which disables all range reduction. If this solves the problem, there might be an issue in the range reduction techniques. In this case, please report the problem to the developers. Otherwise, try to perform a multi-start by setting PRE_pureMultistart to 1 and PRE_maxLocalSearches to a high number, in addition use a powerful upper bounding solver (e.g., IPOPT) and try a local solver from multiple initial points. If a feasible point is found like this, there is an issue with the relaxations of your problem. In this case, make sure your model does not contain conditional statements or loops with a number of iterations not known a priori. If you are using relaxation-only constraints, make sure these are actually valid. If the problem persists, please report it to the developers.

MAiNGO reports a domain violation, e.g., division by 0

This can have several causes: there may be a mistake in the model equations, the bounds on the optimization variables may be too loose, or the relaxations (or rather the underlying interval extensions) get too bad when propagating through the model equations. In the latter case, try locating the problem within your model by, e.g., using the mc_print(,) function. There are two different ways to remedy such situations (except for reformulating the problem or choosing tighter bounds) depending on why the domain violation occurs:
If you know that the argument of interest is within the correct domain for all values of your optimization variables within their box constraints (i.e., without considering additional (linear or nonlinear) constraints), the domain violation is purely caused by weak relaxations. In this case, you can use the functions pos(), neg(), lb_func(), ub_func() and bounding_func() to artificially cut the relaxations (cf. doc/implementedFunctions/Implemented_functions.pdf).
If, however, there are values of you optimization variables within their bounds (that may potentially violate other constraints) that leads to the argument actually violating the domain, the above functions must not be used. Doing so would lead to wrong relaxations and unpredictable behavior. Instead, the points that lead to domain violations need to be excluded via a constraint and the argument can then be restricted using the functions min(,) or max(,) (cf. doc/implementedFunctions/Implemented_functions.pdf).

MAiNGO terminates with a MC++ error saying something about alpha1 and alpha2

There seems to be a problem with the relaxations, please report this to the developers.

MAiNGO finds feasible solutions early on, but the LBD takes forever to converge

This occurs frequently with deterministic global solvers. It is usually a sign for the relaxations being weak. Sometimes it helps to start with tighter bounds on the optimization variables. Otherwise, more effort might be needed to find better relaxations. If not doing so already, you should also consider using range reduction (cf. Section How does MAiNGO work?).

MAiNGO runs forever without finding a feasible point although LBD seems to be converged already

Try using a more powerful UBP solver (e.g., IPOPT or KNITRO), possibly increase UBP_maxStepsPreprocessing. Also set PRE_maxLocalSearches> to 0 in order to enable local searches at the root node. If the problem persists, it might be problematic for local solvers. Try a more powerful solver which may currently not be implemented in MAiNGO and use the point reported by the external solver as an initial point for MAiNGO. Previous experience has shown that one possible cause can be redundant constraints (since these destroy constraint qualifications). If it seems like the lower bound converged, try using simple function evaluations. If your problem contains binary variables, improvements may be needed to the upper bounding strategy that currently does not distinguish between continuous and binary variables except that the latter are checked for integrality at the returned solution point. In this case, please report the problem to the developers.