In large-scale production systems, there are many complex decisions relating to a variety of asset utilisation and cost implications. Efficient and optimisation based planning systems can be beneficial in deriving feasible and low cost plans. A production-planning task usually involves the allocation of resources to products/packs and also the determination of product paths through the plant. In most cases, the mathematical models result in a high dimensional mixed integer linear problem formulation. For larger problems, the resulting mixed integer models involve several thousands of integer and continuous variables. Usually, to serve such problems, explicit consideration is given to the large-scale nature of the problem. The solution methodology usually consists of a linear programming embedded within a branch and bound procedure. For realistic and commercial scale problems, these methods are hopelessly time-consuming. In this work, an efficient nondifferentiable optimiser in co-operation with a linear programming system have been used to optimise the integer decisions in planning problems. The current approach is quite general and is capable of representing a very wide range of planning systems. The methodology outlined in this paper has been implemented in a medium term planning system for solution of large-scale industrial problems.