A methodology is proposed for implementing logic and engineering knowledge within a Branch-and-Bound algorithm and with a purpose to accelerate convergence. The development addresses assignment problems of utility networks with an emphasis on the optimal allocation of units for maintenance problems. Proposed criteria are presented to automatically tailor the solution strategy and fully customise the optimisation solver. Model and problem properties are exploited to reduce the solution space, prioritise the branching of nodes, calculate lower bounds, and prune inferior parts of the binary tree. Comparisons with commercial MILP solvers demonstrate the significant merits of customising the solution search engine to the particular solution space. Extraction of knowledge and analysis of operations is conceptually supported by the graphical environment of the Hardware Composites.