To better govern inventory control of spare parts in tandem with scheduling of maintenance operations in a railway setting, a Railway Spare Inventory & Maintenance Scheduling (R-SIMS) model is developed. To suit the railway industry setting the model includes an inventory control strategy which envokes continuous monitoring and a reorder level as well as an order up-to level. Moreover a condition based maintenance (CBM) strategy is used with periodical inspections. These strategies are combined in a simulation model which assumes stochastic step-wise deterioration of parts as well as stochastic lifetime lengths to predict the expected cost per part per period. The three maintenance scheduling and spare inventory decisions to be minimised in terms of their resulting costs are 1) when to replace parts based on their condition, 2) when to buy new spare parts and 3) how many spare parts to procure at each order. These decisions are represented by values of the three decision variables: the CBM threshold L_p, the reorder level s and the order up-to level S. The minimisation of these decision variables is performed through surrogate modelling, a branch of machine learning optimisation techniques which deals with complex black-box functions. It does so by running experiments and estimating a surrogate function which in turn is optimised mathematically. Multiple surrogate modelling methods are compared in experiments to test their performance, speed and stability. These experiments showed that the Tree-structured Parzen Estimator algorithm as used in the HyperOpt Python library was one of the best performing methods (only rivaled by the MVRSM algorithm), moreover it was shown to be the fastest as well as the most stable of the tested algorithms for this particular problem application. This research contributes to the existing theory by creating a modelling framework for the joint optimisation of spare inventory and maintenance scheduling decisions which is specifically tailored to the railway context and includes possibilities for minimal maintenance. In practice this model can be used by maintenance providers to increase the financial success of their maintenance operations and by infrastructure managers to increase insight into their maintenance providers' operations.