All systems are unreliable in the sense that they degrade with age and usage and eventually fail. The degradation depends on the load (or stress) on the system and the maintenance actions. With higher load the productivity increases but this is achieved at the expense of faster degradation and more frequent failures. This in turn reduces the availability of the system. The yield (expected output per unit time), expected operational cost and the expected profit are functions of the load and the availability. The relationship between load and yield is a complicated function and depends on system reliability and maintenance actions. The thesis deals with determining the optimal load and maintenance actions to maximise the yield (or expected profit) or to minimise the expected cost. The thesis comprises of two parts (Parts I and II). Part I is theoretical in nature and looks at the modelling, analysis and optimisation issues. Part II deals with a real case study involving a dragline used in open cut mining. Part I starts with stress-life models which link the load on the system with the reliability of the system. It examines the empirical data in the form of S-N and P-S-N curves, and as WPP plots reported in the literature. The salient features are identified so as to assess the suitability of different models to model the stress-life data. Various Weibull models are assessed to determine their suitability to model stress-life data at the component level. This is followed by a discussion of the modelling at the system level. Optimal maintenance actions are derived based on component level and system level stress-life models. For the component level models, the optimal maintenance actions are derived for a variety of maintenance strategies (such as the age, block and periodic policies) for a specified load to minimise the asymptotic expected cost per unit time. Following this, the joint optimal choice of load and maintenance is examined. For the system level model, the optimal load is obtained to minimise the expected cost per unit time over the life of the system taking into account different actions such as minor and major preventive maintenance actions and corrective maintenance actions. Part II deals with a case study involving a BE1370W dragline. The dragline is used in open cut coal mining to remove the dirt (overburden) on top of the coal seam. This involves dragging the bucket, which is suspended from the boom, through the muckpile until full and then hoisting and swinging the bucket to the dumping point. The capacity of the bucket size determines the load on the dragline. The system level model studied in Part I is used to determine the optimal load to maximise the yield. The dragline was modelled as comprising of 25 components (or sub-systems). Each component was modelled using a system level model of Part I. The model parameters were estimated using real operating and maintenance data over two years. The optimal load to maximise the yield was derived and a sensitivity study carried out.