The use of object-oriented techniques in software development for total systems engineering projects

Modern traction system design places great demands on system components, parameters and tolerances in order to achieve an ever-increasing level of performance at a competitive price. As a result, the analysis and simulation of traction system components is widely carried out, and advanced modelling techniques exist for the manifold aspects of a complete traction system. In a complete traction system, the interactions between rolling stock, power system, train control, passenger flow etc are extremely complex, and as a result each individual model is forced to accept approximations of the others in its operation. Ideally, an analysis methodology is required which can incorporate the interactions between models, whilst retaining the accurate individual model descriptions. This would allow a complete traction system model to be developed. However, unless the methodology isolates the behaviour of each model from the interactions, an unwieldy complete model results in which a change to any part of the system, such as a change of rolling stock, would require a complete revision of the whole model. In reality, of course, the interactions between rolling stock and the remainder of the system remain unchanged. As a result, the maintenance of the complete model becomes a formidable problem, and this detracts from its usefulness. It is this lack of flexibility, due to software methodologies not reflecting the real system, that has prevented such large models being developed. This point is emphasised if consideration is given to the design and analysis methods used in hardware design. Invariably, when considering the requirements of a new specification, existing systems and approaches are considered, such that they may be modified and re-used to minimise the design effort required. It seems logical, then, that any software methodology used to model hardware systems should reflect the same properties, i.e. the minimum level of change may be carried out to develop new systems. These properties are all demonstrated by using an object-oriented approach. >