Decision Support

Recent work for the British Army has focussed on the peridicities for checking to ensure structural integrity of armoured fighting vehicles in the face of ballistic attack. For a number of RCM reasons this check is reactive - and the frequency of the check is based on a number of implicit risk issues. A tool for calculating a defensible frequency for each structurally-significant item (SSI) has been developed in Visual Basic using the Simple Multi-attribute Rating Technique (SMART); this tool displays the SSI inspection intervals for any chosen vehicle zone (e.g. engine compartment, gearbox compartment, driver's compartment, etc) in ascending inspection interval order and this allows a suitable inspection interval for the zone as a whole to be established. The tool can also be used in reverse by evaluating the risk/cost tradeoff in adjusting the inspection intervals.

We carried out studies for the Royal Navy to establish the optimal upkeep cycle of a major class of warship, based on RCM analyses of the class. Using genetic algorithm (GA) techniques on a mathematical model consisting of more than 32,000 data items, the optimal upkeep cycle for this class of vessel was established, together with estimates of upkeep period duration and dockyard resource loading. This model is now coded in Visual Basic to interact directly with the platform RCM database and a modified version is under consideration by the Royal Australian Navy.

In a series of studies we have demonstrated that such GA techniques could also yield useful results in optimal spares allocation. These studies showed what spares should be carried on board and at locations further back in the logistics chain to support an RCM-derived maintenance strategy. The modelling approach also estimates the degree of uncertainty associated with the final result, given the qualitative nature of input data such as reliability estimates, costs and consequence of stockouts.

In the classic 'bearings only' target motion analysis scenario we have also shown that GA techniques provide solution convergence that is comparable with non-linear least squares and Kalman filtering methods, whilst avoiding the need for complex matrix mathematics.

We have used Kalman filtering techniques to 'observe' otherwise unmeasurable plant data - for example, to estimate seawater exit temperature within a seawater-cooled heat exchanger system, where only seawater ambient temperature and freshwater inlet/outlet temperature measurements are known. In this arrangement, the heat transfer efficiency - and hence tube stack condition - can be estimated and this provides a reliable trigger for tube cleaning maintenance.

Our decision support experience also includes a number of studies using artificial neural network approaches to system modelling within a fault diagnostic arrangement; these approaches have shown a high degree of solution robustness in the presence of noise, together with short model development times.

You are visitor number:

©2011 Steven Consultants Ltd

Site last updated Sept 2011