On the use of genetic algorithm to optimize industrial assets lifecycle management under safety and budget constraints
- EDF R and D, Departement MRI, 6, quai Watier, 78401 Chatou cedex (France)
The purpose of this paper is to describe the method and tool dedicated to optimize investments planning for industrial assets. These investments may either be preventive maintenance tasks, asset enhancements or logistic investments such as spare parts purchases. The two methodological points to investigate in such an issue are: 1. The measure of the profitability of a portfolio of investments 2. The selection and planning of an optimal set of investments 3. The measure of the risk of a portfolio of investments The measure of the profitability of a set of investments in the IPOP tool is synthesised in the Net Present Value indicator. The NPV is the sum of the differences of discounted cash flows (direct costs, forced outages...) between the situations with and without a given investment. These cash flows are calculated through a pseudo-Markov reliability model representing independently the components of the industrial asset and the spare parts inventories. The component model has been widely discussed over the years but the spare part model is a new one based on some approximations that will be discussed. This model, referred as the NPV function, takes for input an investments portfolio and gives its NPV. The second issue is to optimize the NPV. If all investments were independent, this optimization would be an easy calculation, unfortunately there are two sources of dependency. The first one is introduced by the spare part model, as if components are indeed independent in their reliability model, the fact that several components use the same inventory induces a dependency. The second dependency comes from economic, technical or logistic constraints, such as a global maintenance budget limit or a safety requirement limiting the residual risk of failure of a component or group of component, making the aggregation of individual optimum not necessary feasible. The algorithm used to solve such a difficult optimization problem is a genetic algorithm. After a description of the features of the software a test case is presented showing the influence of the optimization algorithm parameters on its efficiency to find an optimal investments planning. (authors)
- Research Organization:
- American Nuclear Society, 555 North Kensington Avenue, La Grange Park, IL 60526 (United States)
- OSTI ID:
- 22212714
- Resource Relation:
- Conference: M and C 2013: 2013 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, Sun Valley, ID (United States), 5-9 May 2013; Other Information: Country of input: France; 7 refs.; Related Information: In: Proceedings of the 2013 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering - M and C 2013| 3016 p.
- Country of Publication:
- United States
- Language:
- English
Similar Records
Stochastic Analysis for Long Term Capital Structures, Systems, and Components Refurbishment and Replacement
Conventional Generation Asset Management with Renewable Portfolio Standards Using Real Options