Estudiante

Proceedings of the 12th WSEAS International Conference on AUTOMATIC CONTROL, MODELLING & SIMULATION

Design of Experiment and Montecarlo Simulation as Support for Gas Turbine Power Plant Availabilty Estimation
ENRICO BRIANO  CLAUDIA CABALLINI * PIETRO GIRIBONE # ROBERTO REVETRIA #  DIP CONSORTIUM Office Tower, Voltri Distripark Europe, 16158 Genoa, ITALY * CIELI – Italian Centre ofExcellence in Integrated Logistics Via Bensa 1, 16132 Genoa, ITALY # DIPTEM – Department of Industrial Production, Technology, Engineering and Modelling Via Opera Pia 15, Genoa, ITALY enrico.briano@dipconsortium.org; claudia.caballini@cieli.unige.it; piero@itim.unige.it; roberto.revetria@unige.it

Abstract - Maintenance is an important aspect in order to guarantee the efficiency of industrial facilities.For power plants the high availability ratios can be obtained only with preventive maintenance but the result costs increases rapidly. In order to reduce the cost level of the maintenance activity, on-condition maintenance is carried out on an increasing subset of components. Only using appropriate reliability models can identify the optimal mix between preventive and on-condition maintenance.Some time the data required by such models are inadequate or missing and the performances of the implemented system can fall quite rapidly. The authors propose an innovative approach based on a hierarchical Montecarlo simulator able to estimate properly the power plant reliability and, at the same time, improve its performances by a fine-tuning of its parameters. A real life case study is thanpresented and discussed.

Key Words - Reliability, Fuzzy Logic, Model Estimation, Montecarlo Simulation, Design of Experiment

1. Introduction
The availability and reliability of a complex system is influenced by a wide range of stochastic factors (i.e. component failures, control breakdowns, etc.), then it is difficult to create “ad hoc” reliability analytical models allowing simulation as theonly approach. Particularly Montecarlo simulation has proven to be very effective in the evaluation of the general availability ratio as well as a way to improve the maintenance plans. Maintenance simulation, however, involves several complex aspects (i.e. model conceptualisation, data collection, statistical analysis on input data, forecasts, etc.) that require ad hoc approach to be solved. Lack ofdata is one of such key aspect that engineers and managers have to face in order to provide a real useful simulation. Among the various aspect of the maintenance related to economical aspects (spare parts inventories, maintenance workload, etc.) two key parameters have to be properly estimated: • MTBF: Mean Time Between Failure; • MTTR: Mean Time To Repair. The MTTR parameter is easy to estimatesince it can be obtained a priori from a work schedule or as a summary of the performed maintenances, while the MTBF is some time very hard to estimate. Since MTBF can be obtained only from sampling real life components or by applying complex analytical models, it is generally available only for high standardized products (i.e. lamp bulbs, microchips, etc.). The proposed approach involves thedesign and implementation of a general-purpose hierarchical Montecarlo simulator in which are known some of the MTBF parameters and a procedure for the estimation of the unknown one. By using Analysis of Variance (ANOVA) techniques the authors successfully identified a configuration of the simulator able to reproduce the behaviour of a real power plant.

ISSN: 1790-5117

223

ISBN:978-954-92600-1-4

Proceedings of the 12th WSEAS International Conference on AUTOMATIC CONTROL, MODELLING & SIMULATION

2. The Implemented Approach
The values of the MTBF for each component in each possible failure configuration are generally known in term of probabilistic distributions, in fact several studies were carried out by many authors in order to define a closed form for such distribution, but...