17 novembre 2023
10h30 – 12h00
en ligne TEAMS
L’inscription est ouverte et gratuite pour tous mais obligatoire
Présentation en anglais
RÉSUMÉ :
Power plants as well as other technical systems are not functioning to the fullest 100 % of the time. Maintenance and inspection on some components require them to be off-line in a scheduled way. However unplanned outages due to failures or external conditions may cause surprises and can lead to situations in which the demand for the plant or system is not satisfied at short term at large costs. The frequency, the average duration and the fraction of time occurrence of such conditions should at least be thought about and studied using qualitative and quantitative RAM (Reliability Availability, Maintainability) models. From experience over a long period helping to improve the (un) availability of power plants, a book was written which is available from the vgbe website at very modest costs. It is a tribute to plant operators and maintenance experts who duly filled in maintenance and failure feedback data hoping that the next plant would be better. Plants have improved by engineers as some examples for newbuildings show. Application of RAM techniques for newbuildings is straightforward using averages for failure frequencies and repair time of components. The amount of mathematics is very modest and professional software assisted with Excel will provide many answers. However, differences between plants not only in terms of component capacities (f.i. 2 * 75 % pumps? 3 * 50 % pumps?) but also in terms of their component failure characteristics should be taken into account. Ageing versus the modelling of teething troubles is still not standard however from vgbe’s KISSY database and subsequent analysis of the raw data in that database the required coefficients are available as is shown. In the presentation degradation-before-failure will be discussed as often mechanical systems do not fail suddenly. There can be a degradation trajectory that, when made visible by measurements, allows to minimize forced unavailability by inspections and overhauls. Such a trajectory may, given sufficient time, also allow other plants and systems to be started in order to meet the demand.
An analysis of raw OREDA data as well as some Licensee Event Records (LERs) data for US nuclear plants helped in gaining understanding of degradations. Finally, some recent efforts in modelling the effects of preventive maintenance (PM) will be presented. A model where after PM a component is as-good-as-new and the bathtub curve starts again is, at least for power plants, a gross simplification of reality. The experience with Arithmetic Reduction of Age (ARA) models as shown in the vgbe R&D project 444 for 204 power plants recently was published in vgbe’s PowerTech Journal. ARA models are worth the effort however for components and subsystems that become “better” in time, one should also think about a model using Fix Efficiency Factors and a model for the appearance of “humps” in the curve for the cumulative number of failures as a function of time. Do such “humps” give an indication that for specific failures a better root cause analysis would improve availability?
BIOGRAPHIE :
Henk Wels reitred as a Sr. Consultant Risk Management & Decision Analysis in 2019 but is still working in the field of RAM with particular interest in power plants. Graduated at Delft University as a Naval Architect & Marine Engineer, he has been working at the dredgers firm Bos Kalis, at the Dutch Ministry of Traffic & Public works (Rijkswaterstaat) and at departments of KEMA, NRG, DNV and DEKRA. Henk has some 35 years of experience in assessing and quantification of the reliability, availability, and maintainability of mechanical and electrical systems. His relation with power plants started in 1988 when a project was initiated by KEMA to gather failure data of all Dutch power plants to improve their forced unavailability. Benchmarking the performance of power plants and their components against other plants, application of FMECA and RCM, maintenance optimization and the qualitative and quantitative modelling of power plants and their crew are his core interests. He is in contact with the power production sector via DNV and via vgbe, not in the least as at vgbe since 1988 data for the unavailability of power plants can be downloaded to MS Excel by vgbe’s database manager for analysis purposes.