استراتژی تکامل برای توربین گاز خطا-شخیص
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|21941||2005||9 صفحه PDF||سفارش دهید||2352 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Applied Energy, Volume 81, Issue 2, June 2005, Pages 222–230
The aim of this investigation is to be able to diagnose gas-path faults in gas turbines by minimising the differences between the observed and simulated data for the engine’s behaviour. The simulated data are generated using a known set of faults as the input to the engine-behaviour aero-thermo model and an appropriate objective function is minimised to yield the best solution to the problem. The application of evolution strategy (ES) in the search for this minimum is an effective, flexible, robust and reliable way of solving engine-diagnostics problems. Adopting this approach leads to a considerable reduction in the overall time taken to obtain a convergent solution when compared with that required using a simple genetic-based algorithm.
During the last two decades, new strategies (such as just-in-time, kanban and lean manufacturing) have been developed and introduced into industry and commerce. Simultaneously, the need for the implementation of more cost-effective maintenance procedures has been increasingly recognised, and so concepts such as total quality management (TQM), which includes total productive management (TPM), have been devised. Thus, lifetime costs for many manufactured artefacts (e.g., gas-turbines) have been reduced. It has also been realised that the main barriers to the more widespread successful application of TPM are the out-of-date perceptions (in the minds of those involved in the operation process) of the benefits of optimising maintenance schedules. The goal, even though unattainable, but which can be approached asymptotically, in any planned maintenance programme should be zero breakdowns during the operation of the artefact. A realisation should also occur of the energy wastages and costs that would ensue if a breakdown occurs. One of the means of reducing such disruptions and ensuant cost is to implement an optimal maintenance schedule for diagnosing the causes from the symptoms and the magnitudes of the faults arising because of say the engine being operated and the consequences for the engine’s performance. Gas-turbine engines, that are operated continually, suffer component deteriorations of performance. Traditionally, the use of gas-path analysis (GPA)  has been popular for assessing the behaviours of engines. It employs the aero-thermodynamic relationship between the performances of the engine’s components and expresses the outcome in terms of the deviations from baseline values. In practice, the effectiveness of a GPA is limited by the relevance, number and accuracies of the measurements obtained from the engine. Additionally, factors such as sensor bias need to be taken into account. The possibility of sensor and component faults occurring simultaneously also should be assessed. A GPA requires that the number of engine-behaviour measurements be not less than the number of performance parameters (e.g., each component’s flow-capacity and efficiency) being examined in order for acceptably accurate diagnostics to be achieved. However, such a situation is unlikely to exist for engines while in service. Other techniques, such as the use of an artificial neural-network (ANN), fuzzy logic or the Bayesian belief network (BBN) have been employed with considerable success. Nevertheless, the fundamental problems of accounting for the engine-components’, non-linear behaviour, measurement noise and sensor-bias remain common challenges. The genetic algorithm (GA) optimisation technique has also been employed during the last couple of years and the credibility of the predictions therefrom appears promising. The research described in this report proposes a new strategy to overcome some of the limitations incurred by using the simple GA-based technique.
نتیجه گیری انگلیسی
The described technique uses strategy parameters to control the objective parameters, so making the convergence of the predicted solution faster and more accurate. The variations of the strategy parameters are controlled by a heuristic function and therefore the population moves towards the final solution in a more orderly and logical way, and so make the diagnosis process more cost-effective.