A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2018; you can also visit <a rel="external noopener" href="https://res.mdpi.com/def5020075118eacc4bbef474af105d2ee82aee08c4ad06fc51850ab125c3ec264b6668bc71e46f71011dfd88174de19cc8ea01c9b29ad94ecdbccde89363ad9080b9d631cb579df8d43d97c810c359a3ccd6d9d059a59cd29bb9c9483df427c980cbb54df5187b1ff2a5acbfe50bd8084f397e42ae14b21fc743d93d52da970f9c243d4890852237a287f67e632aff2aea8e7?filename=&attachment=1">the original URL</a>. The file type is <code>application/pdf</code>.
<i title="MDPI AG">
<a target="_blank" rel="noopener" href="https://fatcat.wiki/container/a2yvk5xhdnhpxjnk6yd33uudqq" style="color: black;">Energies</a>
This article offers reasons to defend the use of generic behavior models as opposed to specific models in applications to determine component degradation. The term generic models refers to models based on operating data from various units, whereas specific models are calculated using operating data taken from a single unit. Moreover, generic models, used in combination with a status indicator, show excellent capacity for detecting anomalies in the equipment and for evaluating the effectiveness<span class="external-identifiers"> <a target="_blank" rel="external noopener noreferrer" href="https://doi.org/10.3390/en11040746">doi:10.3390/en11040746</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/mpwzp6fezrdzxmyd2zjpetewva">fatcat:mpwzp6fezrdzxmyd2zjpetewva</a> </span>
more »... f the maintenance actions, resulting in lower development and maintenance costs for the operating firm. Artificial neural networks and moving means were used to calculate the degradation indicators, based on the remainders in the model. The models were developed from operating data from fourteen wind turbines monitored over several years, and applied to the detection of faults in the bearings on the non-drive end of the generator. The use of generic models may not be recommendable for detecting faults in all cases, and the suitability will depend greatly on the context of the methodology developed to detect each type of fault, according to the element causing the fault and the fault mode, since each methodology requires a greater or lesser degree of precision in the model. Energies 2018, 11, 746 2 of 22 financial costs, etc. By using artificial intelligence techniques in maintenance, it is possible to identify failure patterns in equipment and thus anticipate possible failures     with the final purpose of increasing the life cycle of the wind turbines. The principle of condition-based maintenance  , considers that if it is possible to identify that a component is degraded and might fail within a given period of time, then preventative maintenance can be performed before that failure actually occurs. In the case of a wind turbine, that means increasing the energy produced, since the maintenance task can be performed at times when any energy that might be generated by the wind is negligible or non-existent. At the same time, it enables maintenance costs to be cut by preventing greater damage were the failure to actually occur. Knowing what is going to fail makes it possible to optimize spare part management  and reduce the logistical waiting times involved in performing the maintenance task. Moreover, by using these techniques it is possible to extend the useful service life of the assets  according to the knowledge of the failure. Condition-based maintenance practice can be achieved based on three stages : in a first stage the detection process identify anomalies in the behavior of the equipment  . Once it has been determined that the equipment is not working under normal conditions, it is necessary to identify the type of anomaly occurring in the machine [15, 18,    , this second knowledge stage is named diagnostics. Finally, once the detection and diagnosis have been done, it is possible to estimate how the degradation of that specific failure mode will evolve and subsequently when the failure will probably happen (the third stage, prognosis) [22, 25, 26] . Failure modelling is an extremely complex science; very few data are typically available for failure incidents and, generally speaking, the appearance and development of the failures varies greatly. For that reason there are studies that seek to construct a physical model of the equipment and its control system to simulate failures in order to evaluate the behavior under failure conditions. These methods allow one to design robust fault tolerant control strategies for wind turbines. Most of these methodologies are supported by fuzzy logic and they are useful to train a diagnosis system and isolate the failure [27, 28] . In this field, several studies were conducted: simulations of sensor failures and multiple failures were studied in [29, 30] ; an adaptive fuzzy system for fault tolerant control and cooperative control were applied to detect blade erosion and debris build-up in  ; and two active power control schemes are developed based on adaptive pole placement control and fuzzy gain-scheduled proportional-integral control approaches in  . These studies represent the evolution to condition base operation, and knowing the real heath of equipment and isolating its failure one can adapt the control strategy to extend the remaining life of the equipment, as it was presented in  . Other techniques, that were used to evaluate the health of mechanical equipment, were introduced in , where the authors combined vibration analysis and normal behavior temperature model using artificial neural networks (ANNs). An ANN allows training a black box model without knowing the physical behavior of the equipment, which is extremely complex. The combination of SCADA systems, databases and ANNs are becoming popular in artificial intelligence development due to the fact a deep knowledge of the physical rules that are involved in the process is not needed. Previous contributions have analyzed the use of SCADA data and ANN to create normal behavior models to detect malfunctions: an intelligent system for predictive maintenance applications to health condition monitoring in wind turbines was presented in  . They created normal behavior models and bounds, nevertheless they do not compare different model configurations to improve the goodness of fit the model, nor the utilization of generic models versus specific models. Each equipment is unique and a single model must therefore be generated for each one, based on its specific data  . However, data is not generally available for all units. When the unit is new, there will not be enough data to allow it to be modelled until a full year has passed, in order to have data for each season. This behavior model also varies and has to be recalculated whenever maintenance is carried out on the unit. On the other hand, in operational processes it is not desirable to wait for enough data to be available, it is easy to fall into the trap of using models built from limited historical data from a period of time with high failure risk due to premature damage. As a result, the model will be very imprecise and more
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20180727075329/https://res.mdpi.com/def5020075118eacc4bbef474af105d2ee82aee08c4ad06fc51850ab125c3ec264b6668bc71e46f71011dfd88174de19cc8ea01c9b29ad94ecdbccde89363ad9080b9d631cb579df8d43d97c810c359a3ccd6d9d059a59cd29bb9c9483df427c980cbb54df5187b1ff2a5acbfe50bd8084f397e42ae14b21fc743d93d52da970f9c243d4890852237a287f67e632aff2aea8e7?filename=&attachment=1" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/d5/42/d5424342aa8c92d8e374d1f2483290cbe485146b.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener noreferrer" href="https://doi.org/10.3390/en11040746"> <button class="ui left aligned compact blue labeled icon button serp-button"> <i class="unlock alternate icon" style="background-color: #fb971f;"></i> mdpi.com </button> </a>