Posters

Abstract

Wind turbines equipped with gearboxes present high mechanical complexity due to the large number of internal components, such as gears, shafts, couplings, and bearings, which are subjected to severe and continuous operational loads. To mitigate failure risks, vibration-based Condition Monitoring Systems (CMS) are essential tools for predictive maintenance, enabling early fault detection and informed decision-making. Early diagnosis allows maintenance teams to assess whether repairs can be performed at the top of the tower or require component removal, directly impacting downtime and operational costs. This paper presents a case study involving a wind farm with 90 turbines, where a recurring failure pattern was identified in the inner race of an intermediate shaft bearing within the gearbox. Initially, the Condition Monitoring Center detected severe vibration signatures in some turbines through manual analysis routines. Subsequent boroscopic inspections confirmed fractures on the inner race of the bearing, all exhibiting the same failure pattern. Once this pattern was established, an artificial intelligence–based tool was used to scan historical vibration data from all turbines in the wind farm, identifying 15 machines with the same defect signature. These findings were later confirmed through boroscopic inspections. The presence of fractured bearings posed a significant risk of gearbox seizure, requiring corrective actions to ensure continued turbine operation. Conventional repair methods would require full gearbox replacement using cranes, at an estimated cost of approximately €475,000 per turbine, resulting in a total cost of about €7.13 million for the 15 affected machines, which was financially unfeasible. To gain time and avoid immediate shutdown, additional vibration data analysis was conducted. The results showed that some turbines were able to operate for approximately one week with fractured bearings while maintaining stable vibration levels. This behavior indicated that, despite the severity of the defect, the turbines could operate under controlled conditions. Based on this evidence, the turbines were allowed to resume operation at 50% of nominal power, minimizing downtime losses while alternative repair solutions were investigated. After approximately ten months, a supplier capable of replacing the bearing directly at the top of the tower was identified. This solution reduced the repair cost to approximately €6,300 per turbine, totaling about €95,000 for all 15 machines, and eliminated the need for crane usage. This approach significantly reduced maintenance costs and allowed the turbines to continue operating safely. Additionally, maintaining turbine operation during this period avoided downtime losses estimated at approximately €24,000 per day. This strategy was only possible due to the continuous data provided by the condition monitoring system. This case study demonstrates how CMS-based predictive maintenance supports asset management by enabling early defect identification, defining safe operational conditions even in the presence of severe faults, and optimizing maintenance strategies to preserve asset service life while achieving substantial cost savings.