Posters

Abstract

Predictive maintenance uses data analysis tools to detect anomalies in asset operation and potential damage in equipment, so they can be fixed before resulting in failure. Ideally, predictive maintenance allows the maintenance frequency to be as low as possible to prevent unplanned repairs without doing an excessive amount of preventive maintenance. Implementing predictive maintenance requires real-time condition monitoring of the asset state and performance. In the case of wind turbines, real-time condition monitoring is traditionally focused on the drive train components such as shafts, gearboxes, bearings, and generators using accelerometers, strain gages, pressure, temperature, rotational speed sensors. Inclinometers and wave level sensors, just to mention a few, are also used to monitor the condition of the towers and foundations. All these sensors provide important data from components in the wind turbine's nacelle and tower. In contrast, rotor blades are not continuously monitored even though rotor blade failures are one of the most common factors affecting continuous operation of wind turbines both onshore and offshore. Rotor blade failures could result not only in blade loss, but also in damages to the tower or surrounding turbines, in addition to be a safety concern for nearby populations in the case of onshore installations. The common practice to determine blade integrity involves conducting drone-based visual inspections at pre-determined time intervals, ranging from a few months to up to three years. These inspections provide a snapshot of the blade condition at the time of the inspection but no information on the onset and evolution of damage between scheduled inspections. Real-time monitoring of blades based on the acoustic emission nondestructive evaluation method is a leading alternative to detect the onset of blade damage and evaluate its evolution, reducing the need for frequent visual inspection and increasing wind turbine uptime and energy production. The data provided by real-time monitoring are one of the pillars of true predictive maintenance programs. This paper discusses the development and implementation of a wind turbine blade real-time monitoring technology based on acoustic emission. The technology is built around an acoustic emission Internet of Things device capable of detecting active damage on the blades using micro-electro-mechanical acoustic sensors. The signals detected by the sensors are pre-processed by the Internet of Things device and transmitted, via a Wi-Fi access point in the wind turbine nacelle, or via cellular network, to a data-driven web application residing in the cloud. The data processing is completed in the cloud, where the information extracted is used to calculate proprietary quantities that are tracked over time. These proprietary quantities are indicators of the state and evolution of the wind turbine rotor blades structural health and can be used to classify the blades in a wind farm according to their structural condition. This classification can be used to establish a predictive maintenance program for the wind farm. The technology has been implemented in more than one hundred and fifty wind turbines throughout several wind farms across North America and Europe.