Vibration Analysis on Flexible Pipe Connectors of Parabolic Trough Solar Collectors for Predictive Maintenance

Kurzfassung

Parabolic Trough Collectors (PTC) are the most proven concentrating solar power (CSP) technology. The part of a PTC that is typically the first to fail is the rotation and expansion performing assembly (REPA), which is a flexible pipe connection between the movable receiver tube and the stationary field tubing. The combination of high temperatures, pressures and mechanical loads due to the movement makes this component prone to failure. A malfunction of the REPA can lead to a leakage of the hot heat transfer fluid, potentially causing fires and shut downs of the collector row or the whole power plant. The aim of this work is to develop a predictive maintenance strategy based on the vibrations that occur in the REPAs. A previous work investigated the change in the vibration patterns over the life cycle of REPAs in PTCs for condition monitoring. This thesis expands on this idea by using the change in vibrations to predict the REPAs future malfunctions. To distinguish between the vibrations of a normally functioning REPA and a REPA shortly before a leak occurs, a convolutional neural network is trained for classification. The vibration data used for training is collected in an accelerated life cycle test campaign and analyzed using short time Fourier transform. Two kinds of REPA are tested, first a rotary flex hose assembly and later a ball joint assembly. By employing a machine learning classification model, REPA malfunctions were predicted at least 1000 cycles before they occurred, which translates to more than 2.5 years of operation. This proves, that installing vibration sensors in PTC power plants can help making the operation safer and more cost effective.