Model-based Monitoring of Rotor Systems

点击翻译按钮获取中文摘要

For many typical applications of rotors, e.g. turbines, compressors and pumps, safety and reliability play an important role as failures could result in a shutdown of entire power plants or airplanes. Yet, many defects can be recognized before they get critical as they already influence the behavior of the rotor when they are small. So, condition monitoring could be used to search for these signs and to warn against the arising damage.Most of the currently applied monitoring systems only use measured data for diagnosis. Therefore, it is difficult to specify the location and amount of error and the future behavior of the rotor. This information can be gained when adding a simulation model of the damaged rotor. This is called the model-based monitoring.In this research project, an autonomous system for model-based monitoring is developed. After performing an installation phase, the system automatically monitors the vibrations of the system. If they differ from normal operation, the measurement data will be recorded and compared with typical patterns of errors which are stored in a damage library. Based on the results of this error classification, a suitable simulation is started in order to optimize the fault parameters. The optimized simulation allows a detailed damage diagnosis.Contemporary, the damage library includes unbalance, static bow, angular misalignment and rubbing due to rotor drops. In the following examinations, the damage library should be expanded. The focus lies on further misalignment phenomena. They especially occur in rotors with multiple bearings like they are common in power plants. It should be investigated how misalignment within the pedestal can be allocated to specific bearings during identification and how its amount can be quantified.The second expansion of the library should include misalignment in rigid couplings. In this case, the misaligned components rotate with the rotor, so this error has different characteristics than misalignment in the pedestal. In fact, it is quite similar to static bow of the rotor. There exist two kinds of misalignment in clutches, which are an angular mismatch between the plane faces and an offset in radial direction.In order to examine these failures, the existing test rig has to be updated. It should also be possible to include different amounts of the errors. For the signal-based fault classification, typical patterns for these errors must be found. The error models in simulation should be as simple as possible for fast error identification, yet they must describe the error characteristics correctly. For further verification, several setups should be tested, e.g. small modifications of the test rig or the superposition of several errors.