Jung, Henning

The focus of this dissertation relates to the research field of Structural Health Monitoring (SHM) with the aim to develop new methods for damage diagnosis and prediction based on acceleration measurements. The monitoring of bogie components of railway vehicles is chosen as an application example. Railway vehicles are typically designed for an operationel service life of 30 years or more. In order to ensure a safe and reliable operation over this period, regular maintenance actions must be carried out, which nowadays are based on fixed, predefined intervals. These intervals are limited by a certain operating time or after a defined mileage has been reached. Structural damages that occur between two maintenance actions are rarely detected and, in extreme cases, can lead to derailment. Such events are unpredictable and always result in costly corrective maintenance activities. To avoid such derailment scenarios, the intervals are therefore defined according to a conservative approach (short time intervals, less mileage). Increasing the number of maintenance activities not only raises maintenance costs but also leads to inefficient utilization of components, which are often replaced before the end of their technical service life. Therefore, an intelligent, conditionbased maintenance strategy using an SHM system for damage diagnosis and prediction presents a promising approach to reducing the number of maintenance activities while simultaneously increasing safety, reliability, and cost-effectiveness. Structural damages affect the vibration behavior. Railway vehicles are exposed to constant vibrations during operation, primarily caused by the rolling contact between wheel and rail. The main focus of this research thesis is on the automated evaluation and assessment of vibration measurements concerning defective components. For this purpose, a new method based on Stochastic Subspace Identification (SSI) and kernel density estimation is developed and presented. The core of this method is the Modal Parameter Density Estimation Function (MoDE-F). A key innovation is the use of a large number of different stochastic excitation scenarios in combination with kernel density estimation to determine the modal parameters. This approach leads to the minimization of incorrect identification results that arise from a violation of the white noise assumption for the system excitation – as used in the derivation of the SSI –. The automated evaluation and classification of the generated MoDE-F is realized by using the Jensen-Shannon divergence metric (JSDM). Based on the current state of research, the mode of operation is theoretically analyzed and presented. Numerical simulations show the general applicability to realistic operating conditions. The new method is then applied to a real bogie.

Jung, Henning

Filters

Settings

Sort By

Results per page