Multi-plane unbalance estimation using non-collocated sensor responses and model order reduction techniques

Mass unbalance is a frequently encountered fault in turbomachinery. In multi-stage process applications such as centrifugal compressors, the unbalance is distributed across several planes along the rotor centerline. In such cases, the widely researched single-plane unbalance estimation methods are insufficient, and the conventional multi-plane approaches often require the sensors to be collocated with the unbalance planes, meaning the vibrations are measured at the disks or impellers. While this approach yields accurate predictions, it is often impractical due to design constraints and limited sensor accessibility. To that end, this study presents a novel, multi-plane unbalance estimation algorithm, capable of accurate estimations even with non-collocated sensor measurements. The accuracy of the proposed approach is compared with the traditional collocated approach, in combination with three different model-order reduction (MOR) techniques: Guyan reduction, modified Guyan reduction, and the system equivalent reduction expansion process (SEREP). The proposed approach is validated with a test rig consisting of a thin shaft and two centrally mounted disks, using both simulated vibration data and experimental measurements integrated with the reduction methods.