< Back to previous page

Project

Experimental Dynamic Substructuring: Towards Robust FRF-based Decoupling Strategies

Wind turbine manufacturers and their suppliers are faced with increasing constraints on the noise and vibration requirements of their products, while development time and costs are under pressure. One of the problems in the current development is the end-of-line validation of prototype gearboxes. It is observed that the strict limits imposed on the end-of-line test rig do not always lead to acceptable vibration levels when the gearboxes are installed in the wind turbine. This dissertation investigates how wind turbine gearbox behaviour is correlated in test and field conditions and how measurements can be made less test rig dependent. 

The state-of-the-art of substructuring is used to give an answer to the first research question. Experimental substructure decoupling has been identified as a potential tool to compensate for the influence of the test rig on the vibrational behaviour of the gearbox. Despite some clear advantages over experimental substructure coupling, the decoupling methodology faces a number of difficulties that prohibit large scale applicability. 

In this dissertation, these difficulties are systematically pointed out and solutions are proposed. A relatively simple test rig is used to define an optimal measurement approach and to formulate conditions under which experimental decoupling is feasible. Measurement noise is one of the main problems in substructuring as errors are strongly amplified when the FRF matrices are inverted. A number of existing filtering methods are presented and their ability to filter noise while guaranteeing a good fit to the original FRF is assessed. 

Based on these results and observations during the measurements, a new filtering approach is presented that not only filters random measurement noise, but also corrects for systematic measurement errors. The new method uses each individual measurement, a column in the FRF matrices, as a basis to recalculate the full FRF matrices. Because many FRF matrices can be calculated this way, the decoupling solution consists of many individual solutions of which the average can be used as the decoupled FRF. The sensitivity of the experiment and the uncertainty within the measurements is reflected by the spread on the results.

One of the characteristics of the decoupling problem is that a certain freedom exists in the choice of measurement locations on the test rig. However, despite being critical in the success of experimental decoupling, a clear strategy on the selection of measurement locations has not yet been defined. This dissertation has developed an intelligent algorithm that can select the optimal measurement locations while requiring little accurate knowledge about the test rig and component.

Date:1 Oct 2012 →  27 Nov 2017
Keywords:Dynamic substructuring
Disciplines:Control systems, robotics and automation, Design theories and methods, Mechatronics and robotics, Computer theory
Project type:PhD project