Mechanical Systems and Signal Processing www.elsevier.com/locate/jnlabr/ymssp Mechanical Systems and Signal Processing 19 (2005) 615–631 Noise elimination from measured frequency response functions K.Y. Sanliturk*, O. Cakar Faculty of Mechanical Engineering, Istanbul Technical University, 80191 Gumussuyu, Istanbul, Turkey Received 5 May 2003; received in revised form 2 April 2004; accepted 21 April 2004 Available online 27 July 2004 Abstract It is inevitable that measured signals are contaminated with ‘noise’ when a data acquisition system is used for an experimental measurement. It is also well known in modal testing that the quality of measured frequency response functions (FRFs) is adversely affected by noise originating from test environment as well as electronic devices. This situation often leads to serious difficulties in many applications as they require high-quality FRF data. This paper presents a method based on singular value decomposition (SVD) for the elimination of noise from measured FRFs so as to improve the quality of the measured data. It is believed that this method will enhance the accuracy as well as the success rate of various applications that rely on measured FRFs. r 2004 Elsevier Ltd. All rights reserved. 1. Introduction Measured frequency response functions (FRFs) are used for many purposes including: verification of the theoretical models, model updating, structural modification, determination of excitation forces, fault detection as well as solving general vibration and noise problems. Apart from a few exceptions, it is extremely desirable to acquire high-quality data from vibration tests for many areas of applications. In fact, high-quality FRF data are demanded in almost any application involving matrix inversion and/or differencing operation. However, there are some unavoidable experimental errors and error sources originating from the experimental set-up itself [1–4]. One of the unavoidable error sources that reduce the quality of the measured FRFs significantly is the so-called ‘noise’. The nature of the contamination categorised as ‘noise’ is rather complex as the contributing factors are diverse and complicated. Some of the dominant ARTICLE IN PRESS *Corresponding author. Tel.: +90-212-293-13-00; fax: +90-212-245-07-95. E-mail address: sanliturk@itu.edu.tr (K.Y. Sanliturk). 0888-3270/$-see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.ymssp.2004.04.005