MODIFIED CHAIN-CODE COMPUTER VISION TECHNIQUES FOR INTERROGATION OF VIBRATION SIGNATURES FOR STRUCTURAL FAULT-DETECTION

Vibration testing is a viable method for structural fault diagnosis. Different structural dynamic response variables and parameters that may be considered as candidate objects of interrogation are briefly reviewed. The fault diagnosis process is broken down into three parts, and only the detection process is addressed in this paper. The frequency response function obtained by exciting the structure at a selected reference point is utilized as the preferred form of vibration signature to be used for interrogation. The chain code computer vision technique is modified to evaluate the frequency response function signature as a waveform. Signatures are obtained from a laboratory structure in the form of a ribbed plate, similar to a highway bridge. Cracks are simulated in the structure by cutting through splice plates with a jigsaw. By applying the computer vision technique on the signatures and comparing inspection signatures with benchmark signatures, small cracks are detected consistently. Results of the interrogation are interpreted in a graphical manner. In addition, an automated evaluation technique is presented. The robustness of the technique is verified by contaminating signals with synthetic noise. Successful performance of the technique in the presence of noise indicates the potential for fault diagnosis of large outdoor structures. © 1994 Academic Press Limited.