Fuzzy logic expert system for automated damage detection from changes in strain energy mode shapes

In recent years, researchers have developed several methods for damage detection in structures employing strain energy or curvature mode shapes. Experience shows that mode shape methods are highly sensitive to measurement noise. Such sensitivity is a direct result of the second derivative applied to the displacement mode shapes, which produces curvature mode shapes. Calculation of strain energy mode shapes includes calculation of curvature as well. Consequently, the artifact false peaks (normally an indicator of damage) appear on the strain energy mode shapes, reducing accuracy of damage recognition. Combined with the intrinsic variance in damage peak shapes and amplitudes dependent upon magnitude and location of damage, the false peaks create a major problem for automated detection of damage location. Most often, a highly qualified expert is required to analyze the data and to identify and locate damage. In an attempt to replace the human expert with a computer, this paper describes a fuzzy logic expert system designed to mimic the human decision process. An expert system allows for easy encoding of expert knowledge as a set of rules. "Fuzziness" of the expert system allows better treatment of the uncertainties of the problem and to simplify the expert system itself. The fuzzy expert system has been designed based on a finite element (FE) model of a simple beam and has provided reliable detection of damage for every tested damage scenario (100% recognition). The same system recognized damages with 100% accuracy, and no false positives or negatives on mode shapes acquired by impact testing and/or non-contact laser vibrometer techniques.