Intelligent Diagnosis and Smart Detection of Crack in a Structure from its Vibration Signatures

Abstract

In recent years, there has been a growing interest in the development of structural health monitoring for vibrating structures, especially crack detection methodologies and on-line diagnostic techniques. In the current research, methodologies have been developed for damage detection of a cracked cantilever beam using analytical, fuzzy logic, neural network and fuzzy neuro techniques. The presence of a crack in a structural member introduces a local flexibility that affects its dynamic response. For finding out the deviation in the vibrating signatures of the cracked cantilever beam the local stiffness matrices are taken into account. Theoretical analyses have been carried out to calculate the natural frequencies and mode shapes of the cracked cantilever beam using local stiffness matrices. Strain energy release rate has been used for calculating the local stiffness of the beam. The fuzzy inference system has been designed using the first three relative natural frequencies and mode shapes as input parameters. The output from the fuzzy controller is relative crack location and relative crack depth. Several fuzzy rules have been developed using the vibration signatures of the cantilever beam. A Neural Network technique using multi layered back propagation algorithm has been developed for damage assessment using the first three relative natural frequencies and mode shapes as input parameters and relative crack location and relative crack depth as output parameters. Several training patterns are derived for designing the Neural Network. A hybrid fuzzy-neuro intelligent system has been formulated for fault identification.
The fuzzy controller is designed with six input parameters and two output parameters. The input parameters to the fuzzy system are relative deviation of first three natural frequencies and first three mode shapes. The output parameters of the fuzzy system are initial relative crack depth and initial relative crack location. The input parameters to the neural controller are relative deviation of first three natural frequencies and first three mode shapes along with the interim outputs of fuzzy controller. The output parameters of the fuzzy-neuro system are final relative crack depth and final relative crack location. A series of fuzzy rules and training patterns are derived for the fuzzy and neural system respectively to predict the final crack location and final crack depth.To diagnose the crack in the vibrating structure multiple adaptive neuro-fuzzy inference system (MANFIS) methodology has been applied. The final outputs of the MANFIS are relative crack depth and relative crack location. Several hundred fuzzy rules and neural network training patterns are derived using natural frequencies, mode shapes, crack depths and crack locations.
The proposed research work aims to broaden the development in the area of fault detection of dynamically vibrating structures. This research also addresses the accuracy for detection of crack location and depth with considerably low computational time. The objective of the research is related to design of an intelligent controller for prediction of damage location and severity in a uniform cracked cantilever beam using AI techniques (i.e. Fuzzy, neural, adaptive neuro-fuzzy and Manfis).