Development of a Vibration-Based Crack Diagnostic Application Using the MANFIS Technique

This paper analyses the effectiveness of a newly developed, multiple-crack diagnostic tool in dynamic structures using the multiple adaptive neuro-fuzzy inference system (MANFIS) technique. The effect of crack characteristics on the vibration responses have been investigated with different boundary conditions using finite element and numerical analysis. The first three relative natural frequencies, the difference of the first three average relative mode shapes, crack locations, and crack depths are used to train the fuzzy and neural controllers of the MANFIS system. The MANFIS controller is comprised of an input layer, a hidden layer and an output layer. The fuzzy segment uses the first three relative natural frequencies and the difference of the first three average relative mode shapes as the inputs. The hidden layer processes the outputs from the fuzzy controller. Finally, relative crack locations and relative crack depths are outputs from the developed MANFIS controller. It is observed that the predicted values of relative crack locations and relative crack depths from the formulated technique are well in agreement with the results from experimental analysis. The proposed methodology demonstrates its capability to be a suitable non destructive technique for fault identification in vibrating structures.