Abstract:
Cracks in shafts can be identified as a significant factor for limiting the safe and
reliable operation of machines. Engineers can predict faults using classical approaches.
However, when artificial intelligence approaches are used, the forecasting time for
crack diagnosis improves dramatically. The objective of this study is to detect the
location and depth of the crack in the shaft using a fuzzy logic algorithm. Literature
presents measurements of frequency, mode shape, and structural damping can be used
to assess cracks. However, evaluating mode shape and structural deformation is more
difficult than measuring frequency. Such criteria, however, are insufficiently sensitive
to detect early flaws. This study employs changes in phase angle and natural frequency
as crack indicators. To evaluate the natural frequencies and phase angles of the cracked
shaft utilizing the change in stiffness matrices of the cracked element, theoretical
calculations were performed using Matlab. To verify the theoretical values of natural
frequencies, modal analysis was performed using Ansys. Good agreement is observed
between the results. To detect the location and depth of the crack, the fuzzy logic
technique uses first and second mode natural frequencies and their corresponding phase
angles of the shaft as input parameters. The correlation coefficients for triangular,
trapezoidal, and Gaussian membership functions are all close to one. Also, the average
total errors of the three membership functions with the theoretical values are all less
than 5%. This indicates that results obtained from all membership functions are close
to the theoretical locations and depths of crack. So the proposed fuzzy logic technique
would constitute an efficient tool for real-time crack identification.
