抄録
This research is about improvement of the diagnostic accuracy of the fatigue crack monitoring via the statistical diagnostic method. Our research group proposes an unsupervised damage diagnostic method named SI-F method which diagnoses the damage from detecting the change of correlation between sensors caused by the initiation or propagation of the damage via the statistical evaluation. By the method, correlation between sensors is identified by using the response surface and the change of them is statistically investigated with the F-test. To identify the crack length by the method, identification about the relation between the crack length and the F0 statistic is required. Then in this research, to evaluate effect of the regression error, the noise magnitude and the fluctuation of the external force to the relation, numerical simulation was conducted. For the simulation, two sets of data, one with constant load and one with variable load, are generated and compared. And the applicability of the result of the simulation is experimentally investigated. Finally, the results indicate that the F0 affected by the regression error and the noise magnitude but not affected by the external force.
