A Defect Localization Approach Based on Improved Areal Coordinates and Machine Learning

Table 1

Comparisons of simulation results based on different localization methods.

Methods	Coordinate class	Large error rate (%)	Initial localization error (mm)		Error after eliminating the large ones (mm)
			Average value	Standard deviation	Average value	Standard deviation
Traditional method with no measurement error	Abscissa	0	0.004	0.06	0.004	0.06
	Ordinate	0	3.34-14	5.61-14	3.34-14	5.61-14
Traditional method with measurement error ()	Abscissa	12.64	4.43	29.89	1.25	1.23
	Ordinate	8.35	4.15	37.16	1.13	1.04
Traditional method with measurement error ()	Abscissa	21.17	7.34	39.36	1.64	1.32
	Ordinate	15.24	7.06	49.00	1.44	1.23
New method with measurement error ()	Abscissa	2.05	1.07	1.36	0.94	0.90
	Ordinate	7.51	1.78	3.08	1.06	1.01
New method with measurement error ()	Abscissa	6.35	1.76	2.13	1.32	1.10
	Ordinate	14.87	2.84	4.59	1.35	1.17