Ultrasonic Guided Waves-Based Monitoring of Rail Head: Laboratory and Field Tests

Table 3

Test 1. Multivariate analysis: percentage of outliers for various transducer pairs, frequency bandwidths, and damage types. The best and the worst combination of wavelet processing and damage index features are presented.

			Defect type
			Transverse	Longitudinal	Oblique	Internal
DATA	Best/Worst	Combination	Success rate
Central sensors high-freq	B	c_I_(1-3,6,9)	92.8%	48.8%	22.5%
	W	b_II_(9,11)	57.1%	0%	2.5%
Central sensors low-freq	B	a_IV_(1-3,5-11)	93.2%	65%	78.8%
	W	a_II_(5,9)	41.8%	0%	0%
Gage sensors high-freq	B	a_II_(1-4,8-9)	58.1%	0%	41.7%
	W	a_IV_(3,5)	23.5%	0%	5%
Gage sensors low-freq	B	a_III_(1-3,5-11)	70.4%	8.8%	68.3%
	W	a_IV_(3,5)	33.8%	0%	13.3%
Central sensors high-freq	B	c_I_(1-2,4-9,11)				57.1%
	W	b_II_(4,5)				0%
Central sensors low-freq	B	f_IV_(1-3,5-6,8-9,11)				77.9%
	W	b_I_(1,2,10)				0%
Gage sensors high-freq	B	f_I_(1,4-9)				41.7%
	W	a_IV_(3,5)				0%
Gage sensors low-freq	B	d_III_(2-11)				60%
	W	a_IV_(3,5)				0%