Pipeline Leak Aperture Recognition Based on Wavelet Packet Analysis and a Deep Belief Network with ICR

Table 1

Simulation results of different methods in different leakage apertures.

methods	Neurons number	Recognition accuracy (%)	Average running-time(s)
			Training time	Testing time
	400-200-100-65-5	98.98	42.32	10.25
DBN	400-200-100-65-5	98.7	58.25	15.63
LSTSVM	-	98.58	16.23	4.11
LSSVM	-	98.42	20.61	6.75
SVM	-	98.31	22.15	7.04
BPNN	400-100-5	98.91	15.13	4.41