An Effective Fault Feature Extraction Method for Gas Turbine Generator System Diagnosis

Table 5

Accuracies of SVM classifier under various kernels of KPCA and mother wavelets.

Mother wavelet	Level	KPCA with linear kernel		KPCA with RBF kernel and		KPCA with polynomial kernel and
		Accuracy	Feature number	Accuracy	Feature number	Accuracy	Feature number
Db3	3	95.48%	34	91.12%	86	67.22%	22
	4	94.20%	62	90.08%	110	64.20%	28
	5	95.13%	110	81.24%	132	87.36%	45
Db4	3	95.12%	25	88.24%	87	70.00%	18
	4	96.26%	54	87.34%	106	81.00 %	27
	5	95.55%	105	83.42%	132	74.26%	39
Db5	3	84.40%	26	82.53%	89	66.36%	20
	4	93.23%	54	85.32%	110	64.06%	27
	5	93.49%	105	80.00%	132	81.04%	38
Db6	3	94.20%	26	84.05%	98	68.34%	19
	4	95.32%	53	85.64%	104	66.78%	24
	5	93.78%	102	88.04%	130	85.30%	38
Average		93.85%	—	85.59%	—	73%	—