Fault Detection for Multimodal Process Using Quality-Relevant Kernel Neighborhood Preserving Embedding

Table 2

The monitoring accuracy rate, false alarm rate, missing alarm rate based on LLE, KPLS and QKNPE methods.

Failure	Method	Statistics	Accuracy	Rate of false alarm	Rate of missing report
1	LLE		73.07	1.51	30.82
			99.60	3.02	0
	KPLS		99.03	6.73	0
			98.93	5.13	0
	QKNPE		99.93	0.50	0
			99.53	3.52	0
2	LLE		99.40	2.01	0.09
			86.93	2.01	17.08
	KPLS		97.40	6.22	0
			58.07	3.56	55.86
	QKNPE		99.80	0.75	0
			99.07	3.26	0.09
3	LLE		98.87	2.43	0
			61.60	1.29	70.79
	KPLS		95.67	6.98	0
			96.45	2.30	4.57
	QKNPE		99.60	0.86	0
			98.73	2.72	0
4	LLE		63.00	3.23	87.52
			84.33	1.67	36.61
	KPLS		91.07	8.42	8.65
			92.05	1.88	14.91
	QKNPE		96.27	0.89	7.99
			95.07	2.34	8.82
5	LLE		76.27	5.59	96.01
			86.87	1.50	59.47
	KPLS		87.80	11.09	11.33
			92.93	2.13	18.24
	QKNPE		96.87	1.25	10.63
			95.60	2.42	12.29
6	LLE		88.53	5.58	93.07
			92.07	1.43	98.02
	KPLS		87.42	11.19	16.41
			94.15	2.41	31.58
	QKNPE		97.60	1.43	25.84
			96.20	2.64	19.80