Table 4.

Two-step Rolling Time Horizon Analysis: Selected Model Performances for Fault 7 (easy), 8 (moderate), and 15 (challenging)

Fault	Sample Bin	Fault Detection Rate	Optimal Feature Subset Size	Accuracy	False Alarm Rate	AUC
Fault-7 Feed Temperature Change	19–20*	0.83	21	0.92	0.00	1.00
	19-20	0.83	10	0.91	0.00	1.00
	39-40	1.00	1	1.00	0.00	1.00
	59-60	1.00	1	1.00	0.00	1.00
	79–80*	1.00	2	1.00	0.00	1.00
	79-80	1.00	1	1.00	0.00	1.00
	99–100*	1.00	6	1.00	0.00	1.00
	99-100	1.00	1	1.00	0.00	1.00
	119-120	0.75	1	0.88	0.00	1.00
Fault-8 pH Sensor Drift	19–20*	0.83	52	0.85	0.13	0.93
	19-20	0.83	49	0.85	0.12	0.93
	39-40	0.98	4	0.99	0.00	1.00
	59–60*	0.97	3	0.99	0.00	1.00
	59-60	0.97	2	0.98	0.00	1.00
	79–80*	0.98	32	0.99	0.00	1.00
	79-80	0.98	9	0.99	0.00	1.00
	99–100*	0.98	66	0.99	0.00	1.00
	99-100	0.98	3	0.99	0.00	1.00
	119–120*	0.99	70	1.00	0.00	1.00
	119-120	0.99	1	0.99	0.00	1.00
Fault-15 Contamination	19–20*	0.57	204	0.58	0.41	0.61
	19-20	0.57	201	0.58	0.41	0.61
	39-40	0.94	58	0.96	0.02	0.99
	59–60*	0.99	57	1.00	0.00	1.00
	59-60	0.99	18	0.99	0.01	1.00
	79–80*	0.99	72	1.00	0.00	1.00
	79-80	0.99	42	0.99	0.00	1.00
	99–100*	1.00	77	1.00	0.00	1.00
	99-100	1.00	5	1.00	0.00	1.00
	119–120*	1.00	145	1.00	0.00	1.00
	119-120	1.00	8	1.00	0.00	1.00