Kriging Metamodeling in Rotordynamics: Application for Predicting Critical Speeds and Vibrations of a Flexible Rotor

Table 11

Relative generalization error on the variance of the amplitude for different regression orders and correlation functions. Nonsymmetric case with 600 samples.

	Gaussian	Linear	Exponential	Cubic
Zero-order regression
		1.86	1.38	9.30
		4.09	3.51	4.17
		3.39	1.79	1.53 10^−1
		1.45	8.85	6.52
	3.75	1.01	4.82
	5.28		4.28	3.39
	1.09	1.70	9.23
		1.41	1.07	3.95
First-order regression
		1.31	1.28	8.56
		3.28	3.28	3.71
		3.45	1.81	1.55
		1.44	7.39	5.28 10^−2
	1.27	1.78	5.42
	3.97		4.50	2.02
	5.00	3.89	1.89
		1.60	1.67	7.52
Second-order regression
		1.28	1.26	8.44 10^−3
		3.10 10^−2	3.21	3.57
		3.47	1.88	1.67
		1.44	7.40	5.40
		2.68 10^−5	2.79
		3.33	3.72	1.64 10^−3
		3.03	2.35 10^−5	4.64
		1.46	1.50	7.14 10^−4