Uncertainty Quantification of Vibroacoustics with Deep Neural Networks and Catmull–Clark Subdivision Surfaces

Table 5

DNN prediction statistical results.

Frequency (Hz)	Variables	RMSE	MAE	MAPE
50	1-D radius	7.17E − 04	6.14E − 04	1.02
	1-D thickness	1.24E − 05	9.16E − 06	2.82E − 02
	1-D density	1.94E − 05	1.72E − 05	3.91E − 02
	1-D Young’s modulus	3.59E − 05	2.78E − 05	5.94E − 02
	1-D Poisson ratio	6.74E − 06	4.71E − 06	1.05E − 02
	5-D variables	1.04E − 03	6.86E − 04	1.80
300	1-D radius	2.13E − 03	1.77E − 03	0.17
	1-D thickness	1.89E − 04	1.58E − 04	2.12E − 02
	1-D density	3.19E − 04	2.26E − 04	2.58E − 02
	1-D Young’s modulus	1.44E − 04	1.05E − 04	1.17E − 02
	1-D Poisson ratio	6.49E − 05	4.92E − 05	5.57E − 03
	5-D variables	5.59E − 02	2.20E − 02	2.44