As can be shown in TABLE I, different numbers of
training data can cause the differences in prediction
performance. However, the increase of the group number has
not improved the prediction accuracy obviously. Even when
the group number is increased to 370, its prediction
performance is worse than the 333-group one. It is mainly
because of the characteristics and relevance of flooding data.
Consequently, the 333-group data are utilized to train the
LS-SVM model for prediction of the flooding velocity in the
following sections.
B. Comparison of LS-SVM Model with Traditional Models
Three traditional empirical models are adopted to predict
the flooding velocity of DN50 plastic short-cascade ring,
DN50 metal Pall ring and DN38 metal Intalox saddle. Then,
they are used to make the comparison with the LS-SVM
model. The first correlation is the Bain-Hougein equation
[14,15], which is an improvement and regression of the
Sherwood chart, formulated in (16):
As can be shown in TABLE I, different numbers of
training data can cause the differences in prediction
performance. However, the increase of the group number has
not improved the prediction accuracy obviously. Even when
the group number is increased to 370, its prediction
performance is worse than the 333-group one. It is mainly
because of the characteristics and relevance of flooding data.
Consequently, the 333-group data are utilized to train the
LS-SVM model for prediction of the flooding velocity in the
following sections.
B. Comparison of LS-SVM Model with Traditional Models
Three traditional empirical models are adopted to predict
the flooding velocity of DN50 plastic short-cascade ring,
DN50 metal Pall ring and DN38 metal Intalox saddle. Then,
they are used to make the comparison with the LS-SVM
model. The first correlation is the Bain-Hougein equation
[14,15], which is an improvement and regression of the
Sherwood chart, formulated in (16):
การแปล กรุณารอสักครู่..