The 440-group data contain most of common packings in
domestic market, including the old and the new type of
packings with all kinds of diameters and materials, and almost
cover all the random packings used in industrial processes.
The 333-group data from the earlier years of the 440-group
are used to train the LS-SVM model, and the other 107-group
data from the later years are selected as the test set to predict
the flooding velocity. The old packings are used to train the
LS-SVM model and then predict the flooding velocity of the
new packings. However, some packing constants are difficult
to obtain of many new packings for the traditional models. In
order to obtain the comparison results of LS-SVM model and
traditional empirical models, the test data are mixed with the
early years and the later years, while the 333-group train data
are obtained from the earlier years.