RESULTS AND DISCUSSION
Experiments were conducted in two regimes:
- the sample tube was empty
- the sample tube was packed
Empty sample tube
Typical results for Model A and Model B are presented in Fig 2 and Table
2 for empty sample tube. In the figures we present comparison of
experimental (solid line) and calculated (circles) signal curves for
both models for few experiments. In the tables we present values of
dispersion coefficient as well as f-values for all experiments.
In the Fig. 2 we present results of simulations obtained for model A
(left panel) and for model B (right panel) for the same operating
conditions. More experimental results are presented in Table 2. For
model A in the Fig. 2a (Experiment No. E4) we observe good, then in Fig.
2b (Experiment No. E7) average and in Fig. 2c (Experiment No. E12)
rather poor fit. Poor fits for model B are not observed. We can conclude
that in this case interactions between zone 2a and zone 2b are
significant. f-values ranged between 0.921 and 0.962 for the Model A and
between 0.973 and 0.993 for Model B and the observation justifies
presented conclusion. The best fits for Model A correspond to the worse
for Model B.
Let’s trace how DL-values varies with temperature and
pressure for both models. In addition to DL-values there
are also presented products of multiplying of DL by
pressure and by temperature raised to the -3/2 power. This product
should have a constant value, and, it actually is almost constant both
for Model A and B with one exception. Result of experiment E1 seems to
be an outlying result. Dixon’s Q-test confirmed this supposition for
both models and the result was discarded. Statistics of remaining
results are included in Table 2. The first and most important conclusion
drawn is that small values of standard deviation and the confidence
interval justified correctness of the method of identification of
dispersion coefficient. It is easily to observe that standard deviation
and confidence interval for Model B are about 40% smaller than for
Model A. It additionally indicates that DL coefficients
should be evaluated using model B. This conclusion is important –
average percentage difference between DL-values
determined from model A and from model B reach 15 percent. So, in the
case of empty tube, model B can be recommended as more precise than
model A.