STADYN Wave Equation Program 4: Eta Limiting, and More on Norm Matching

In our last post we broached the subject of different norm matching methods for the actual and computed velocity-time histories at the pile top. In this post we will go into $\eta$ limiting, while at the same time running both norms to get a better feel for the differences in the results.

Before we begin, one clarification is in order: CAPWAP’s Match Quality and the use of the 1-norm in STADYN are similar in mathematical concept but different in execution. That’s because the Match Quality weights different part of the force-time history (in their case) differently, whereas STADYN goes for a simple minimum sum difference.

One characteristic of the inverse case both in the original study and in the modifications shown in the last post are very large absolute values of $\eta$ . These are products of the search routine, but they are not very realistic in terms of characterising the soil around the pile. To illustrate, we bring back up one of the results from the last post, showing the optimisation track using the 2-norm and phi-based Poisson’s Ratio (which will now be the program standard):

Note that the #8 track ( $\eta$ for the lowest shaft layer) has a value approaching -30; this is obviously very unrealistic.

In principle, as with $\xi$ , the absolute value of $\eta$ should not exceed unity; however, unlike $\xi$ there is no formal reason why this should be the case. But how much should we vary $\eta$ ? To answer this question, and to continue our investigation of the norm issue, we will examine a matrix of cases as follows:

$\eta$ will be run for values of 1, 2, 3 and unlimited (the last has already been done.)
Each of these will be run for both the 1-norm and 2-norm matching.

A summary of the results are shown below

Changed Parameter	Difference		Static Load, kN		Average Shaft $\xi$		Toe $\xi$		Toe $\eta$
Norm	1	2	1	2	1	2	1	2	1	2
\| $\eta$ \| < 1	0.3364	0.003690	811	1490	-0.364	-0.149	-0.62	-0.311	-0.175	0.611
\| $\eta$ \| < 2	0.2381	0.002626	278	223	-0.091	-0.06	-0.588	-0.316	-0.781	-0.0385
\| $\eta$ \| < 3	0.1806	0.001707	172	207	0.324	0.42	-.832	0.823	-1.01	1.45
Unrestricted $\eta$	0.1344	0.001456	300	218	-0.329	-0.183	-0.491	0.804	8.19	1.52
$\nu = f(\xi,\eta)$	0.1484	0.001495	278	187	-0.383	-0.53	0.792	0.366	3.116	1.814

To see how this actually looks, consider the runs where | $\eta$ | < 3. We will use the 2-norm results.

Velocity-Time Output — Impedance*Velocity Comparison, 2-norm, eta limiting = 3.

Optimization Track — Optimisation Track, 2-norm, eta limiting = 3

The results indicate the following:

The average shaft values of $\xi$ tend to be negative. This is contrary to the cohesive nature of the soils. The interface issue needs to be revisited.
The toe values do not exhibit a consistent pattern. This is probably due to the fact that they are compensating for changes in values along the shaft.
As values of | $\eta$ | are allowed to increase, with the 2-norm the result of the simulated static load test become fairly consistent. This is not the case with the 1-norm. Although limiting | $\eta$ | to unity is too restrictive, it is possible to achieve consistent results without removing all limits on $\eta$ .
The velocity (actually impedance*velocity) history matching is similar to what we have seen before with the unlimited eta case.
The optimisation track starts by exploring the limits of $\eta$ , but then “pulls back” to values away from the limits. This indicates that, while limiting values “within the box,” i.e., the absolute values of $\eta$ < 1, is too restrictive, reasonable results can be obtained with some $\eta$ limiting.

Based on these results, $\eta$ limiting will be incorporated into the program. The next topic to be considered are changes in the soil properties along the surface of the pile, as was discussed in the last post.