| 1. |
For the Barton-Bandis
criterion, changed from basic friction angle to
residual friction angle. |
| 2. |
In the friction angle
dialog for the Barton-Bandis criterion, changed
the output of the dialog from phib to phir. Added
menu items for calculating a residual friction
angle based on phib, and wet and dry Schmidt hammer
rebound values (according to the Barton & Choubey,
1976 paper). Users have the option to either use
the calculator or enter their own reduced friction
angles. |
| 3. |
Improved the Levenberg-Marquardt
curve fitting technique for Barton-Bandis data
so that it converges on a wide range of input data.
Did this by first performing a Simplex fit, and
then using its results as the seed value for the
Levenberg-Marquardt routine. |
| 4. |
Improved the performance
of the Levenberg-Marquardt and Simplex routines
for fitting the Hoek-Brown criterion to lab data.
First, a Mohr-Coulomb fit is done to obtain a good
starting value of uniaxial compressive strength,
which is then used to determine initial search
values in the Levenberg-Marquardt and Simplex curve
fitting routines. |
| 5. |
Fixed a bug in the
program that lopped off the last data point whenever
data was copied from one RocData input spreadsheet
to another. |
| 6. |
Increased the display
size of data points. |
| 7. |
For the analyses of
Hoek-Brown lab data, Generalized Hoek-Brown field
data, and both triaxial and direct shear data with
the Power Curve, included inverted forms of the
respective curves. This allows tensile data points,
or any data to the left of the valid ranges for
these curves to be properly considered in the curve
fitting process. Prior to this feature, such points
were simply ignored. Changed the calculation of
residual values to include errors associated with
these points.
The changes were done for the Levenberg-Marquardt
and Simplex curve fitting routines. |
| 8. |
Fixed two bugs in the
Simplex routines. The first was caused by attempts
to read beyond the allocated range of an array.
The second bug slowed down convergence: the second
best point, instead of the best, found in an iteration
was used in the next step. |
| 9. |
Fixed a bug in the
main view of RocData that displayed only a short
segment of the instantaneous Mohr-Coulomb curve,
whenever triaxial lab data was analyzed with the
Power Curve strength model. |
| 10. |
Fixed a
bug that caused the equivalent Mohr-Coulomb strength
curve to be displayed with a dashed line instead
of a solid line, whenever the display Mogi line
option was turned on. |
| 1. |
Fixed bug in Power
Curve lab data analysis dialog that caused program
to crash when entering or modifying grid data. |
| 2. |
Checked the parameter
limits for all strength models types and fixed
a few that were wrong. |
| 3. |
Fixed a bug in the
Amoeba algorithm for the Simplex method (for all
the strength models). I changed the line “Swap(y[1],
y[ilo])” to “Swap(y[0], y[ilo])”. |
| 4. |
Fixed a memory leak
(with the aparam variable) in the CurveFit_xxx_Simplex
functions for the different strength methods. |
| 5. |
Replaced the automatic
differentiation approach in the tausignhb_F function
with derivatives explicitly obtained using Mathematica. |
| 6. |
Modified
the Levenberg-Marquardt method for lab data so
that tensile strength values are included during
the curve fitting process. Added code that allowed
errors to be calculated for points to the left
of a Hoek-Brown curve (i.e. sig3 that resulted
in m*sig3/sigc + 1 < 0). (Modified the Levenberg-Marquardt
algorithm in RocLab as well.) |
