Return to Main
Page
JOINT
CSHRP/NEW BRUNSWICK BAYESIAN APPLICATION
5.0 RESULTS 5.1 ITERATION 1 The following are the resultant model equations for the thin, thick and padding overlay model for the first iteration where:
5.1.1 Thin Overlay 5.1.2 Thick Overlay with Milling 5.1.3 Thick Overlay with Padding
5.2 ITERATION 2 The following are the resultant model equations for the thin and thick overlay models for the second iteration where:
5.2.1 Thin Overlay 5.2.2 Thick Overlay 6.0 DISCUSSION AND
RECOMMENDATIONS Bayesian methodology is a viable tool
that has applications for NBDOT . It has demonstrated
that it has the potential to assist in optimizing
rehabilitation design of overlays by predicting rutting
performance relative to Thin and Thick overlays. The
XLBayes software is user friendly once the user is
familiar with MSExcel spreadsheet software. However,
reference documents for assisting in the interpretation
of output graphs and data are limited and it is
recommended that additional documentation be developed in
this area. Problems encountered during this modeling process generated some recommendations for future modeling efforts. When the results for the predicted rut measurements versus the actual measured rut value were viewed in a barchart it was noted that the resultant differences were not consistent. On some contracts the predicted rut value was higher than the actual measured value and on others it was lower. To better address this observation, for any future development on this model it is recommended that an additional variable representing strength be added. It was also noted that the inference space for the variables in the encoded matrices was not the same as those recorded in the actual databases collected for the models. Therefore, it is recommended that the
data range over which the experts were originally encoded
be changed and the experts be reencoded for this change
as well as for any additional strength variable if future
development of this rutting model is contemplated. It is recommended that any future use
of Bayesian methodology with respect to developing a
predictive rutting model for NBDOT should be addressed as
a two stage model: Stage one would be from a design
perspective. Input to the model would be those variables
readily available to the Designers such as existing RCI,
strength values, traffic, thickness, age, and rut
measurements. The model could be used to determine
rutting for a range of critical variables and place the
results into a graphical format to assist the Designer in
his decision making. Stage two would be addressed
from a construction perspective. Input to the model would
be those variables readily available to the construction
engineer such as thickness of A/C new, thickness of A/C
total, % A/C, traffic, insitu density. With the
available information on insitu density, % air voids, %
AC in the mix and estimates of accumulated traffic, a
performance graph could be generated using any of these
variables versus traffic to determine the years to reach
a threshold rutting value in millimeters and therefore
rate the work. It is recommended that during the
initial steps of the methodology, that in addition to
providing input on potential variables to include in the
model, the experts be required to indicate how they feel
these variables would influence the model. After the
variables are selected then it is recommended that
statistical analysis initially be performed to determine
which variables are significant, which are highly
correlated etc. This step would help identify variables
that could be a problem in the analysis and allows them
to be addressed at an early stage. Another recommendation is to provide
experts a few days between encoding each model when
asking them to encode more than one. It is recommended that knowledge and
experience received from this project be shared with
interested students and professors from the universities
in New Brunswick. REFERENCES 1. Mark Nickeson, XLBAYES addin
Module for Microsoft Excel 5.0, Vemax Management Inc. 2. John B.L. Robinison, PhD.,P.Eng,
NBDOT ESAL Forecaster Program, D.C. Campbell Chair in
Highway Construction and Pavement Research, Department of
Civil Engineering, University of New Brunswick, in
conjunction with the Sensitivity Analysis of Load
Equivalency Factor Data Report. |
