|
| | | |
Improving Bridge Deterioration Modelling Using Rainfall Data from the Bureau of Meteorology
Huang, Q., Ong, K.L. and Alahakoon, D.
appropriate maintenance is paramount. Often, au-
thorities are faced with limited funding and available
contractors who are able to carry out the maintenance
checks and works. Therefore, a predictive model that
can forecast the future state of a bridge component
will enable the authority to prioritise and deploy re-
sources to where it is most needed. The challenge
faced in this paper is the requirement from the Victo-
rian road authorities to develop an eective predictive
model. Prior attempts have been made by using dif-
ferent techniques to construct an alternate predictive
model but with limited results. The problem lie in the
data itself. With data manually recorded by dier-
ent contractors, it is noisy and erroneous. Attempts
to data cleaning has led to little improvement in the
overall model performance. Finally we turned to data
augmentation to increase the proportion of reliable
data. In our quest to do so, we ended up pulling rain-
fall data from the BoM to augment the data provided
by VicRoads. We consider rainfall data as a candidate
for augmentation because literature in civil engineer-
ing has correlated bridge component deterioration to
the presence of water moisture. Since high rainfall
contributes to increased deterioration, leveraging the
rainfall information should lead to improved predic-
tive performance. Initial experiments on the predic-
tive performance of the baseline and \high rainfall"
models suggest the viability of this approach. |
Cite as: Huang, Q., Ong, K.L. and Alahakoon, D. (2015). Improving Bridge Deterioration Modelling Using Rainfall Data from the Bureau of Meteorology. In Proc. Thirteenth Australasian Data Mining Conference (AusDM 2015) Sydney, Australia. CRPIT, 168. Ong, K.L., Zhao, Y., Stone, M.G. and Islam, M.Z. Eds., ACS. 161-167 |
(from crpit.com)
(local if available)
|
|