Page 57 - Undergraduate Research Opportunities Program Proceedings 2017-18

School of Engineering
Department of Civil & Environmental Engineering

Supervisor: CHANG Chih-chen / CIVL
Student: SAN JUAN Justin David Quitalig / CIGBM Course: UROP2100, Fall
UROP3100, Spring

Hand-drawn-schematic-structural diagrams are always adopted to analyze the response of civil structures for further
design purposes. Analysis of such structures is often complex and better completed by a computer, while learning
specialized software is often out of reach for students due to time and cost. To address this issue, this study develops a
method to analyze the structures directly from the hand-drawn diagrams, namely Image-based Structural Analysis (ISA).
Selective search and a deep Convolutional Neural Network (CNN) are designed to detect objects in the images. Based on
the bounding boxes and class information of the objects, a finite element model is constructed for determining the
response of the structure. To this end, results show that the proposed system is capable of reconstructing the structural
system with reasonable accuracy and it is suitable for mobile deployability. The code is now available at:
https://github.com/JSANJ/isa_python.

Personalized Learning for Civil Engineering Using Recurrent Neural Network

Supervisor: CHANG Chih-chen / CIVL
Student: TAN Ke Yi / COMP Course: UROP2100, Fall
UROP3100, Spring

Knowledge tracing, the technique used to model learners’ masteries of knowledge being tutored and understand their
learning process, is fundamental in the domain of intelligent tutoring system. Bayesian Knowledge Tracing and Deep
Knowledge Tracing are two popular implementations of knowledge tracing. In this paper, we investigate the feasibility of
modelling students’ learning behavior with recurrent neural network within a traditional classroom. Students’
interactions with coursework of structural analysis are transformed into concept vectors. The learning dynamics of
students are then constructed with recurrent neural network. This learning dynamics system is intended to help
instructors with evaluating students’ learning states. With this framework, instructors can trace students’ performance
and arrange the teaching sequence for more effective learning of students.

UROP Research on the Assessment of Population Exposure to Air Pollution

Supervisor: LAU Alexis Kai Hon / CIVL
Student: FAN Huiying / EVMT Course: UROP1100, Fall
YAU Hoi Ying /EVMT UROP1100, Fall

With the increasing demand for understanding the population exposure of air pollution in Hong Kong, a huge amount of
air quality data were collected from various monitor stations. To ensure the collected data reflects the air quality the
public were exposed to with high reliability, there is a need for data cleaning before the analysing stage. By conducting
data tests including physical range and extreme value check, Signal’s gradient test, calibration drift adjustment and data
frequency conversion with reference to the expected range of data, our group has developed a data cleaning tool in
Python to correct faulty data through smoothening the gradients and adjusting for calibration drifts.

Field Measurements of Pressure Surges in Water Supply Pipeline Systems for Leakage Detection
Supervisor: LEE Joseph Hun-wei / CIVL Co-Supervisor: CHAN Shu Ning / CIVL
Student: WONG Leong Sang / CIVL Course: UROP1100, Fall
UROP2100, Spring

47