UTHM Institutional Repository

Investigative model of rail accident and incident causes using statistical modelling approach

Ismail, Shamsullarifin (2016) Investigative model of rail accident and incident causes using statistical modelling approach. Masters thesis, Universiti Tun Hussein Onn Malaysia.


Download (579kB)


Nowadays, railway transportation becomes a popular choice among commuter as transportation mode to travel from one place to another. Thus, it makes the industry grows faster especially at urban area. The complexity of rail network required high level of safety features to prevent any interruption. For that purpose, this thesis will show a proper procedure on how the prediction model of accident need to be conducted using regression model. From the root cause analysis, the most contributory factor to influence the accident can be identified. “Ishikawa diagram” is a popular tool to identify problem occurring from the root where it begins. Process of identifying required bundles of accident and incident investigation report at least for 5 years and this thesis used data starting from 1999 to 2014. It was taken from several sources on Australian Railways website. Analysis from Ishikawa shows there are ten main factors involved to influences an accident. Those factors are “train driver mistake”, “other’s human mistake”, “weather influence”, “track problem”, “train problem”, “signaling error”, “maintenance error”, “communication error”, “procedure error”, and “others”. Each factor with positive correlation coefficient value to the type of accident and incident were taken as parameter. Then, before completing the prediction model formula, some of hypothesis needs to be tested to know which model among regression model is suitable and give a better prediction result. Dispersion test is a test to calculate dispersion value to know either data is under dispersion for value less than 1 (Poisson model is appropriate) or over dispersion for value more than 1 (Negative binomial is appropriate). Then, Vuong test is applied to measure which model has a better result between those two models. From the hypothesis, this thesis shows that Zero-inflated model is the most fitted model to predict accident and incident cases of collision, derailment and SPAD. In some country, they may have different system of rail and geography, thus it should have different possibilities to influence accident and incident. However, this method and procedure are available to use for them to identify and predict the most influencing factor that contributes to the accident occurrences.

Item Type: Thesis (Masters)
Subjects: T Technology > TF Railroad engineering and operation
Depositing User: Mr. Mohammad Shaifulrip Ithnin
Date Deposited: 09 May 2017 07:33
Last Modified: 09 May 2017 07:33
URI: http://eprints.uthm.edu.my/id/eprint/9113
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item


Downloads per month over past year