DEVELOPMENT OF A TRACK SAFETY PERFORMANCE MODEL FOR TRACK SAFETY PERFORMANCE ASSESSMENT IN INDONESIAN RAILWAY CASES (A CASE STUDY IN OPERATIONAL AREAS 1 JAKARTA, 2 BANDUNG, AND 3 CIREBON OF PT KERETA API INDONESIA (PERSERO))

• Main Author: Arisikam, Dicky
• Format: Dissertations
• Language: Indonesia
• Subjects: Teknik sipil
• Online Access: https://digilib.itb.ac.id/gdl/view/80779
• Institution: Institut Teknologi Bandung

The majority of Train Accidents (TA) that occurred in Indonesia from 2015– 2020 were caused by infrastructure factors, such as the condition of tracks, the presence of bridges, and the quality of signal, telecommunication, and electricity. To reduce these TA, need a maintenance or improvement of infrastructure, with a focus on locations with high risks to TA. Conventionally, the prioritization of maintenance is determined based on the number of TA that have occurred at a location within a time period. However, considering the random of TA data and the phenomenon of Regression to The Mean (RTM), so prioritizing maintenance locations based on the number of TA can be bias and the locations were selected is not locations with a high risk of TA. Therefore, a comprehensive approach is required to assess the safety performance of tracks. This approach can assist railway operators in selecting locations that need priority maintenance, so that infrastructure maintenance or improvement can be carried out more effectively and efficiently. This research is to develop a track safety performance model that can be utilized to assess the safety performance of tracks and select locations requiring priority maintenance or improvement. The model depicts the associative relationship between TA as the dependent variable with exposure (train frequency and length of tracks) and various infrastructure factors (tracks, bridges, and signal) as independent variables. In this context, infrastructure factors play a role as risk factors influencing TA. The model is constructed using the Generalized Linear Model (GLM) with specifications of Poisson Regression (RP), Negative Binomial (NB), Zero Inflated Poisson (ZIP), and Zero Inflated Negative Binomial (ZINB). TA data taken in segment or track between two stations in Operational Areas (OA) of 1 Jakarta, 2 Bandung, and 3 Cirebon from 2015–2020 are utilized as the basis for modeling. Totally there were 379 tracks from 3 OA. The selection of the model is based on tests of dispersion values, goodness-of-fit tests, and Vuong tests. The modeling results show that the NB regression model is the best model to describe the association relationship between TA and infrastructure factors. Variables associated with TA in Indonesia (according to study location) are train frequency (Train/day), length of track (Km), train speed (Km/hour), curve length of track with a radius of 500 m to ? 1000 m (Km), number of vulnerable location (Number), the length of electricity network (Km), and single or double track. The output of the is an estimated number of TA for each entity and then it used to assess track safety performance. Furthermore, the results of the assessment of track safety performance based on the Performance Indicator of Safety Performance Function (SPF) and Empirical Bayes (EB) showed that there were 47 and 102 dangerous segments. Then, based on the Data Envelopment Analysis (DEA) methods (DEA-TA and DEA-SPF) show the safety rating of track from 379 segments, does not mention whether the location is dangerous or not. The track ranked at the top from Performance Indicator and DEA method are the most dangerous track and need priority maintenance. Based on the results of the assessment using the Performance Indicator SPF and EB and the DEA-SPF method, the differences are not significant, this is because the variables used in the analysis are same, they are estimated TA and infrastructure variables from SPF, while the DEA-TA method shows different results, because it does not use parameters from SPF. Based on the advantages and limitations of these four assessments, the Performance Indicator of EB is the most appropriate assessment of track safety performance for the TA because it uses the TA data, the TA expectation number from SPF as analysis parameters and thresholds, and consider the effect of bias on RTM.