This report is a study of departure punctuality at the SJ train depot in Hagalund. The purpose of the report has been to find underlying patterns and reasons behind late departures.
The theoretical framework that has been used is a combination of Lean and Six Sigma, where the management tool Lean has its basis in optimal resource utilization and minimizing of waste. The Lean concepts that have primarily been used are Visualization and Standardization, this since those are areas in which SJ have been lacking.
The practical work behind the report has been done using the Six Sigma method DMAIC (Define, Measure, Analyze, Improve, Control), where a large focus has been put on Measuring and Analyzing.
The quantitative data that’s been used has come directly from SJ’s own late departure reports, where trains departing 5 minutes or more past the scheduled time are considered to be late. This lateness is automatically registered, where the scheduled departure time is called Right Time (RT) and RT > 5 hence indicated a late departure.
The reason behind the lateness is also noted for all departures, but this data is entered manually and the reason is chosen from a limited, predefined list of lateness codes (JDE codes).
A data mining of the late departure statistics for the Timetable period 2012 (December 11th 2011 to December 8th 2012) revealed large flaws with the manual lateness reporting, where inconsistent usage of the JDE codes made it impossible to discern any underlying patterns in lateness factors.
To circumvent the data flaws an experiment was mad during November 2012, where all late departure reporting during the month was monitored to ensure proper JDE code usage. The result revealed a large previously unknown source of delay, “Human error”, which had hitherto been hidden in the catch-all code “Miscellaneous”.
A visualization of the automatically collected departure data, the RT data, in turn revealed clear issues during personal shift changes, and also concluded 1pm-6pm CET to be a late departure heavy time of the day. The visualization of departure data was also compared to the visualization of trains’ time spent at the depot, the so called turn time, where trains spending less than three hours at Depot Hagalund could be shown to affect the general departure punctuality to a higher degree than other trains. Through use of regression analysis it could also be shown that trains arriving late with a short turn time, to a higher degree also departed late, whilst trains with a longer turn time were seemingly statistically unaffected by delays in arrival.