Tillståndsbaserat underhåll av spårväxlar genom statistisk processtyrning : En fallstudie enligt DMAIC

Sammanfattning: Switches, which are critical components of the Swedish railway, have a neglected maintenance cost that is three times as high as their current annual maintenance cost. Between 2017 and 2018, switches’ reported faults increased by 38 % and about one-third of them caused delays on the regular railway traffic. The purpose of this master thesis is to present recommendations of how condition-based maintenance could reduce the occurrence of faults in railway switches that affects the regular railway traffic. Condition-based maintenance is a cost-effective strategy designed to monitor and plan maintenance according to the condition of a device and is suitable for remote-controlled monitoring. To fulfill the purpose the thesis was divided into three milestones that were accomplished by applying the problem-solving method DMAIC (Define, Measure, Analyse, Improve and Control). The first milestone included the Define and Measure steps and aimed to investigate which fault caused the largest number of delay minutes per fault. Different categories of causes that affects the railway traffic were analysed. Among them, Material-weakening/Aging and Broken component resulted in many faults and delay minutes. The faults were sorted into groups at component level. Faults caused by Gearbox were identified as those causing the largest delays in the railway traffic. The result from the first milestone with the Define and Measure steps were then used for the second milestone. The second milestone included the Analyse step where it was investigated if the identified faults in Gearbox could be foreseen. This investigation was first conducted through the analysis of alarms recorded in one of Trafikverket’s databases. A graphical analysis of the data showed that no relationship could be identified between the faults in Gearbox and the recorded alarms of the database. Then, it was investigated if faults in Gearbox could be foreseen using statistical process control charts based on switching time. Statistical process control monitors a process using real time data. However, in this thesis we used historical data from 2018 to perform our analysis. The available data had deficiencies in quality due to truncation of the switching time. The truncation meant that the decimals were removed. The control charts issued out-of-control situations where the existing database did not record any alarm. Moreover, data on the switching time of several switches showed significant autocorrelation that affects the calculation of the control limits. However, the results appeared complex to interpret most likely because of the truncation and the autocorrelation of the data. A further graphical analysis of the switching time and the mean of the switching time indicated that 69 % of the switches had a probable relationship between faults in Gearbox and switching time. The third milestone included the Improve and Control steps and provided recommendations of how to reduce the occurrence of faults in switches. The analysis conducted in the previous milestones led to the following recommendations: Increase measurement accuracy when measuring switching time, Establish control charts for the switching time based on statistical process control and explain potential causes of the observed autocorrelation, and Improve reporting procedures of faults in the database.