Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Data-driven maintenance in railway
Thesis (MEng) -- Stellenbosch University, 2022.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA |
| Published: |
Stellenbosch : Stellenbosch University
2022
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867614100504182784 |
|---|---|
| access_status_str | Open Access |
| author | Bezuidenhout, Mariska |
| author2 | Cornelius Jacobus, Fourie |
| author_browse | Bezuidenhout, Mariska Cornelius Jacobus, Fourie |
| author_facet | Cornelius Jacobus, Fourie Bezuidenhout, Mariska |
| author_sort | Bezuidenhout, Mariska |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng) -- Stellenbosch University, 2022. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/126100 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA |
| last_indexed | 2026-06-10T12:46:40.081Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2022 |
| publishDateRange | 2022 |
| publishDateSort | 2022 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/126100 Data-driven maintenance in railway Bezuidenhout, Mariska Cornelius Jacobus, Fourie Jooste, Johannes Lodewyk Lucke, Dominik Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Railroads -- Maintenance and repair Big data Railroads—South Africa—Safety measures Rolling Stock Machine learning Railroads—South Africa -- Monitoring UCTD Thesis (MEng) -- Stellenbosch University, 2022. ENGLISH ABSTRACT: Railways are an efficient way of providing public transport to millions of people all over the world. Creating rolling-stock assets is expensive and time-consuming, and therefore proper maintenance is essential to prolonging the use of assets. Maintenance is a complex and knowledge-intensive _eld; and the railway sector is no exception. Advancements in digitalisation and condition-monitoring technology have made it possible to track various functions and components on rolling stock through onboard sensors. Data is collected while trains are operating, with the purpose to use it for predictive maintenance. This involves acting proactively after early signs of actual or possible failure are detected. This process offers many advantages, such as increased reliability, availability and safety. The problem, however, is that even with the potential offered by data-driven solutions, vast amounts of data are collected without it being used effectively due to the scale thereof. The purpose of this study is to investigate how rail operators can analyse data gathered through condition-monitoring technology, and incorporate it into predictive maintenance of railway assets. A case study was applied to a South African railway operator, namely the Passenger Rail Agency of South Africa (PRASA). Data was gathered from TrainTracer, a condition-monitoring technology that had been fitted to PRASA's X'Trapolis commuter trains. The Team Data Science Process (TDSP) serves as the base for a data-science process and was applied to the pantograph subsystem. Pantograph bounce { the undesired phenomenon where contact between the pantograph and contact wire is momentarily broken { was studied. Pantograph-bounce data was acquired, wrangled and analysed to discover relationships and factors leading to failure. Classification, a sub-paradigm under supervised machine learning, was used to investigate the cause of pantograph bounce as being due to either faulty sensors, faulty pantographs or faulty infrastructure. A k-Nearest-Neighbours (KNN) classification model concluded that 58.4% of the pantograph faults were caused by faulty pantographs, 38.4% by infrastructure, and 3.2% by sensors. The infrastructure areas and routes in which the most pantograph bounces were observed were also identified. The TDSP and the classification model results demonstrated the potential of data gathered on railway assets. The pantograph was used as a candidate subsystem to study for this project; however, these processes can also be applied to other components or subsystems. In essence, rail operators can leverage the potential of condition-monitoring technology for predictive maintenance and ultimately serve passengers at a lower cost, and in an efficient and safe manner. AFRIKAANS OPSOMMING: Spoorweö is 'n d'NItroffende mnnier orn wöreldwyd 0'RnbAre vervoer aan miljoene mense te versW. Die skep van spoorvoertuigbates is duur en tydrowend; derhAIwe is behoorlike instandhouding nood- snAkIik om die gebruik bAtes te verleng. Instandhouding is 'n ingewikkelde en kennisintensicwe veld, en binne die spoorwegsektor is dit gcon uitsondering nie_ Vooruitgng in digitalisering en kondisiemoniteringstegnologie het dit moontlik gemaak om verskeie funksies en komponente van spoorvoortuie deur middel VAn ingeboude sensors te kontroleer. DAtA word versnmel terwyl treine bcdryf word, met die doel om dit vir voorspellingsin- stAndhouding te gebruik. Dit behels pronktiewe optrcde nadAt vrooö tokens VAn werklike of moontlike fAIing bespeur word. Hierdie proscs bi«i bAie voordele, onder Ander 'n toenAme in betroubAarheid, beskikbAArheid en veiligheid. Die problecrn is egter dat, selfs met die potensiAAI WAt deur opItNings gebied word, word 'n enorme dAta versnmel sonder dut dit — eenvoudig wccn.s die ornvgng dAArvAn — d'NItreffend gebruik word. Die dool van hierdie studie is om te ondersoek hoe spoorwegondernemings datA kAn ontlecd WAt deur middel kondisiemoniteringstegnologie versAmeI word, en dit by voorspellingsinstandhouding van spoorwegbAt€s kAn inkorporcwr. 'n GovAIIcstudie is op 'n Suid-AfrikAAnse spoorwegonderneming, nAAmIik die Passenger Rail Agency of South Africa (PRXSA), t.ocgepsw. DAtA is versnmel van TTAinmcer, 'n kondisiemoniteringstegnologie WAArmee PRASA se X'TnpoIis-pendeItreine toegerus is. Die TBSP (TOam DAta Science Process)- prosesmodel, WAt AS die grondsIAg vir 'n datAwetenskapproses dien, is op die toegepas. Puntogrnnfwip — die ongewenste verskynsel WAArtydens kontAk tussen die pantograf en kontAkdraad momenteel onderbreek word — is bßtudcwr. Rou dAtA oor puntogrAAfwip is bekom, in makliker bruikbAre formate omskep, en ontl«xi om verhoudings en fAktore te ontdek WAt tot fAIing lei. KIAssifikA-sie, 'n mnsjienleormetode, is gebruik om die oorsAAk vnn puntogrAAfwip as wcons öf foutiewe sensors, öf foutiewe pantogrnwe öf fouticwe te ondersoek. Die het tot die slotsom gekom dat 38.4% VAn die pnntograafsteurings dour foutiewe pnntograwo, 38.4% deur die en deur sensors veroorsnnk word. Die infrswtuktuur- areas en -roetes waar die pantografwippe voorgekom het, is 00k geidentifis«r. Die TDSP en die resultAte die klnssifi"siemodel het die potensiAAI VAn dAta "'At op spoorwcg- bAt€s versamel word, Vir hierdie projek is die pantograaf as knndidnntsubstelsel vir bestudering gebruik; hierdie prcßc-sso kAn egter 00k op Ander komponento of sutFteIseIs toegepas word. Spoorwegondernemings kAn die Intensiaal van kondisiemoniteringstegnolwie uenlik vir Voor- spellingsinstandhouding aanwend, en passasiers uiteindelik teen 'n laer koste, en op 'n doeltre ende en veilige manier, bedien. Masters 2022-11-14T10:47:12Z 2023-01-16T12:50:05Z 2022-11-14T10:47:12Z 2023-01-16T12:50:05Z 2022-11 Thesis http://hdl.handle.net/10019.1/126100 en_ZA Stellenbosch University xvii, 120 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Railroads -- Maintenance and repair Big data Railroads—South Africa—Safety measures Rolling Stock Machine learning Railroads—South Africa -- Monitoring UCTD Bezuidenhout, Mariska Data-driven maintenance in railway |
| title | Data-driven maintenance in railway |
| title_full | Data-driven maintenance in railway |
| title_fullStr | Data-driven maintenance in railway |
| title_full_unstemmed | Data-driven maintenance in railway |
| title_short | Data-driven maintenance in railway |
| title_sort | data driven maintenance in railway |
| topic | Railroads -- Maintenance and repair Big data Railroads—South Africa—Safety measures Rolling Stock Machine learning Railroads—South Africa -- Monitoring UCTD |
| url | http://hdl.handle.net/10019.1/126100 |
| work_keys_str_mv | AT bezuidenhoutmariska datadrivenmaintenanceinrailway |