Přístupnostní navigace
E-application
Search Search Close
Master's Thesis
Author of thesis: Bc. Martin Vašíček
Acad. year: 2025/2026
Supervisor: Ing. Petr Hadraba, Ph.D.
Reviewer: prof. Ing. Zdeněk Hadaš, Ph.D.
This master's thesis addresses the design of a detection system for predictive maintenance of railway tracks based on the processing of voltage records from stationary piezoelectric sensors mounted directly on the rail. The core of the solution is a convolutional neural network in an autoencoder architecture trained exclusively on healthy train passages. Its task is to detect deviations from the typical signal pattern corresponding to the normal operating state of the track. The real data is futher acquired using a custom implementation that extracts individual wheelsets. The real data is supplemented with synthetic data obtained from a physical track simulator. The validation of the trained model is performed on three levels, namely the testing of four sizes of wheel flats, three types of rail defects, and the application of the model to a real recording. The results confirm the system's ability to detect shape-based deviations and at the same time to distinguish a systematic change in the track condition from a local defect of a passing vehicle based on the number of affected wheelsets in the recording. All Python scripts are attached.
convolutional neural network, autoencoder, anomaly detection, reconstruction error, piezoelectric sensor, predictive maintenance, railway track, rail defects, wheel flat, data extraction, synthetic data, finite element method
Date of defence
11.06.2026
Result of the defence
Defended (thesis was successfully defended)
Grading
B
Process of defence
Student obeznámil komisi s výsledky své DP. Po přečtení posudků následovaly dotazy oponenta (viz posudek) a komise: Dotazy ke grafické interpretaci (prezentace sl. 14) Volba jednoho průjezdu pro validaci. Popis autoenkodéru. Volba velikosti jader. Čas potřebný pro naučení neuronové sítě. Student reagoval uspokojivě na mnoho dotazů oponenta i komise.
Language of thesis
Czech
Faculty
Fakulta strojního inženýrství
Department
Institute of Automation and Computer Science
Study programme
Applied Computer Science and Control (N-AIŘ-P)
Composition of Committee
doc. Ing. Pavel Škrabánek, Ph.D. (místopředseda) prof. Ing. Zdeněk Hadaš, Ph.D. (člen) Ing. Jiří Kurfürst, Ph.D. (člen) Ing. Jiří Kovář, Ph.D. (člen) prof. Ing. Dagmar Janáčová, CSc. (člen) prof. Ing. Jiří Jaroš, Ph.D. (člen) prof. Ing. Miroslav Fikar, DrSc. (předseda) prof. Ing. Vladimír Vašek, CSc., dr. h. c., FEng. (člen)
Supervisor’s reportIng. Petr Hadraba, Ph.D.
Grade proposed by supervisor: B
Reviewer’s reportprof. Ing. Zdeněk Hadaš, Ph.D.
Grade proposed by reviewer: C
Responsibility: Mgr. et Mgr. Hana Odstrčilová