Fault detection in diesel engines using deep learning and Qarma algorithms
Rogov, Oleg (2022)
Diplomityö
Rogov, Oleg
2022
School of Energy Systems, Konetekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2022052538698
https://urn.fi/URN:NBN:fi-fe2022052538698
Tiivistelmä
QARMA, known as quantitative associated rule mining algorithm, is one of the recently developed algorithms for classification, that needs more testing on the real data. Neural Networks, that nowadays only tend to improve, are very common in solving different classification problems.
This work compares the state-of-the-art algorithm of QARMA with the established approach of Neural Networks by training them on the open-source dataset of the diesel engine faults, which contains 4 types of different engine conditions. Deep Neural Network was created by me using Python language with its different data processing libraries, while QARM algorithm was ran in a special graphical user interface under the control of the copyright holder. The evaluation of this algorithms is performed with the usage of accuracy metrics and confusion matrix. One of the goals of the work was also to improve the accuracy for diesel faults detection in comparing to the existing results, that were described in one of the articles – 5% improvement was reached with the usage of QARMA and Deep Neural Network. Few recommendations on how to improve the research algorithms are proposed.
This work compares the state-of-the-art algorithm of QARMA with the established approach of Neural Networks by training them on the open-source dataset of the diesel engine faults, which contains 4 types of different engine conditions. Deep Neural Network was created by me using Python language with its different data processing libraries, while QARM algorithm was ran in a special graphical user interface under the control of the copyright holder. The evaluation of this algorithms is performed with the usage of accuracy metrics and confusion matrix. One of the goals of the work was also to improve the accuracy for diesel faults detection in comparing to the existing results, that were described in one of the articles – 5% improvement was reached with the usage of QARMA and Deep Neural Network. Few recommendations on how to improve the research algorithms are proposed.