Institutional Repository of Vilnius University
Juostinio konvejerio būsenų klasifikavimas naudojant giliuosius neuroninius tinklus ir duomenų augmentacijos metodus /
Abstract
dc:descriptionThis work explores the application of deep neural networks for conveyor belt load and defect state classification, using various data augmentation methods aimed at improving the classification accuracy of deep learning models. Tension data signals from five different conveyor belt load weight conditions (0.5 kg, 1 kg, 2 kg, 3 kg, 5 kg) and one damaged belt condition were collected and analyzed. Four deep neural network models based on the FCN, CNN-LSTM, ResNet, and InceptionTime architectures were developed and applied to classify conveyor belt states. Different time series augmentations such as Laplace noise, Gaussian noise, uniform noise, and magnitude warping were applied to the data collected during the study. Furthermore, new conveyor belt tension signals were generated using a TimeVAE model. The study has shown that deep neural network models based on FCN, ResNet, and InceptionTime architectures are able to classify conveyor belt states accurately and efficiently. The research has also shown that various data augmentation methods can improve the accuracy of these models, for example, the combined addition of Laplace and Gaussian noise improved the FCN models baseline (without augmentation) classification accuracy with 2.0 s length signals by 4.5% to 92.6% ± 1.54%. The FCN model has demonstrated the best accuracy and performance despite having the lowest amount of trainable parameters, thus demonstrating the importance of selecting and optimizing the right architecture when developing models for specific tasks.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Pikšrys, Armantas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210647802&prefLang=en_US