JavaScript is disabled for your browser. Some features of this site may not work without it.
| dc.contributor.author | Tsedal, Mequanent Admasu | |
| dc.date.accessioned | 2024-12-19T07:05:18Z | |
| dc.date.available | 2024-12-19T07:05:18Z | |
| dc.date.issued | 2024-02 | |
| dc.identifier.uri | http://ir.bdu.edu.et/handle/123456789/16405 | |
| dc.description.abstract | Fused Deposition Modelling is an Additive Manufacturing technology capable of producing simple and complex 3D structures using plastic materials. However, like any manufacturing process, Fused Deposition Manufacturing is susceptible to defects that can lead to mechanical failures and safety hazards. Common defect types include warping, layer shifting, and under extrusion. Deep Learning algorithms show promising solutions to address these problems. Although promising the research area still needs considering improvement and real-time implementation. This thesis proposes a multi-classification model utilizing the ResNet50 pre-trained Convolutional Neural Network algorithm involving adjustments to the network structure and hyperparameters to detect these defects. To enhance the model's performance and visualize the effect of batch normalization and drop out techniques are employed besides a random search tuner. The algorithm is implemented in Python and trained to classify the displayed defect type using the layer-by-layer collected dataset under consistent lighting to prevent the illumination conditions. The printing material used in this study is PLA. The model with dropout has a classification test accuracy of 98.63% which is the best model out of the four revealed models. The model prediction time shows feasibility to be implemented in real-time scenarios. Keywords: Fused Deposition Modelling, Additive Manufacturing, ResNet50, Convolutional Neural Network, Random search tuner | en_US |
| dc.language.iso | en_US | en_US |
| dc.subject | Mechanical and Industrial Engineering | en_US |
| dc.title | Fused Deposition Modelling Defect Detection using Customized ResNet50-based Convolutional Neural Network | en_US |
| dc.type | Thesis | en_US |
