Rizqi Fitri Naryanto, Mera Kartika Delimayanti, Agustien Naryaningsih, Rizky Adi, Bima Aji Setiawan
This research aims to create a deep learning model utilizing Artificial Neural Networks (ANN) and Convolutional Neural Networks (CNN) to detect and classify damage or faults in diesel engines effectively. The main contribution of this study entails the creation of a classification model aimed at problem identification in diesel engines. The study used the DEFault dataset, which has 3,500 rows of data classified into four distinct labels. The DEFault dataset comprises four discrete noise levels, namely 0dB, 15dB, 30dB, and 60 dB. The findings indicate that both models have demonstrated satisfactory outcomes regarding model assessment. The model performance is most ideal when trained on the DEFault dataset with 60dB white noise, whereas the dataset with 0dB white noise leads to the least favourable model performance. The results suggest that artificial neural networks (ANN) perform better than convolutional neural networks (CNN) in datasets with higher levels of white noise. Conversely, CNN demonstrates superior performance in datasets with lower levels of white noise. An additional investigation could prove advantageous in implementing the deep learning model on a device that can identify diesel engine faults in real-time. © Little Lion Scientific.
Mechanical Engineering Department, Engineering Faculty, Universitas Negeri Semarang, Indonesia; Department of Computer and Informatics Engineering, Politeknik Negeri Jakarta, Indonesia; Fisheries Research Center, National Research, and Innovation Agency (BRIN), Indonesia
