Akademik Veri Yönetim Sistemi
SELCUK 13TH INTERNATIONAL CONFERENCE ON APPLIED SCIENCES, Konya, Türkiye, 12 - 14 Aralık 2025, ss.307-318, (Tam Metin Bildiri)
The early detection of faults in rolling element bearings, which are among the most critical components of rotating machinery, and the prediction of their vibration behavior are of vital importance for predictive maintenance strategies. Deep learning methods, particularly Long Short-Term Memory (LSTM) networks, have achieved strong performance in the analysis of vibration data with time-series characteristics. However, the predictive performance of LSTM models is highly sensitive to the proper selection of hyperparameters such as hidden layer size and learning rate, and manual tuning of these parameters often leads to suboptimal configurations. In this study, a hybrid ABC-LSTM model optimized by the Artificial Bee Colony (ABC) algorithm is proposed to accurately forecast bearing vibration signals. The effectiveness of the model is evaluated using a vibration dataset from the Case Western Reserve University (CWRU) Bearing Data Center, which includes different fault types such as ball, inner race, and outer race faults, as well as different fault sizes (0.007", 0.014"). The ABC algorithm is employed to optimize the topology and training parameters of the LSTM network in order to minimize the prediction error. The results, assessed in terms of RMSE, MAE, and sMAPE, show that the proposed hybrid approach exhibits superior performance in learning complex and noisy vibration signals. This study presents a self-optimizing and robust deep learning framework for fault prediction in industrial systems.