| dc.description.abstract | The use of machine learning technology has shown great potential in supporting asset integrity monitoring and management. Asset integrity ensures that physical assets, such as industrial infrastructure, oil facilities, and pipelines, operate safely, reliably, and according to standards. In the effort to enhance asset reliability and safety, effective predictive methods are crucial for detecting potential issues and failures before they occur. This research aims to investigate the implementation of Machine Learning techniques, specifically the hybrid LSTM-RNN (Recurrent Neural Network-Long Short-Term Memory) model, to support asset integrity monitoring and maintenance. This hybrid model combines the strengths of both techniques to improve predictive performance and analysis in dynamic environments. Subsequently, the model is tested and evaluated using independent time series datasets to measure its predictive performance and analytical capabilities in detecting potential damages, anomalies, or significant changes in asset conditions. Based on 100 modeling iterations and predictions on the same dataset, the Confidence Intervals were obtained as follows: RNN achieved an average prediction result of 16.6, followed by LSTM with an average of 15.3, RNN-LSTM with an average of 15.2, and LSTM-RNN with an average of 15.0. The research results indicate that the hybrid LSTM-RNN model can provide more accurate and reliable predictions and analyses of asset integrity compared to traditional methods. | en_US |
