An approach towards the implementation of a reliable resilience model based on machine learning

Machine Learning tools to enhance systems' resilience received an increased impetus driven by energy transition, climate change and digitalization, but critical challenges on system requirement definition and reliability of learning processes need to be addressed. This study proposes a systematic framework based on system engineering and focused on the reliability of the learning process of the Hidden Markov Model (HMM) coupled with the Baum-Welsh algorithm. The HMM hidden states may represent the precursors of accidental events, being the states between a regular performance and a failure of a sub-system. The Baum-Welch algorithm, estimating the parameters of the HMM, iteratively updates the estimates of the state transition and observation probabilities. The framework was applied to a real case of LNG bunkering, showing that the system can learn from incomplete data, improve the learning quality given a new set of observations, make predictions about the latent states and enhance system resilience. The novelty of this work lies in ensuring the learning process and contributing to the attainment of an explainable, robust, and interpretable data-driven approach.

Automated summary

Machine learning tools have gained increased importance in enhancing system resilience due to energy transition, climate change, and digitalization. However, challenges remain in defining system requirements and ensuring the reliability of the learning process. This study proposes a systematic framework based on system engineering to improve the reliability of learning process using the Hidden Markov Model (HMM) coupled with the Baum-Welch algorithm. The framework was applied to a real case of LNG bunkering, demonstrating its ability to learn from incomplete data, improve learning quality, make predictions, and enhance system resilience. The novelty of this work lies in ensuring the learning process and contributing to the development of an explainable, robust, and interpretable data-driven approach.