Abstract:
This final year project explored the implementation of a predictive maintenance strategy within Sonatrach LNG1/Z complex, aiming to enhance maintenance management within the complex. Facing the limitations of traditional maintenance approaches, which often lead to unexpected downtime and high costs, the primary objective was to develop a system capable of anticipating equipment failures before they occur, using artificial intelligence.
The study began with a detailed presentation of the operational context and specific challenges at the GNL1/Z complex. And then explore the fundamentals of artificial intelligence, including Machine Learning and Deep Learning architectures, and MLPs, CNNs, GRUs, LSTMs algorithms. The core of the approach involved analyzing historical equipment data to identify the most suitable algorithm for this work.
The implementation shows that Convolutional Neural Networks (CNNs) is the most effective algorithm for this application. The trained CNN model achieved a remarkable accuracy of 91.58% in predicting failures. This high performance allows the LNG1/Z complex to reliably anticipate potential breakdowns, for faster and more effective maintenance interventions. This will minimize unplanned downtime, reduce operational costs, optimize equipment lifespan, and increase facility safety. This approach represents a major strategic optimization for the continuous improvement of performance and competitiveness at Sonatrach LNG1/Z complex.
