Methodology for diagnosing the technical condition of aviation gas turbine engines using recurrent neural networks (RNN) and long short-term memory networks (LSTM)

Civil Aviation High TECHNOLOGIES, Год журнала: 2025, Номер 27(6), С. 21 - 41

Опубликована: Янв. 11, 2025

This study presents a method for diagnosing the technical condition of aviation gas turbine engines (GTE) using recurrent neural networks (RNN) and long short-term memory (LSTM). The primary focus is on comparing effectiveness these models forecasting key operating parameters GTEs, such as vibrations, turbine-inlet temperatures, rotor speeds low high pressure. research involved thorough data cleaning normalization, including handling missing values, normalization Min-Max Scaling, outlier removal, decorrelation, time series smoothing. RNN LSTM were trained backpropagation through (BPTT) algorithm to accurately forecast GTE parameters. results show that both demonstrate accuracy, but perform better in most For vibration (VIB_N1FNT1, VIB_N1FNT2, VIB_N2FNT1, VIB_N2FNT2), achieved lower RMSE MAE confirming their higher accuracy. temperature (EGT1 EGT2), also showed accuracy rates. Meanwhile, some speed (N21 N22). findings emphasize necessity choosing appropriate model based nature specifics be forecast. Future may developing hybrid approaches combine advantages achieve optimal GTEs.

Язык: Английский

AI-Driven Predictive Maintenance in Mining: A Systematic Literature Review on Fault Detection, Digital Twins, and Intelligent Asset Management

Applied Sciences, Год журнала: 2025, Номер 15(6), С. 3337 - 3337

Опубликована: Март 19, 2025

The mining industry faces increasing challenges in maintaining high production levels while minimizing unplanned failures and operational costs. Critical assets, such as crushers, conveyor belts, mills, ventilation systems, operate under extreme conditions, leading to accelerated wear failure risks. Traditional maintenance strategies often fail prevent unexpected downtimes, safety hazards, economic losses. As a response, industries are integrating predictive monitoring technologies, including machine learning, the Internet of Things, digital twins, enhance early fault detection optimize strategies. This Systematic Literature Review analyzes 166 high-impact studies from Scopus Web Science, identifying key trends algorithms, hybrid AI models, real-time techniques. findings highlight adoption deep reinforcement twins for anomaly process optimization. Additionally, AI-driven methods improving sensor-based data acquisition asset management, extending equipment lifecycles reducing failures. Despite these advancements, standardization, model scalability, system interoperability persist, requiring further research. Future work should focus on applications, explainable academia-industry collaboration accelerate implementation intelligent solutions, ensuring greater reliability, efficiency, sustainability operations.

Язык: Английский