В статье представлен метод прогнозирования гидравлического давления в системах водоснабжения с использованием рекуррентных нейронных сетей, в частности модели с долгой краткосрочной памятью (LSTM). Основываясь на данных, получаемых с пьезоэлектрических датчиков давления, интегрированных в сеть через протокол MQTT, исследование демонстрирует значительное превосходство LSTM- модели над традиционными моделями. Это подчеркивает не только высокую точность прогноза, но и возможность более эффективного управления системами водоснабжения. Преимущество данной архитектуры заключается в ее гибкости и способности адаптироваться к изменяющимся условиям эксплуатации, благодаря обработке новых данных, поступающих на сервер в реальном времени. Базовые принципы построения и настройки модели LSTM могут быть легко модифицированы для прогнозирования различных параметров, что расширяет перспективы использования рекуррентных нейросетей в управлении инженерными системами.

The method for forecasting hydraulic pressure in water supply systems using recurrent neural networks, specifically the long short-term memory (LSTM) model is presented. Based on data obtained from piezoelectric pressure sensors integrated into the network via the MQTT protocol, the study demonstrates a significant superiority of the LSTM model over traditional models. This fact highlights not only the high accuracy of the forecast but also the potential for more efficient water supply system management. The advantage of this architecture lies in its flexibility and ability to adapt to changing operational conditions through real-time processing of new data sent to the server. The basic principles of constructing and configuring the LSTM model can be easily modified to forecast various system parameters, expanding the prospects for using recurrent neural networks in managing engineering systems.

Thermally grown oxide (TGO) layers—primarily alumina (Al2O3)—provide
oxidation resistance and high-temperature protection for thermal barrier coatings. However, during their service in humid and hot environments, water vapor accelerates TGO
degradation by stabilizing metastable alumina phases (e.g., θ-Al2O3) and inhibiting their
transformation to the thermodynamically stable α-Al2O3, a phenomenon which has been
shown in numerous experimental studies. However, the microscopic mechanisms by which
water vapor affects the phase stability and transformation of alumina remain unresolved.
This study employs first-principles calculations to investigate hydrogen’s role in altering
vacancy formation, aggregation, and atomic migration in θ- and α-Al2O3. The results
reveal that hydrogen incorporation reduces the formation energies for aluminum and
oxygen vacancies by up to 40%, promoting defect generation and clustering; increases
aluminum migration barriers by 25–30% while lowering oxygen migration barriers by
15–20%, creating asymmetric diffusion kinetics; and stabilizes oxygen-deficient sublattices,
disrupting the structural reorganization required for θ- to α-Al2O3 transitions. These effects
collectively sustain metastable θ-Al2O3 growth and delay phase stabilization. By linking
hydrogen-induced defect dynamics to macroscopic coating degradation, this work provides
atomic-scale insights for designing moisture-resistant thermal barrier coatings through the
targeted inhibition of vacancy-mediated pathways.