Predicting structural deterioration of large-scale building clusters using snapshot data: an integrated Markov-LSTM model

Building Research & Information, Год журнала: 2025, Номер unknown, С. 1 - 17

Опубликована: Март 5, 2025

To ensure a safe environment for occupants, predicting long-term structural deterioration of buildings is critical. However, existing models have limited capability to predict with mathematical tractability and accuracy, especially large-scale building clusters. address this gap, study aims establish new integrated Markov-LSTM model, combining the strengths model-driven data-driven methods, enhanced prediction. Specifically, proposed two-stage inhomogeneous Markov chain allows process be tractable through derivation analytical transition probabilities. further improve long short-term memory (LSTM) employed residuals calculated from Markov-based predictions true values. The performance model evaluated two case studies, using snapshot data results demonstrate significant improvements over benchmark models, reduction mean absolute error (MAE) by an average 0.1780 (and 0.3292), squared (MSE) 0.1421 0.5717), percentage (MAPE) 4.7778% 13.2736%) in Case 1 2). This contributes research practice prediction providing both focusing on clusters, supporting more effective condition-based maintenance.

Язык: Английский

Graph comparison efficient conditional generative adversarial networks for parameter identification of synchronous generators

Expert Systems with Applications, Год журнала: 2025, Номер unknown, С. 126449 - 126449

Опубликована: Янв. 1, 2025

Язык: Английский

An Audio-Based Motor-Fault Diagnosis System with SOM-LSTM

Applied Sciences, Год журнала: 2024, Номер 14(18), С. 8229 - 8229

Опубликована: Сен. 12, 2024

This paper combines self-organizing mapping (SOM) and a long short-term memory network (SOM-LSTM) to construct an audio-based motor-fault diagnosis system for identifying the operating states of rotary motor. first uses audio signal collector measure motor sound data, fast Fourier transform (FFT) convert actual measured sound–time-domain into frequency-domain signal, normalizes calibrates ensure consistency accuracy signal. Secondly, SOM is used further analyze characterized waveforms in order reveal intrinsic structure pattern data. The LSTM process secondary data generated via SOM. Dimensional aggregation prediction sequence long-term dependencies accurately identify different possible abnormal patterns. also experimental design Taguchi method optimize parameters SOM-LSTM increase execution efficiency fault diagnosis. Finally, applied real-time monitoring operation, work type performed, tests under loads environments are attempted evaluate its feasibility. completion this provides diagnostic strategy that can be followed when it comes faults. Through system, conditions equipment detected, which help with preventive maintenance, make more efficient save lot time costs, improve industry’s ability monitor operation information.

Язык: Английский