A novel ensemble network based on CNN‐AM‐BiLSTM learner for temperature prediction of distillation columns

The Canadian Journal of Chemical Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: March 5, 2025

Abstract In recent years, complexity has significantly increased in chemical processes where a distillation column serves as crucial unit. It is worthwhile to develop an accurate and reliable predictive model maintain the steady operation condition of column. Although data‐driven models that do not rely on any prior knowledge present promising approach, they encounter challenges associated with nonlinearity dynamic behaviour within process data. To tackle these challenges, deep learning‐based combined distilled spatiotemporal attention ensemble network (CDSAEN) proposed. The CDSAEN constructed by sequentially integrating multiple base learners, which are iteratively generated decreasing span lengths through boosting method implemented specially designed extraction evaluation function. learner, convolutional neural (CNN), mechanism (AM), bidirectional long short‐term memory (BiLSTM) utilized adaptively capture intricate features establish robust mapping relationship from inputs output. Real‐world data system plant reconstructed time series dataset subsequently fed into for training forecast temperature apparatus advance. results exhibited effectiveness reliability. Additionally, comparison six other approaches, proposed attained superior performance mean absolute error (MAE) = 0.084, root squared (RMSE) 0.108, R 2 0.974. This study can provide support maintaining stable columns processes.

Language: Английский

---

A Deep Learning-Based Acoustic Signal Analysis Method for Monitoring the Distillation Columns’ Potential Faults

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(16), P. 7026 - 7026

Published: Aug. 10, 2024

Distillation columns are vital for substance separation and purification in various industries, where malfunctions can lead to equipment damage, compromised product quality, production interruptions, environmental harm. Early fault detection using AI-driven methods like deep learning mitigate downtime safety risks. This study employed a lab-scale distillation column collect passive acoustic signals under normal conditions three potential faults: flooding, dry tray, leakage. Signal processing techniques were used extract features from low signal-to-noise ratios weak time-domain characteristics. A learning-based feature recognition method was then applied, achieving an average accuracy of 99.03% on Mel-frequency cepstral coefficient (MFCC) spectrogram datasets. demonstrated robust performance across different types limited data scenarios, effectively predicting detecting faults columns.

Language: Английский

---

State monitoring and fault prediction of centrifugal compressors based on long short–term memory and principal component analysis (LSTM-PCA)

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2433 - e2433

Published: Oct. 21, 2024

Centrifugal compressors are widely used in the petroleum and natural gas industry for compression, reinjection, transportation. Early fault identification evolution prediction centrifugal can improve equipment safety reduce maintenance operating costs. This article proposes a dynamic process monitoring method based on long short-term memory (LSTM) principal component analysis (PCA). constructs sliding window at each sampling point, which contains 100 data from past current time points, uses LSTM to predict 30 future points. At same time, this is also combined with PCA threshold construct new LSTM-PCA algorithm. And was validated using compressor data. The results show that effectively detect anomalies, improvements significantly reduced false positive rate of detected make multi-step advance predictions system behavior after faults occur.

Language: Английский

---