Denoising stacked autoencoders‐based near‐infrared quality monitoring method via robust samples evaluation

This paper proposes a denoising stacked autoencoders‐based near‐infrared spectroscopy on‐line quality monitoring model via robust sample evaluation. The previous related work tends to focus on the near infrared spectrum data from the high dimension, multicollinearity, and information redundancy, and...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Canadian journal of chemical engineering Ročník 101; číslo 5; s. 2693 - 2703
Hlavní autori: Lv, Jiapeng, Chen, Zihao, Luan, Xiaoli, Liu, Fei
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Hoboken, USA John Wiley & Sons, Inc 01.05.2023
Wiley Subscription Services, Inc
Predmet:
Denoising stacked auto‐encoders
Distillation
Feature extraction
Moisture effects
Monitoring
Near infrared radiation
near‐infrared spectroscopy
Noise reduction
Nonlinearity
quality monitoring
Redundancy
robust sample evaluation
Robustness
Samples
Sensitivity
Spectrum analysis
ISSN:0008-4034, 1939-019X
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:This paper proposes a denoising stacked autoencoders‐based near‐infrared spectroscopy on‐line quality monitoring model via robust sample evaluation. The previous related work tends to focus on the near infrared spectrum data from the high dimension, multicollinearity, and information redundancy, and so on, but pays less attention to its inherent nonlinearity and sensitivity caused by internal and external factors (i.e., particle size, colour, moisture, uniformity, temperature, and so on). First, this paper aims to achieve feature extraction in nonlinearity with a denoising stack autoencoder, overcoming the impact of over‐sensitivity described as given distributed noise. Second, we propose a robust sample evaluation method derived from the robust statistics to tell and eliminate the individual samples contrary to the potentially statistical rules learned by the established model from the population samples and retrain the model with a more robust training set. The near‐infrared spectrum data derived from the distillation process of 2, 6 xylenol are used as a case in this paper to verify the validity and accuracy of the monitoring model proposed above.
Bibliografia:Funding information
National Natural Science Foundation of China, Grant/Award Numbers: 61833007, 61991402
ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0008-4034
1939-019X
DOI:10.1002/cjce.24684