A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature Extraction

To address the challenges of spectral unmixing and feature extraction in the hot jet mixtures of two types of aero-engine hot jets, Fourier transform infrared spectrometry was employed to precisely measure these mixtures. The collected data encompassed the spectra of hot jets generated individually...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhenping Kang, Yurong Liao, Xinyan Yang, Zhaoming Li
Format: Article
Language:English
Published: MDPI AG 2025-03-01
Series:Remote Sensing
Subjects:
Online Access:https://www.mdpi.com/2072-4292/17/7/1147
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849769629166075904
author Zhenping Kang
Yurong Liao
Xinyan Yang
Zhaoming Li
author_facet Zhenping Kang
Yurong Liao
Xinyan Yang
Zhaoming Li
author_sort Zhenping Kang
collection DOAJ
description To address the challenges of spectral unmixing and feature extraction in the hot jet mixtures of two types of aero-engine hot jets, Fourier transform infrared spectrometry was employed to precisely measure these mixtures. The collected data encompassed the spectra of hot jets generated individually by the two distinct engine models, as well as those of the mutually mixed hot jets. In this paper, a mixed spectral unmixing algorithm based on VCA was put forward. Initially, the vertex component analysis (VCA) algorithm was utilized to decompose the mixed spectra. By comparing with the separately measured actual pure spectra, it was found that the mean RMSE of the hot jet pure spectra extracted by VCA for the two engines was 0.34846, and the mean SAM reached 0.00096, thus validating the effectiveness of the algorithm. Subsequently, the least squares (LS) algorithm was applied to ascertain the abundance values of the mixed spectra. Among the mixed samples, the average abundance values of the two pure spectra were 0.78 and 0.22, respectively. To further extract the spectral features after unmixing, an innovative one-dimensional convolutional multi-head self-attention mechanism neural network (MHSA-CNN) algorithm was devised in this study. This algorithm can accurately pinpoint the key wave crests of the features at 2282–2283 cm<sup>−1</sup> and 2388–2389 cm<sup>−1</sup>. The research findings offer crucial technical backing for the intelligent fault diagnosis of aero-engines and contribute to enhancing the accuracy and reliability of engine operating condition monitoring.
format Article
id doaj-art-1b17c9fb1d164e02b90be4dfa7184d26
institution DOAJ
issn 2072-4292
language English
publishDate 2025-03-01
publisher MDPI AG
record_format Article
series Remote Sensing
spelling doaj-art-1b17c9fb1d164e02b90be4dfa7184d262025-08-20T03:03:21ZengMDPI AGRemote Sensing2072-42922025-03-01177114710.3390/rs17071147A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature ExtractionZhenping Kang0Yurong Liao1Xinyan Yang2Zhaoming Li3Department of Electronic and Optical Engineering, Space Engineering University, Beijing 101416, ChinaDepartment of Electronic and Optical Engineering, Space Engineering University, Beijing 101416, ChinaDepartment of Electronic and Optical Engineering, Space Engineering University, Beijing 101416, ChinaDepartment of Electronic and Optical Engineering, Space Engineering University, Beijing 101416, ChinaTo address the challenges of spectral unmixing and feature extraction in the hot jet mixtures of two types of aero-engine hot jets, Fourier transform infrared spectrometry was employed to precisely measure these mixtures. The collected data encompassed the spectra of hot jets generated individually by the two distinct engine models, as well as those of the mutually mixed hot jets. In this paper, a mixed spectral unmixing algorithm based on VCA was put forward. Initially, the vertex component analysis (VCA) algorithm was utilized to decompose the mixed spectra. By comparing with the separately measured actual pure spectra, it was found that the mean RMSE of the hot jet pure spectra extracted by VCA for the two engines was 0.34846, and the mean SAM reached 0.00096, thus validating the effectiveness of the algorithm. Subsequently, the least squares (LS) algorithm was applied to ascertain the abundance values of the mixed spectra. Among the mixed samples, the average abundance values of the two pure spectra were 0.78 and 0.22, respectively. To further extract the spectral features after unmixing, an innovative one-dimensional convolutional multi-head self-attention mechanism neural network (MHSA-CNN) algorithm was devised in this study. This algorithm can accurately pinpoint the key wave crests of the features at 2282–2283 cm<sup>−1</sup> and 2388–2389 cm<sup>−1</sup>. The research findings offer crucial technical backing for the intelligent fault diagnosis of aero-engines and contribute to enhancing the accuracy and reliability of engine operating condition monitoring.https://www.mdpi.com/2072-4292/17/7/1147FT-IRaero-engineunmixingfeature extraction
spellingShingle Zhenping Kang
Yurong Liao
Xinyan Yang
Zhaoming Li
A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature Extraction
Remote Sensing
FT-IR
aero-engine
unmixing
feature extraction
title A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature Extraction
title_full A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature Extraction
title_fullStr A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature Extraction
title_full_unstemmed A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature Extraction
title_short A VCA Combined MHSA-CNN for Aero-Engine Hot Jet Remote Sensing Mixed Spectral Feature Extraction
title_sort vca combined mhsa cnn for aero engine hot jet remote sensing mixed spectral feature extraction
topic FT-IR
aero-engine
unmixing
feature extraction
url https://www.mdpi.com/2072-4292/17/7/1147
work_keys_str_mv AT zhenpingkang avcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction
AT yurongliao avcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction
AT xinyanyang avcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction
AT zhaomingli avcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction
AT zhenpingkang vcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction
AT yurongliao vcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction
AT xinyanyang vcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction
AT zhaomingli vcacombinedmhsacnnforaeroenginehotjetremotesensingmixedspectralfeatureextraction