Sensor Fault Diagnosis Using Spectral Principal Component Analysis and CNN Deep Learning

Mou Jianqiang, Cui Shan

Abstract

A data driven methodology for sensor fault diagnosis in sensor network using principal component analysis (PCA) of coherence spectrum and convolutional neural network (CNN) deep learning is proposed. The methodology was evaluated with the measurement data of a sensor network for ambient relative humidity (RH) monitoring of a chemical laboratory. The results demonstrated accuracy up to 99% for sensor fault diagnosis in the sensor network functioning across a large spectrum of frequencies for environmental monitoring.

Download

IMEKO-TC6-2025-024.pdf

DOI

10.21014/tc6-2025.024

IMEKO TC

TC6 - Digitalization

Back to the proceedings