Title: An empirical study of fault diagnosis methods of a dissolved oxygen sensor based on ResNet-50

Authors: Pu Yang; Qinghao Liu; Boning Wang; Weiran Li; Zhenbo Li; Ming Sun

Addresses: College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China; National Innovation Centre for Digital Fishery, Beijing, 100083, China; Engineering and Technology Research Centre, Internet of Things in Agriculture, Beijing, 100083, China ' College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China ' College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China; National Innovation Centre for Digital Fishery, Beijing, 100083, China; Engineering and Technology Research Centre, Internet of Things in Agriculture, Beijing, 100083, China ' College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China; National Innovation Centre for Digital Fishery, Beijing, 100083, China; Engineering and Technology Research Centre, Internet of Things in Agriculture, Beijing, 100083, China ' College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China; National Innovation Centre for Digital Fishery, Beijing, 100083, China; Engineering and Technology Research Centre, Internet of Things in Agriculture, Beijing, 100083, China ' College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China; National Innovation Centre for Digital Fishery, Beijing, 100083, China; Engineering and Technology Research Centre, Internet of Things in Agriculture, Beijing, 100083, China

Abstract: This work diagnoses online faults by classifying the real-time data collected by a dissolved oxygen sensor. Based on the three types of fault classification of dissolved oxygen parameters viz., complete failure faults, fixed deviation faults, and drifting faults, VGGNet, GoogLeNet, and ResNet-50 deep learning neural network models are used, respectively, to detect the faults. The experimental results show that the performance of the ResNet-50 model is the best, with a 98% dissolved oxygen failure accuracy rate. Whereas the accuracy rate of VGGNet and GoogLeNet can reach up to 0.9. The comparative experimental results of VGGNet and GoogLeNet show that the training loss function converges at 0.2. In contrast, the loss function of VGGNet and GoogLeNet after 150 rounds of training can only be reduced to about 0.3 and 0.5, respectively. The proposed model (ResNet-50) has good accuracy and reasonable generalisation ability.

Keywords: dissolved oxygen; DO; fault diagnosis; deep learning; ResNet-50; prediction accuracy.

DOI: 10.1504/IJSNET.2022.124566

International Journal of Sensor Networks, 2022 Vol.39 No.3, pp.205 - 214

Received: 02 Nov 2021
Accepted: 24 Nov 2021
Published online: 28 Jul 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article