TY - JOUR
T1 - Sliced Wasserstein cycle consistency generative adversarial networks for fault data augmentation of an industrial robot
AU - Pu, Ziqiang
AU - Cabrera, Diego
AU - Li, Chuan
AU - Valente de Oliveira, José
N1 - Publisher Copyright:
© 2023 Elsevier Ltd
PY - 2023/7/15
Y1 - 2023/7/15
N2 - We investigate the role of the loss function in cycle consistency generative adversarial networks (CycleGANs). Namely, the sliced Wasserstein distance is proposed for this type of generative model. Both the unconditional and the conditional CycleGANs with and without squeeze-and-excitation mechanisms are considered. Two data sets are used in the evaluation of the models, i.e., the well-known MNIST and a real-world in-house data set acquired for an industrial robot fault diagnosis. A comprehensive set of experiments show that, for both the unconditional and the conditional cases, sliced Wasserstein distance outperforms classic Wasserstein distance in CycleGANs. For the robot faulty data augmentation a model compatibility of 99.73% (conditional case) and 99.21% (unconditional case) were observed. In some cases, the improvement in convergence efficiency was higher than 2 (two) orders of magnitude.
AB - We investigate the role of the loss function in cycle consistency generative adversarial networks (CycleGANs). Namely, the sliced Wasserstein distance is proposed for this type of generative model. Both the unconditional and the conditional CycleGANs with and without squeeze-and-excitation mechanisms are considered. Two data sets are used in the evaluation of the models, i.e., the well-known MNIST and a real-world in-house data set acquired for an industrial robot fault diagnosis. A comprehensive set of experiments show that, for both the unconditional and the conditional cases, sliced Wasserstein distance outperforms classic Wasserstein distance in CycleGANs. For the robot faulty data augmentation a model compatibility of 99.73% (conditional case) and 99.21% (unconditional case) were observed. In some cases, the improvement in convergence efficiency was higher than 2 (two) orders of magnitude.
KW - Conditional cycle consistency generative adversarial networks
KW - Cycle consistency generative adversarial networks
KW - Generative adversarial networks
KW - Industrial robots
KW - Scarce faulty data augmentation
KW - Sliced Wasserstein distance
UR - http://www.scopus.com/inward/record.url?scp=85150792725&partnerID=8YFLogxK
U2 - 10.1016/j.eswa.2023.119754
DO - 10.1016/j.eswa.2023.119754
M3 - Article
AN - SCOPUS:85150792725
SN - 0957-4174
VL - 222
JO - Expert Systems with Applications
JF - Expert Systems with Applications
M1 - 119754
ER -