Diffusion Model based Time-Series Data Generation Method for Augmenting Programmable Motion Fault Data of Collaborative Robots 


Vol. 14,  No. 2, pp. 113-122, Feb.  2025
https://doi.org/10.3745/TKIPS.2025.14.2.113


PDF
  Abstract

Collaborative robots, with their programmable features, can perform various tasks based on operator commands. However, due to diverse patterns of sensor data depending on operating conditions, it is difficult to acquire sufficient data for fault diagnosis. This paper proposes a Diffusion Model-based fault data generation method considering time-series sensor data characteristics of collaborative robots. The method involves injecting missing values into normal data at equal intervals, imputing the missing values using a Diffusion Model trained on fault features, and merging the imputed values to generate synthetic fault data. To validate the method, torque data from a faulty actuator was used to generate fault data under various conditions. Cosine similarity analysis showed an average similarity of 0.985 between the synthetic and actual fault data, confirming the synthetic data reflects real patterns. Additionally, fault classification using a TCN(Temporal CNN) showed over 91% accuracy across conditions, proving the generated data captures actual fault characteristics and is useful for fault diagnosis.

  Statistics
  Cite this article

[IEEE Style]

M. S. Choi, J. S. Kim, J. W. Lee, "Diffusion Model based Time-Series Data Generation Method for Augmenting Programmable Motion Fault Data of Collaborative Robots," The Transactions of the Korea Information Processing Society, vol. 14, no. 2, pp. 113-122, 2025. DOI: https://doi.org/10.3745/TKIPS.2025.14.2.113.

[ACM Style]

Min Seo Choi, Jin Se Kim, and Jung Won Lee. 2025. Diffusion Model based Time-Series Data Generation Method for Augmenting Programmable Motion Fault Data of Collaborative Robots. The Transactions of the Korea Information Processing Society, 14, 2, (2025), 113-122. DOI: https://doi.org/10.3745/TKIPS.2025.14.2.113.

Forms

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)

Old Journals

Indexing

All publications of TKIPS is indexed in DOI, EBSCO, Google Scholar, Crossref, and CrossCheck.

TKIPS is also selected as the Journal for Accreditation by NRF (National Research Foundation of Korea).

Related Journals