EUVIMEDEUVIMED entdecken
Crossref

Cross-domain adaptive imputation method for industrial time series based on joint time domain and frequency domain modeling

Huixin Tian, Shikang Lu, Hongyi Wang, Kun Li

Measurement Science and Technology · 29.05.2026 · 37 (22) · 226001

DOI: 10.1088/1361-6501/ae6abd ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 28.05.2026. Verfasst von Huixin Tian und 3 weiteren Autor:innen. Die Arbeit nennt 41 Literaturangaben. Gefördert von: Natural Science Foundation of Tianjin Municipality.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The missing values are a common issue in data collected by sensors in the industrial field. They severely affect the reliability of subsequent data-driven monitoring and control of industrial processes. Accurate imputation is crucial for reliable industrial system operation. However, existing methods often decouple time domain and frequency domain information, limiting their ability to model complex temporal dynamics. In high missing-rate scenarios, they further suffer from accumulated pre-imputation bias and error propagation. To address these challenges, this paper proposes a cross-training adaptive time series imputation (CATI) model based on joint time domain and frequency domain modeling. First, a time domain and frequency domain cross-training strategy is developed to achieve collaborative modeling of time domain and frequency domain features through cross-domain information sharing. Then, an adaptive dynamic loss (ADL) and a pre-imputation updating mechanism are designed to effectively reduce error accumulation, thereby improving the model robustness and imputation accuracy. Finally, the proposed CATI model is validated on six public datasets and one real-world industrial dataset. Moreover, ablation studies further confirm the essential contributions and robustness of each module in the imputation framework.

Förderung

  • Natural Science Foundation of Tianjin Municipality

Zitieren

APA

Tian, H., Lu, S., Wang, H., & Li, K. (2026). Cross-domain adaptive imputation method for industrial time series based on joint time domain and frequency domain modeling. Measurement Science and Technology, 37(22), 226001. https://doi.org/10.1088/1361-6501/ae6abd

Vancouver

Tian H, Lu S, Wang H, Li K. Cross-domain adaptive imputation method for industrial time series based on joint time domain and frequency domain modeling. Measurement Science and Technology. 2026;37(22):226001. doi:10.1088/1361-6501/ae6abd

BibTeX
@article{Tian2026,
  author = {Tian, Huixin and Lu, Shikang and Wang, Hongyi and Li, Kun},
  title = {Cross-domain adaptive imputation method for industrial time series based on joint time domain and frequency domain modeling},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {22},
  pages = {226001},
  doi = {10.1088/1361-6501/ae6abd}
}
Kostenlos starten →