Functional Principal Component Analysis for Extrapolating Multistream Longitudinal Data

Chung, Seokhyun; Kontar, Raed Al

doi:10.1109/tr.2020.3035084

Cited by 6 publications

(4 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The presence of such categorical disparities often results in significant variations in CM signals among units based on their respective labels. That said, in practices such as degradation-based CM, these variations remain dormant initially and become evident only at a later stage due to cumulative effects [4]. For example, while initially presenting as healthy, lithium-ion batteries from a disqualified lot often exhibit abnormal degradation trends at the later stage, unacceptably deviating from the degradation trend of normal batteries, as presented in our case study in Sec.…”

Section: Introductionmentioning

confidence: 73%

“…While facilitating fast adaptation to online data, such parametric modeling results in severe vulnerability to model misspecification. To tackle this, another line of two-step approaches inspired by functional data analysis has been proposed [21,4,22]. They estimate eigenfunctions from historical CM signals and express each signal as a linear combination of the eigenfunctions.…”

Section: Related Workmentioning

confidence: 99%

“…We call this as the dormant stage. This is common in CM practices, where the future evolution of CM signals is dependent on different operational modes or configurations (i.e., labels), but their heterogeneity is not easily detected in the early stage [4]. Predicting a future trend solely based on online observations at x ∈ (0, 3] would be very challenging in such a case.…”

Section: Simulation Studymentioning

confidence: 99%

“…Conceptually, it can be thought of as dimensionality reduction for functional data, which encodes functions to the corresponding finite-dimensional vectors in the space "spanned" by a finite number of orthonormal basis functions [4]. Here the projected vectors and the basis functions are called functional principal component (FPC) scores and eigenfunctions, respectively.…”

Section: B Functional Data Augmentationmentioning

confidence: 99%

See 3 more Smart Citations

Federated Condition Monitoring Signal Prediction With Improved Generalization

Chung

Kontar

2024

IEEE Trans. Rel.

View full text Add to dashboard Cite

Building a predictive model that rapidly adapts to real-time condition monitoring (CM) signals is critical for engineering systems/units. Unfortunately, many current methods suffer from a trade-off between representation power and agility in online settings. For instance, parametric methods that assume an underlying functional form for CM signals facilitate efficient online prediction updates. However, this simplification leads to vulnerability to model specifications and an inability to capture complex signals. On the other hand, approaches based on overparameterized or non-parametric models can excel at explaining complex nonlinear signals, but real-time updates for such models pose a challenging task. In this paper, we propose a neural process-based approach that addresses this trade-off. It encodes available observations within a CM signal into a representation space and then reconstructs the signal's history and evolution for prediction. Once trained, the model can encode an arbitrary number of observations without requiring retraining, enabling on-the-spot real-time predictions along with quantified uncertainty and can be readily updated as more online data is gathered. Furthermore, our model is designed to incorporate qualitative information (i.e., labels) from individual units. This integration not only enhances individualized predictions for each unit but also enables joint inference for both signals and their associated labels. Numerical studies on both synthetic and real-world data in reliability engineering highlight the advantageous features of our model in real-time adaptation, enhanced signal prediction with uncertainty quantification, and joint prediction for labels and signals.

show abstract

Section: Introductionmentioning

confidence: 73%

Section: Related Workmentioning

confidence: 99%

Section: Simulation Studymentioning

confidence: 99%

Section: B Functional Data Augmentationmentioning

confidence: 99%

See 2 more Smart Citations