Physics-informed machine learning for sensor fault detection with flight test data

Silva, Brian M. de; Callaham, Jared; Jonker, Jonathan; Goebel, Nicholas; Klemisch, Jennifer; McDonald, D. C.; Hicks, Nathan D.; Kutz, J. Nathan; Brunton, Steven L.; Aravkin, Aleksandr Y.

doi:10.48550/arxiv.2006.13380

Cited by 4 publications

(6 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The term y k − H k x− k is referred to as innovation or residual and describes the difference between the prediction and actual measurements. Here, we use the moving innovation covariance to describe the characteristics of anomalous behavior, following Mehra and Peschon [44], Hajiyev [45] and Silva et al [39,40], which is defined as…”

Section: Kalman Filtermentioning

confidence: 99%

“…Fathi et al [37] and Jiang and Liu [38] used KF and its variants to denoise observation data and proposed KF-DMD, EnKF-DMD, and DMD-KF-W methods to reconstruct noisy deterministic dynamic systems and random dynamical systems. Silva et al [39,40] proposed a physically enhanced data-driven fault detection method that uses dynamic mode decomposition with control (DMDc) and Kalman filter to generate residuals and also provide a decision tree for fault diagnosis. To the best of our knowledge, this is the first time in the field of FDIR that a combination of the DMD and KF methods was used.…”

Section: Introductionmentioning

confidence: 99%

“…To the best of our knowledge, this is the first time in the field of FDIR that a combination of the DMD and KF methods was used. However, Silva et al [39,40] only provide the binary classification of whether or not the fault occurs; moreover, extraction and analysis of fault features are insufficient. For FDIR problems, it is essential to accurately determine the fault type for fault isolation, controller reconfiguration, and faulttolerance control.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A data-driven hybrid sensor fault detection/diagnosis method with flight test data

Song,

Chen,

Wang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Fault diagnosis of multi-source signal systems is critical for the safe and reliable operation of modern industrial systems, and accurate fault diagnosis of systems based on multi-source signals remains challenging. This study proposes a data-driven hybrid fault detection/diagnosis method to identify sensor faults in complex systems through interactions between multiple sensors. Dynamic Mode Decomposition with control (DMDc) is used to obtain the approximate model of the investigated system from multi-source signals, extract the underlying physical mechanisms, combined with the Kalman filter observer to generate the residual between the observed data and the predicted data. Then the residual (moving innovation covariance matrix V k) is input into the k-nearest neighbor (kNN) classification algorithm for fault detection and diagnosis. The effectiveness of the proposed method was evaluated using the flight test braking system dataset. The results showed that the accuracy of the proposed method in fault detection and diagnosis (with accuracies of 100% and 100%, respectively) was significantly improved than that of using raw signal data (76.6% and 6.38%) or raw signal data and V k (80.85% and 42.55%). The analysis of different parameters including fault severity, algorithm hyperparameter k, and sensor type showed that the proposed method has high robustness, generalization ability, and practicality.

show abstract

Section: Kalman Filtermentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A data-driven hybrid sensor fault detection/diagnosis method with flight test data

Song,

Chen,

Wang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…A sensor anomaly is a degradation of its performance to a certain threshold, usually caused by catastrophic failures and more subtle failures, for instance, outdated calibration, sensor deformation, and low-frequency oscillations [6]. The performance cannot be directly characterized and is implicit in the flight test data, which is characterized by long time series, imbalance and small differences between classes.…”

Section: Introductionmentioning

confidence: 99%

Imbalanced Aircraft Data Anomaly Detection

Hao¹,

Ji²,

Yuan³

et al. 2023

Preprint

View full text Add to dashboard Cite

“…The physics-informed machine learning has been growing fast in popularity and applied to several unique forward and inverse problems [47,48,49,50,51,52,53,54]. Extensive reviews of the current state in physics-informed machine learning are available in literature [55,56].…”

Section: Introductionmentioning

confidence: 99%

Physics and Equality Constrained Artificial Neural Networks: Application to Forward and Inverse Problems with Multi-fidelity Data Fusion

Basir,

Senocak

2021

Preprint

View full text Add to dashboard Cite

Physics-informed neural networks (PINNs) have been proposed to learn the solution of partial differential equations (PDE). In PINNs, the residual form of the PDE of interest and its boundary conditions are lumped into a composite objective function as an unconstrained optimization problem, which is then used to train a deep feed-forward neural network.Here, we show that this specific way of formulating the objective function is the source of severe limitations in the PINN approach when applied to different kinds of PDEs. To address these limitations, we propose a versatile framework that can tackle both inverse and forward problems. The framework is adept at multi-fidelity data fusion and can seamlessly constrain the governing physics equations with proper initial and boundary conditions. The backbone of the proposed framework is a nonlinear, equality-constrained optimization problem formulation aimed at minimizing a loss functional, where an augmented Lagrangian method (ALM) is used to formally convert a constrained-optimization problem into an unconstrainedoptimization problem. We implement the ALM within a stochastic, gradient-descent type training algorithm in a way that scrupulously focuses on meeting the constraints without sacrificing other loss terms. Additionally, as a modification of the original residual layers, we propose lean residual layers in our neural network architecture to address the so-called vanishing-gradient problem. We demonstrate the efficacy and versatility of our physics-and equality-constrained deep-learning framework by applying it to learn the solutions of various multi-dimensional PDEs, including a nonlinear inverse problem from the hydrology field with multi-fidelity data fusion. The results produced with our proposed model match exact solutions very closely for all the cases considered.

show abstract

Physics-informed machine learning for sensor fault detection with flight test data

Cited by 4 publications

References 35 publications

A data-driven hybrid sensor fault detection/diagnosis method with flight test data

A data-driven hybrid sensor fault detection/diagnosis method with flight test data

Imbalanced Aircraft Data Anomaly Detection

Physics and Equality Constrained Artificial Neural Networks: Application to Forward and Inverse Problems with Multi-fidelity Data Fusion

Contact Info

Product

Resources

About