Coupling data-driven and model-based methods to improve fault diagnosis

Atoui, Mohamed Amine; Cohen, Achraf

doi:10.1016/j.compind.2021.103401

Cited by 36 publications

(6 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Pagnier and Chertkov [23] presented a hybrid framework that incorporated physics modeling of power systems into graph neural network, which contributed to reliable and explainable real-time predictions. Atoui and Cohen [24] combined model-based methods with pure data-driven methods using structured residuals, improving the classification rates on single faults, multiple simultaneous faults, and unknown operation conditions. Besides, in classical GNNs, such as graph convolution network (GCN) [25], approximate personalized propagation of neural predictions (APPNP) [26], ARMA neural network [27], a fixed adjacency matrix is used, which neglects the personalized information propagation needs in each layer.…”

Section: Literature Reviewmentioning

confidence: 99%

Adaptive physics-encoded graph neural network for health stage assessment of liquid-propellant rocket engines

Ma,

Wang,

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

The improvement of reliable health monitoring system (HMS) for liquid-propellant rocket engines (LREs) is a crucial part for reusable launch vehicle, which contributes to providing competitive and cost-effective propulsion systems. Thus, it accentuates the need for reliable and quick health stage assessment of system and follow-up damage-mitigating control. In this paper, we propose a novel adaptive physics-encoded graph neural network (APGNN) for health stage assessment of LREs. Our approach embeds the relations of different sensors obtained through expert experience, which contributes to constructing a physical graph layer. To better capture the information contained in all the sensor data, a novel convolutional layer of adaptive auto-regressive moving average (ARMA) filters is designed, which considers the personalized information propagation needs of each neural network layer. The performance of the proposed method is quantified with data obtained from physics simulations and real-world engineering systems. The results show that our model has potential applicability for the health stage assessment of LREs with high accuracy.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Adaptive physics-encoded graph neural network for health stage assessment of liquid-propellant rocket engines

Ma,

Wang,

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…Prognosis models are divided into model-based models, data-driven models and hybrid models. Model-based approach is defined as models that rely on physic configuration of system [19,20]. It includes physics-based model and digital simulation of the system.…”

Section: Review Of Reviewsmentioning

confidence: 99%

A review of failure prognostics for predictive maintenance of offshore wind turbines

Zhang

Vatn

Rasheed³

2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Offshore wind turbines (OWTs) are important facilities for wind power generation because of their low land use and high electricity output. However, the harsh environment and remote location of offshore sites make it difficult to conduct maintenance on turbines. To upkeep OWTs cost-effectively, predictive maintenance (PdM) is an appealing strategy for offshore wind industry. The heart of PdM is failure prognostics, which aims to predict an asset’s remaining useful life (RUL) based on condition monitoring (CM). To provide references to PdM of OWTs, this paper presents a systematic review of failure prognostic models for wind turbines. In this review, data-driven models, model-based models, and hybrid models are classified and presented for model selection. The findings reveal that it is promising to develop hybrid models in the future and combine the advantages of data-driven and model-based models. Currently, the internal combinations of machine learning methods and statistical approaches in data-driven models are more common than exterior linkages between data-driven models and model-based models. The limitations and strengths of different models are discussed, and opportunities for developing hybrid models are highlighted in the conclusion.

show abstract

“…Moreover, model-based and data driven hybrid method, establish the system model directly by integrating the multi-source sensor data information, reduce the impact of measurement noise, and use an effective feature recognition method for fault diagnosis [20,21]. Typical cases include [22,23]. The combination of model-based and datadriven methods usually refers to two key techniques: system modeling and dynamic system state estimation.…”

Section: Introductionmentioning

confidence: 99%

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

Coupling data-driven and model-based methods to improve fault diagnosis

Cited by 36 publications

References 32 publications

Adaptive physics-encoded graph neural network for health stage assessment of liquid-propellant rocket engines

Adaptive physics-encoded graph neural network for health stage assessment of liquid-propellant rocket engines

A review of failure prognostics for predictive maintenance of offshore wind turbines

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

Contact Info

Product

Resources

About