Physics informed neural networks for fault severity identification of axial piston pumps

Wang, Zhiying; Zhou, Zheng; Xu, Wengang; Sun, Chuang; Yan, Ruqiang

doi:10.1016/j.jmsy.2023.10.002

Cited by 6 publications

(1 citation statement)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhao et al [19] proposed a method for online parameter identification using a least squares recursive algorithm, enabling real-time, online identification for the diagnosis of abnormal leakage conditions in heat exchangers. Intelligent parameter identification methods include neural networks [20,21], particle swarm algorithms [22], genetic algorithms (GA) [23], and reinforcement learning (RL) technology. Wang et al [24] utilized a Radial Basis Function (RBF) neural network to establish quantitative relationships between system input variables and model parameters, enabling the identification of unknown parameters and obtaining a relatively accurate model for the heat exchange energy-saving system.…”

Section: Introductionmentioning

confidence: 99%

Dynamic Modeling of Heat Exchangers Based on Mechanism and Reinforcement Learning Synergy

Sun,

Jia,

Zhao

et al. 2024

Buildings

View full text Add to dashboard Cite

The current lack of a high-precision, real-time model applicable to the control optimization process of heat exchange systems, especially the difficulty in determining the overall heat transfer coefficient K of heat exchanger operating parameters in real time, is a prominent issue. This paper mainly unfolds the following work: 1. We propose a dynamic model for the control and optimization of the heat exchanger operation. By constructing a system to collect real-time operating data on the flow rates and temperatures on both sides of the heat exchanger, the parameter identification of the overall heat transfer coefficient K is performed. Subsequently, by combining this with mechanistic equations, a novel heat exchanger model is established based on the fusion of mechanistic principles and reinforcement learning. 2. We validate the new model, where the average relative error between the model’s temperature output values and the actual measured values is below 5%, indicating the high identification accuracy of the model. Moreover, under variations in the temperature and flow rate, the overall heat transfer coefficient K demonstrates the correct patterns of change. 3. To further enhance the model’s identification accuracy, a study on the reward functions in reinforcement learning is conducted. A model with the Logarithmic Mean Temperature Difference (LMTD) as the reward function exhibits a high identification accuracy. However, upon comparison, a model using the Arithmetic Mean Temperature Difference (AMTD) for relative error as the reward function shows an even higher identification accuracy. The model is validated under various operating conditions, such as changes in the flow rate on the hot side, demonstrating good scalability and applicability. This research contributes to providing a high-precision dynamic parameter basis for the precise control of heat exchange systems, offering significant guidance for the control optimization of actual heat exchange system operations.

show abstract

Section: Introductionmentioning

confidence: 99%

Dynamic Modeling of Heat Exchangers Based on Mechanism and Reinforcement Learning Synergy

Sun,

Jia,

Zhao

et al. 2024

Buildings

View full text Add to dashboard Cite

show abstract

Enhancing weld line visibility prediction in injection molding using physics-informed neural networks

Pieressa,

Baruffa,

Sorgato

et al. 2024

J Intell Manuf

View full text Add to dashboard Cite

This study introduces a novel approach using Physics-Informed Neural Networks (PINN) to predict weld line visibility in injection-molded components based on process parameters. Leveraging PINNs, the research aims to minimize experimental tests and numerical simulations, thus reducing computational efforts, to make the classification models for surface defects more easily implementable in an industrial environment. By correlating weld line visibility with the Frozen Layer Ratio (FLR) threshold, identified through limited experimental data and simulations, the study generates synthetic datasets for pre-training neural networks. This study demonstrates that a quality classification model pre-trained with PINN-generated datasets achieves comparable performance to a randomly initialized network in terms of Recall and Area Under the Curve (AUC) metrics, with a substantial reduction of 78% in the need for experimental points. Furthermore, it achieves similar accuracy levels with 74% fewer experimental points. The results demonstrate the robustness and accuracy of neural networks pre-trained with PINNs in predicting weld line visibility, offering a promising approach to minimizing experimental efforts and computational resources.

show abstract

Wavelet-driven differentiable architecture search for planetary gear fault diagnosis

Wang,

Zhou,

Yang

et al. 2024

Journal of Manufacturing Systems

View full text Add to dashboard Cite

Physics informed neural networks for fault severity identification of axial piston pumps

Cited by 6 publications

References 42 publications

Dynamic Modeling of Heat Exchangers Based on Mechanism and Reinforcement Learning Synergy

Dynamic Modeling of Heat Exchangers Based on Mechanism and Reinforcement Learning Synergy

Enhancing weld line visibility prediction in injection molding using physics-informed neural networks

Wavelet-driven differentiable architecture search for planetary gear fault diagnosis

Contact Info

Product

Resources

About