Physics Constrained High-Precision Data-Driven Modeling for Multi-Path Ultrasonic Flow Meter in Natural Gas...

Area of Science:

• Mechanical Engineering
• Fluid Dynamics
• Data Science

Background:

• Ultrasonic flow meters are essential for natural gas transportation, requiring accurate flow velocity data for performance analysis.
• Existing modeling methods face challenges in achieving the high accuracy demanded by practical applications.

Purpose of the Study:

• To develop an accurate and interpretable ultrasonic flow meter modeling method.
• To enhance the analysis of metering performance and flow processes in natural gas pipelines.

Main Methods:

• A hybrid approach combining data learning with industrial physics knowledge for ultrasonic flow meter modeling.
• Utilizing pipeline flow field velocity distribution for data preprocessing and loss function design.
• Building flow velocity prediction models under various working conditions.

Main Results:

• The proposed method achieves prediction results closely matching real acoustic path flow velocity distributions.
• Experimental validation demonstrates prediction errors can be controlled within 1%.
• The model exhibits high accuracy and interpretability.

Conclusions:

• The developed method effectively meets the stringent accuracy requirements for ultrasonic flow meter modeling.
• This approach supports improved performance analysis and operational efficiency in natural gas transportation.
• Integrating physics-based knowledge with data learning offers a robust solution for complex flow measurement challenges.