Session: 05-16 Instrumentation IV: AI-based Improvements

Paper Number: 126589

126589 - Neural Network Based Digital Twin for Performance Prediction of Water Brake Dynamometer 

Water brake dynamometer, the core component of aerospace engine testing facilities, is widely used in turbine component testing and turboshaft engine testing. Since the status of the water brake dynamometer system cannot be monitored in detail, equipment maintenance is performed solely based on the operator's experience, resulting in high risks in the dynamometer equipment operation. Post-test inspections may even reveal damage to the dynamometer bearings and key components. Cavitation and other phenomena require urgent technical solutions to improve health monitoring of key equipment and experimental safety.

This paper proposes a performance prediction method for water brake dynamometers based on machine learning. By conducting physical correlation analysis of key parameters, the method constructs a physical model of water brake dynamometer operation. Parameter coupling analysis and sensitivity testing are performed to enhance operating database of water brake dynamometer. The dynamometer runs the database and completes verification of the digital model. Subsequently, a performance prediction model for water brake dynamometer is built based on digital twin technology and experimental data, enabling an accurate mapping of the dynamometer's operational state. After turbine test, predicted operating parameters of the digital model show that the dynamic mean error between predicted values and actual values of multiple core component temperatures is less than 1%. Considering the sensitivity of data changes, these prediction error values are acceptable，which provide valuable reference information.

Presenting Author: Shuo Song Northwestern Polytechnical University

Presenting Author Biography: Shuo Song is a second-year graduate student in Aerospace Engineering at Northwestern Polytechnical University. Shuo grew up in China,and obtained a Bachelor's degree in Aircraft Power Engineering at Northwestern Polytechnical University. During his graduate school, he focuses on researching engine performance prediction and related fields.
While engaging in theoretical studies, he made multiple visits to experimental sites, interacting with engineering personnel and summarizing practical experiences. He participated in several engineering projects, refining performance prediction models based on experimental data. He is dedicated to conducting research in fields such as machine learning and digital twins.
After completing his Master's studies, Shuo is preparing to pursue a Ph.D. abroad, focusing on further researching performance prediction and health monitoring of engineering machinery. He is committed to exploring innovative approaches in related fields.

Authors:

Shuo Song Northwestern Polytechnical University
Hong Xiao Northwestern Polytechnical University
Leibo Jiang Northwestern Polytechnical University
Yufeng Liang Northwestern Polytechnical University