Prediction of High-Speed Hydraulic Dynamometer Safety Envelope Base on Deep Learning Neural Network

High-speed hydraulic dynamometer is widely used for turbine component experiment in aircraft engine area. It usually works through measuring main performance parameters of turbine for verifying design method by absorbing huge shaft work which is passed from turbine. However, its water flow can generate strong turbulence phenomena between impellers in dynamometer’s case. In addition, high water temperature may generate cavitation phenomenon and cause high-frequency pressure pulsation. Both can lead to dynamometer’s performance degradation. Artificial experience diagnosis, though not suggested, is used commonly to prevent high-speed hydraulic dynamometer from working unsteadily. This method depends on the experience of workers, which may cause fuzzy definition and lead to safety hazard. In this paper, we propose a two-stage model of highspeed hydraulic dynamometer based on deep learning neural network. It utilizes the ideas of Transformer Model, which makes our model become more sensitive and stable. Accuracy and stability are proved by verifying actual device operation data. Based on two-stage model, we can draw work safety envelope by predicting performance parameters that delimit the safety boundary. Follow this guide workers can be able to make operation safer and stabler.