Digitalization of Test Rigs for Model Acceptance Testing of Hydraulic Machines: Gaussian-Process-Based Digital Twin of a Turbine on a Hydraulic Test Rig for Efficiency Analysis and Test Automation

Abstract

Systematic measuring of hydraulic machine's performance data at hydraulic model test rigs involves comprehensive expert knowledge of the test rig engineer and is time-consuming, as the multidimensional operation space must be explored manually. Expertise is needed to account for specific performance limits of test-rig equipment, model turbines, storage pumps, or pump turbines as well as some hysteresis of the setup due to flow separation and dealing with operation point-dependent gains between controlled parameters and measured quantities. To free valuable manpower from the burden of shift work, the measurement procedure is subject to automation. However, testing new automation concepts involves safety-critical or expensive tests and consumes valuable time at tests too. To develop and validate new automation concepts offline, we present a hybrid data-driven modeling approach to create a digital twin of the combined setup of a hydraulic turbine on a hydraulic test rig. Special attention paid to the modeling approach is consistent with existing mathematical relations governed by either physical models or norms. The presented approach allows the separation of test rig and hydraulic model machine under test and this enables offline validation of automation concepts on different combinations of test rigs and hydraulic model machines. The modeling approach is demonstrated by performing cross-validation using recorded measurement data.