The Rotate Stress of Steam Turbine prediction method based on Stacking Ensemble Learning

The rotor is one of the most durable parts for a steam turbine. The main shaft, blades or impellers will generate huge transient stress due to high rotate speed, strong torque, and high-temperature steam in the starting up and stopping and other processes, thus, the rotor will face the acid test. As a result, the analysis and prediction of the transient stress of the rotor become the key point in steam turbine life management. The steam turbine is always in high-speed rotation under different poor working conditions, and it is difficult to measure the transient stress directly. In actual production, the finite element analysis is the main method to obtain the stress distribution, lacking measurement accuracy and timeliness. In this paper, we proposed a novel calculation method based on data-driven. Firstly, building a single prediction model based on the Lasso regression, the Elastic net, and the Random forest. Secondly, utilize an ensemble learning based on the stacking to combines a single prediction model to acquire better prediction accuracy. Finally, the case result indicates the effectiveness of the proposed method.