Artificial intelligence based operational strategy development and implementation for vibration reduction of a supercritical steam turbine shaft bearing

Abstract

The vibrations of bearings holding the high-speed shaft of a steam turbine are critically controlled for the safe and reliable power generation at the power plants. In this paper, two artificial intelligence (AI) process models, i.e., artificial neural network (ANN) and support vector machine (SVM) based relative vibration modeling of a steam turbine shaft bearing of a 660 MW supercritical steam turbine system is presented. After extensive data processing and machine learning based visualization tests performed on the raw operational data, ANN and SVM models are trained, validated and compared by external validation tests. ANN has outperformed SVM in terms of better prediction capability and is, therefore, deployed for simulating the constructed operating scenarios. ANN process model is tested for the complete load range of power plant, i.e., from 353 MW to 662 MW and 4.07% reduction in the relative vibration of the bearing is predicted by the network. Further, various vibration reduction operating strategies are developed and tested on the validated and robust ANN process model. A selected operating strategy which has predicted a promising reduction in the relative vibration of bearing is selected. In order to confirm the effectiveness of the prediction of the ANN process model, the selected operating strategy is implemented on the actual operation of the power plant. The resulting reduction in the relative vibrations of the turbine’s bearing, which is less than the alarm limit, are confirmed. This cements the role of ANN process model to be used as an operational excellence tool resulting in vibration reduction of high-speed rotating equipment.