Robustness optimization of gas turbine performance evaluation against sensor failures

Gas turbines operate in harsh environments for long periods of time and their performance will inevitably degrade. Real-time evaluation of gas turbine performance is of great importance for both safety and economy. Considering that gas turbine sensors often fail in harsh environments, in order to improve the reliability of gas turbines against sensor failures, a model for optimizing the robustness of gas turbine performance evaluation against sensor failures is proposed. This model combines just-in-time and ensemble learning algorithms based on deep neural networks. By building local models and then ensemble learning, the influence of parameter changes on the global model is reduced. In this paper, taking gas turbine efficiency as an example, the robustness optimization effect of different models is tested by several robustness evaluation methods. It is found that the proposed model can better optimize the robustness of the evaluation, with the highest accuracy and best fit under various disturbances.