Abstract

In order to remain competitive, gas turbine OEMs must ensure high availability of their products. In addition to designing for product performance and product quality, OEMs provide long-term maintenance plans to their customers to keep their engines working smoothly. Such maintenance plans are made possible by the monitoring and advanced analytics of telemetry data coming from engine assets. For example, detection of problematic operating conditions and prediction of future part failures needing maintenance or overhaul requires collection of data and analytical or machine learning models to process that data. However, implementing and deploying such real-time models necessitates upstream simulation tools. The goal of this paper is to present 3D simulation technology that enables these predictive-maintenance use cases. We will discuss: a) Thermo-mechanical simulation of gas turbines. Here, Siemens Simcenter 3D Thermal-Multiphysics is used to build a complete representation of the whole engine through the use of thousands of separate boundary conditions that adequately represent the transient behavior of the whole engine. Those boundary conditions represent the flow conditions, in various regimes, for a range of rotational speeds and in different sections of the engine. They make use of built-in thermal correlations, of “expressions” that can automatically access specific model data (e.g., fluid temperature, material properties, solve-time computed results, rotational speed or radius), or of proprietary correlations obtained from in-house measurements, experience or CFD. b) Correlation of simulation data to test data. Here, we will discuss a novel adjoint-based sensitivity approach to rapidly compute sensitivities to many design parameters, such as heat-transfer coefficients, and heat-pickup rates, and utilize those sensitivities to rapidly correlate and update a transient thermal-mechanical model to temperature measurements. c) Model-order reduction. Here, we will discuss physics-based (intrusive) and AI-based (non-intrusive) methods to extract lower-dimensional approximations of the whole engine thermo-mechanical model that run in near-real time. The impact of model non-linearities will be emphasized. These reduced-order models can be deployed within edge devices or in the cloud to act as “virtual sensors” or “digital twins” that augment measured telemetry data. Development and deployment of these analytics tools involves several technology bricks and the exchange of large amounts of real-time data between them. As such, we will briefly touch upon the requirements for an industrial data pipeline to facilitate the connection between these bricks. We will demonstrate these topics by means of a real gas-turbine engine model.