Physics Informed Machine Learning Model for Local Creep Prediction in Turbine Blade

Abstract Gas turbine blades are an essential part of modern power generation systems, where they operate under high temperature, pressure and centrifugal load for prolonged period. Over time, these operating conditions lead to creep deformation, which can cause significant damage to the turbine. Conventionally, Creep assessment is carried out using Finite Element Method (FEM) which is a time-consuming process. Various material models namely strain hardening, time hardening, Norton Bailey Law, Garofalo Law etc. are used to describe the creep response of a given material, particularly metals and crystalline solids. The motivation behind this work is to reduce the time to run creep assessment. This paper focuses on exploring hybrid machine learning techniques to predict the local creep strain in turbine blades. This hybrid machine learning model can prove to be effective for quick sensitivity checks, decision making on service request and to study manufacturing related deviations in design. Multiple approaches were evaluated such as neural network, physics informed neural network, stress prediction using machine learning followed by creep prediction using physics-based formulation, ensemble neural network, generalized linear model (GLM) and time series method. Data driven approach significantly reduced the creep run time from few days to minutes and therefore provide flexibility in terms of analysing multiple models with different boundary conditions, material cases etc.