.png){kind=icon}
.png){kind=icon}
Return
Physics-Constrained Machine Learning Surrogate Model for Time-Dependent Behavior of Ceramic Matrix Composites
.png){kind=icon}
PRE
AI
DOI：10.1016/j.compositesb.2025.112825.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
A physics-informed recurrent neural network based surrogate model is developed to emulate the nonlinear, time-dependent constitutive behavior of ceramic matrix composites (CMCs) driven by microscale matrix damage and constituent creep. Physics-informed constraints are introduced into the surrogate model through regularization to ground the prediction in physics and improve its predictive capabilities. The model is trained using data generated by the high-fidelity generalized method of cells, employing appropriate constituent-level creep and damage models. The microscale repeating unit cell is loaded under creep fatigue conditions representative of turbine engine environments. The surrogate model predicts, as a function of variable input stress sequence, temperature, and microstructural features, the resulting strain history response while satisfying constraints related to creep rate, isochoric inelastic deformation, and strain energy density. The trained surrogate accurately matches the strain history over various input data with significant computational efficiency improvements. The benefits of employing physics-informed constraints, including the need for less training data and greater accuracy, are demonstrated. Finally, sensitivity analysis is conducted to highlight the effect of microstructural features on the strain history response. This work demonstrates the feasibility of developing, training, and executing surrogate models for CMCs with complex microstructures, nonlinear time-dependent material response, and under non-monotonic loading.
Keywords:
physics-informed neural network
ceramic matrix composites
surrogate model
microscale damage
creep fatigue
strain history prediction
Journal
IF:
14.2
Papers:
1.2W
Citations:
8.9W