A novel wasserstein autoencoder-enhanced thermo-mechanical coupled reduced-order model for high pressure turbine blades life monitoring

To facilitate effective assessment of loads (e.g., temperature and stress) and life management of high-pressure turbine (HPT) blades, a wasserstein autoencoder (WAE)-enhanced thermodynamically coupled reduced-order model (ROM) is proposed in this paper. The advanced ROM for the nonlinear thermomechanical coupling fields is developed by introducing the deep learning model of WAE in the proper orthogonal decomposition (POD) method. The proposed method improves the prediction accuracy of loads in locally focused regions and generalization performance. The accuracy and efficiency of this method are validated through 30 sets of validation conditions. Results indicate that the proposed approach achieves higher accuracy and better generalization performance than traditional POD-based methods, with errors maintained within 10. Additionally, computational speed is improved by nearly 1400 times compared to conventional numerical methods. The WAE-enhanced ROM is applied for load and life assessment of the HPT blades throughout their service life. The evaluation time for a single aeroengine performance parameter is 1.7 s, and for a single flight evaluation, it is 67 s, which highlights the effectiveness of the proposed method in enabling the assessment of the loads and remaining life of HPT blades.