Damage simulations play a critical role in assessing structural integrity and understanding material failure under various conditions. However, their computational cost often poses a significant challenge, especially in scenarios requiring repeated simulations such as design optimization, uncertainty quantification, and real-time decision-making. Traditional model reduction techniques such as the proper orthogonal decomposition and especially hyper-reduction methods [1] can reduce the simulation time significantly but often necessitate intrusive modifications to existing simulation codes and are bound to the classical finite element framework. Especially in the context of digital twins, where real-time monitoring and predictive capabilities are essential, the need for rapid predictions of damage evolution based on sensor measurements is essential. In this contribution, an autoencoder-based non-intrusive model reduction framework for gradient-extended damage simulations [2] is introduced. The autoencoder is trained to capture complex, nonlinear relationships within the data, enabling significant reductions in computation time while maintaining high accuracy [3]. By creating a surrogate model that directly approximates the latent space representation of the autoencoder, it is possible to incorporate real-world data directly into the reduced order model. The framework is validated using real pressure sensor data and it is shown that a good agreement with a reference FEM solution is achieved. Regarding computational efficiency, the autoencoder-based non-intrusive model reduction framework demonstrates a significant speedup compared to traditional finite element simulations, leading to real-time predictions of the structural response. [1] C. Farhat, P. Avery, T. Chapman and J. Cortial, Dimensional reduction of nonlinear finite element dynamic models with finite rotations and energy-based mesh sampling and weighting for computational efficiency. Int. J. Numer. Methods Eng. (2014) 98(9): 625 - 662. [2] T. Brepols, S. Wulfinghoff and S. Reese, A gradient-extended two-surface damage-plasticity model for large deformations. Int. J. Plast. (2020) 129: 102635. [3] T. Simpson, N. Dervilis and E. Chatzi, Machine learning approach to model order reduction of nonlinear systems via autoencoder and LSTM networks. J. Eng. Mech. (2021) 147(10): 04021061.