The stochastic nature of material damage evolution requires the development of both monitoring and modeling methods, capable of providing information related to evolving and multiscale material states. While several monitoring methods related to fracture have been proposed, recent needs for real-time assessment have created the need to connect microstructural scale data acquisition with machine learning and data-driven modeling. In this context, this talk presents a novel approach to leverage multi-physics nondestructive evaluation (NDE) datasets in a data processing framework capable of providing diagnostics and uncertainty quantification for fracture at the material scale. Data was generated using a double sharp notch specimen of an aluminum alloy in experiments conducted by coupling mechanical testing inside a scanning electron microscope with NDE techniques including Acoustic Emission monitoring and Digital Image Correlation. The main innovation of this approach is the fact that a combination of machine learning methods can be used to categorize the initiation of cracking and the subsequent crack growth. The optimal machine learning model is trained using Claude Shannon’s Information Entropy, while prediction of the next damage state is achieved through a trained Bayesian Inference model. This model is further sampled with the Markov Chain Monte Carlo Method to quantify the bounds of certainty. A discussion is also provided on how this approach can further be implemented in data-intensive, real-time applications.