This talk will discuss the development of a machine learning-enabled, multiscale-multiphysics computational platform for multifunctional piezocomposites. It will constitute a digital twin for detecting location-specific damage evolution from surface electric field sensors. The integrated platform incorporates modules involving bottom-up and top-down multiscale modeling coupled with various machine learning operations. The first module involves development of a finite deformation parametrically upscaled coupled constitutive-damage models (PUCCDM) for structural-scale electromechanical response, by hierarchical modeling of microstructures undergoing progressive damage. The PUCCDM incorporates microstructural morphology in its coefficients in the form of representative aggregated microstructural parameters (RAMPs), determined using machine learning on data generated by micromechanical analysis. The micromechanical model consists of a coupled electromechanical finite deformation phase field model for crack initiation and propagation in nonuniform piezocomposite microstructures. Finally, coupled convolutional neural and long-short term memory networks (ConvLSTM) are deployed to predict current damage for correlating sensor-based electric signals to subsurface damage indicators.