Abstract: In order to estimate both the interface toughness and separation strength between micron-thick copper film and Al2O3 substrate, peel test measurements with film thicknesses 1―15 microns have been performed and the peel forces with the curvature radiuses at crack tip have been measured. Furthermore, a finite element model is presented to simulate the peeling process. In our simulations the interface between film and substrate is modeled by Cohesive Zone Model (CZM) and the constitutive relation of copper film is modeled by strain gradient plasticity (SGP) theory to characterize the size effect. Based on the experiments and simulations, an inverse analysis based on neural network method is adopted to determine the interfacial parameters. The inputs of the neural network are peel force and curvature radius at crack tip, the outputs are interface toughness and separation strength. The neutral network is trained by 66 simulated data, and the trained network can predict the interface toughness (0.045N/mm) and separation strength (26.6MPa) from measured data.