Material-Oriented Regularization Toward Solving Manufacturing Inverse Problem in Ion Beam Microprocessing

Abstract

Abstract A material-oriented regularization (MOR) methodology is developed to solve manufacturing inverse problem of estimating the manufacture input process parameters for a required output performance, demonstrated by ion beam microprocessing of tungsten components in future fusion reactors. The MOR methodology is explored as following steps: forward problem modeling, identification of characteristic material loading, and solving the inverse problem via the characteristic material loading. A thermodynamic model is established in forward problem scheme by comprehensively incorporating material constraints of tungsten, to simulate the output of residual surface stresses in top layer of several μm that determines fatigue performance of the microprocessed tungsten component. With the experimentally verified model, all material loading variables, i.e., thermal, elastic strain, and plastic strain energies can be explicitly described under the processing load of thermal energy input. Among the material loading variables, stored elastic strain energy is identified as characteristic material loading with a highest sensitivity in correlation to residual surface stresses, as process signature. The processing load of 2.1–4.2 J/cm2 is derived for a required residual surface stress in range of 0–1500 MPa within 15 μm depth, with an upper bound of the relative error of 4.7–11.7% for the inverse problem solution. The MOR enables comprehensive incorporation of material constraints with a self-convergence effect to effectively relax the ill-posedness of manufacturing inverse problems, otherwise in conventional regularizations such constraints have to be empirically adjusted in compromise with data fitting.