Chemical vapor deposited (CVD) ZnS, widely used in infrared windows and conformal radomes for high-speed aircraft, requires accurate characterization of its plastic constitutive behavior to ensure service reliability. However, the complex deformation mechanisms of CVD-ZnS at the micro-/nanoscale make it challenging to achieve high-precision parameter identification using conventional approaches. In this study, a novel multiobjective parameter inversion method based on a metaheuristic-tuned neural network is proposed to establish a reliable framework for characterizing the micro-nanoscale mechanical behavior of CVD-ZnS. An automated Python-driven finite element simulation was developed to construct a large-scale nanoindentation data set. A high-accuracy surrogate model was then built by integrating the Crested Porcupine Optimizer (CPO) with a back-propagation neural network (BPNN), and its superiority was validated against several other metaheuristic algorithms. The surrogate model was embedded into the NSGA-II multiobjective optimization framework, and entropy-weighting analysis was employed for comprehensive evaluation, leading to the identification of the optimal Johnson-Cook (J-C) constitutive parameters of CVD-ZnS. The results demonstrated that the CPO-BPNN improved prediction accuracy by approximately 40-56% compared with the traditional BPNN. The final inverted J-C parameters (A = 322.8 MPa, B = 1014.1 MPa, n = 0.32) effectively captured the mechanical responses in both nanoindentation and nanoscratch experiments. Furthermore, nanoscratch experiments revealed the deformation mechanisms of CVD-ZnS and identified an equivalent residual groove depth at the brittle-to-ductile transition (DBT) point of 133.96 ± 4.64 nm, which serves as a geometric indicator of the transition region. Overall, the proposed framework provides a reliable and integrated route for precise inversion and characterization of the plastic constitutive behavior of CVD-ZnS at the micro-/nanoscale.
Yao et al. (Tue,) studied this question.