In-process detection of hot-embossed two-level microstructures by the green light sensing

The microlens distributed on microprism surface is applied in high value-added optical products. During hot embossing, it, however, is difficult to detect the microprism and microlens heights on macro surface. Hence, an optical sensing is proposed according to energy consumption as light passing through the hot-embossed two level microstructures. The objective is to in-process recognize the two-level micro-formed sizes. In order to depict the nonlinearity between optical-sensing and micro-forming heights, the Naive Bayes method (NB) and artificial neural network (ANN) were employed to distinguish the two-level heights based on the processed experiments. It is shown that the green light responds to micro-forming heights best by illumination gradients. Theoretically and experimentally, the denser microlens array is more sensitive than the microprism array. By rapidly updating the database with illumination, process variables and micro-forming heights in the small volume, the NB recognizes two-level microstructure sizes more accurately than the ANNs for efficient production.

Automated summary

This study proposes an optical sensing method for real-time recognition of the size of two-level microstructures during hot embossing. Experimental comparison shows that green light provides the best response to micro-forming heights, and the Naive Bayes method can accurately identify microstructure sizes.