Study On Multi-Dimension Optical Sensing Control Technology For Hot-Embossing Micro-Forming Of Composite Microlens Array

The two-dimensional structure in which the microlens array is compounded on the surface of the microprism array has comprehensive micro-optical properties such as light condensing,anti-reflection and diffuse reflection.In the process of its rapid hot-embossing,there are several problems such as temperature,pressure and other sensor accuracy drift,equipment mechanical movement error,etc.And the micro-forming accuracy depends on the field experience to adjust the process variables,resulting in extremely low processing efficiency.Therefore,according to the interference differences of the two micro dimensions on the refraction and reflection of light in different wavelength bands,an online optical-sensing detection method for the forming dimensions of the composite micro-lenses array was developed.Besides,offline and online detection were combined to build a correlation database related to optiacal sensing,process variables and two forming dimensions.Then BP,RBF neural network and naive Bayes method were employed to establish the online detection model of micro-forming heights and the selection of hot-embossing process variables model,aiming to solve the nonlinear relationship between two-dimensional micro-forming heights and optical sensing illumination as well as that between two-dimensional micro-forming heights and process variables,and realize the hotembossing precision forming of composite micro-lens arrays.First,an online detection method for hot-embossing micro-forming heights was developed and the illuminantion during the micro-forming process was characterized to correlate the twodimensional forming heights.And a simulation and an experimental platform on the hotembossing machine were built for the online detection of hot-embossing micro-forming heights,achieving the data communication among the hot-embossing machine,the detection device and computer software and hardware.An analysis method of optical sensing illuminantion gradient was proposed.The results show that the green light with a wavelength of about 550 nm is the most sensitive to the hot-embossing micro-forming dimensions,and the sensitivity of the small micro lens array is 30.5% higher than that of the micro prism array.Then,due to the normal distribution of micro-forming heights,the naive Bayes model detected microprism and microlens heights with the detection errors of 6.71% and 6.57%,respectively,which were lower than the BP and RBF neural network models.In addition,the microprism and microlens forming errors of the process variables selected by the naive Bayes model using the combined selection strategy were also the lowest,which were 1.39% and0.82%,respectively.Therefore,the naive Bayes model can realize fast and precise online detection and process variables selection on small-scale data.Finally,based on the forming errors detected by the micro-forming heights online detection model,the proportional link feedback control method was employed to establish a micro-forming accuracy feedback control system to correct the target forming heights,simultaneously,the hot-embossing process variables selection model was used to adjust the multiple process variables corresponding to the two-dimensional forming heights.The experimental results show that the forming heights of microprism and microlens array can be controlled at 55 ± 2 μm and 8 ± 1 μm,which can realize fast,precise and stable forming of composite microlens array.