Tool condition monitoring is critical in ultra-precision manufacturing in order to optimize the performance of the overall process, while maintaining the desired part quality. Recently, Deep Learning has been successfully applied in numerous classification tasks in manufacturing, often to forecast part quality. In this paper, a novel Deep Learning data-driven modeling framework is presented, which includes fusion of multiple stacked sparse autoencoders for tool condition monitoring in ultra-precision machining. The proposed computational framework consists of two main structures. A training model that is designed with the ability to process multiple parallel feature spaces to learn the lower-level features; and a feature fusion structure that is used to learn the higher-level features and associations to tool wear. To achieve this learning structure, a modified loss function is utilized that enhances the feature extraction and classification tasks. A dataset from a real manufacturing process is used to demonstrate the performance of the proposed framework. Experimental results and simulations show that the proposed method successfully classifies the ultra-precision machining case study with over 96% accuracy, while also outperforms comparable methodologies.

Tool wear prediction is of critical importance to maintain the desired part quality and improve productivity. Inspired by the successful application of deep learning in many condition monitoring tasks. A novel modeling framework is presented, which includes multiple stacked sparse autoencoders and a nonlinear regression function for tool wear prediction. Multiple stacked sparse autoencoders consists of two main structures: one model is designed with multi-dimensional stacked sparse autoencoders, which can learn more features from different feature-domains in the raw vibration signal; and another single-dimensional stacked sparse autoencoders is used for feature fusion and deeper features learning. And a modified loss function is applied that improves the learning ability. In addition, due to the good properties of tool wear process in nonstationarity and complex nonlinear, a nonlinear regression function is utilized to enhance the progressive tool wear prediction tasks. A dataset from a real manufacturing process is used to evaluate the performance of the proposed modeling framework. Experimental results show that tool wear can be predicted accurately and stably by the proposed tool wear predictive model, which outperforms the already developed methods.

The light from the side surfaces of the test flip-chip light-emitting diode (FCLED) dies is reflected, refracted or absorbed by neighbour dies during the measurement of wafer-level FCLED dies in production lines. A notable measurement deviation is caused by the neighbour-die effect, which is not considered in current industry practice. In this paper, Monte Carlo ray-tracing simulations are used to study the measurement deviations caused by the neighbour-die effect and extension ratios of the film. The simulation results show that the maximal deviation of radiant flux impinging the photodiode can reach 5.5%, if the die is tested without any neighbour dies, or is surrounded by a set of neighbour dies at an extension ratio of 1.1. Moreover, the dependence between the measurement results and neighbour cases for different extension ratios is also investigated. Then, a modified calibration method is proposed and studied. The proposed technique can be used to improve the calibration and measurement accuracy of the test equipment used for measurement of wafer-level FCLED dies in production lines.