Intelligent Prediction of TFT-LCD's Quality Based on BPNN

Thin Film Transistor Liquid Crystal Display (TFT-LCD) is a leading flat-panel technology known for its high resolution, energy efficiency, and slim design, widely integrated into monitors, TVs, and smartphones. Current research predominantly focuses on defect detection in TFT-LCD images, neglecting the relationship between production equipment parameters and product quality scores. This study introduces a reverse neural network model with L1 norm feature selection (L1-BP), using equipment parameters as inputs and quality scores as outputs. L1 regularization identifies pivotal features, reducing noise impact for enhanced prediction accuracy and model interpretability. Moreover, the reverse neural network effectively captures intricate nonlinear patterns in TFT-LCD quality data, further improving prediction precision. Experimental findings demonstrate superior performance over prior research in MSE values, showcasing stable model efficacy and presenting a novel avenue for optimizing quality in TFT-LCD manufacturing.