ANALYSIS OF DETECTION SYSTEM FOR COVER TAPE OFFSET IN THE TAP AND REEL PROCESS USING NEURAL NET TIME SERIES METHOD

This technical report presents a comprehensive study on the detection of cover tape offset or misalignment during the tape and reel process, which is crucial for packaging electronic components into individual pockets of carrier tape. The research aims to develop an efficient system utilizing the Raspberry Pi Camera Module for detecting and analyzing cover tape misalignment. The methodology involves integrating the Raspberry Pi Camera Module with a microcontroller to capture and process images of the carrier tape, employing image processing techniques for misalignment detection. The resulting data is displayed in a user-friendly dashboard format using Node-RED. Additionally, the data is analyzed in MATLAB Neural Net Time Series for predictive analysis. The findings of this research, including the analysis of training results, demonstrate the successful implementation of a reliable cover tape misalignment detection system. Notably, the Bayesian Regularization (BR) training algorithm outperformed the Scaled Conjugate Gradient (SCG) training algorithm for cover tape offset's predictive analysis, exhibiting lower Mean Squared Error (MSE) with 0.0015874 for BR compared to 0.0017839 for SCG, consistently lower Mean Absolute Error (MAE) values, stronger linear correlations, and superior overall performance. It emphasizes its effectiveness for accurate predictions.