Cigarette weight detection is crucial for the quality of cigarette production, requiring regular calibration of key parameters to maintain accurate weight control. Traditional methods for adjusting these parameters are often time-consuming, wasteful, and fail to provide timely updates, which can affect the accuracy of real-time weight assessment. This study introduces an innovative online optimization system designed to refine weight detection parameters. The system incorporates an automated sampling process and utilizes the XGBoost regression model for prediction. These features streamline the optimization, improving both automation and the accuracy of online weight measurements. Results show that the system decreases the time needed for parameter adjustment by 65.72% and reduces cigarette consumption by 83.33%. Moreover, it increases the compliance rate with weight standards during quality checks by 41%. Thus, this advanced diagnostic and optimization approach significantly enhances the precision of real-time weight monitoring, contributing to higher product quality and more efficient manufacturing.