Real-Time Analysis of Screwdriving Defect Trends

This repository contains the project developed during my academic internship, focused on real-time defect trend analysis in semi-automatic screwdriving systems. The analysis includes data extraction, statistical process control, and inferential modeling to monitor and reduce errors in the production process.

Project Structure

Documentation

• Presentazione_Leonardo_Livi.pdf: The presentation given during the academic defense of the project.

Scripts

• DF_BUILD.py: This script processes raw data to create a structured DataFrame for analysis.
• MAIN_COND_STD.py: Aggregates and processes data to prepare it for statistical analysis.
• Derive_procedure_STD.py: Calculates control limits used in PowerBI visualizations to monitor out-of-control processes.
• Poisson_model_errors.R: An R script that fits a Poisson regression model to estimate error rates in the screwdriving process.

Dashboard

• xBar_Chart.pbix: A PowerBI file that visualizes the control charts for monitoring screwdriving operations in real time.

How to Run the Project

1. Data Preparation: Run DF_BUILD.py to create a structured DataFrame from raw data files.
2. Data Aggregation: Use MAIN_COND_STD.py to process and aggregate the data for analysis.
3. Control Limit Estimation: Execute Derive_procedure_STD.py to calculate control limits used in PowerBI.
4. Statistical Modeling: Use Poisson_model_errors.R to fit a Poisson regression model and estimate the error rates.
5. Visualization: Open xBar_Chart.pbix in PowerBI to view the real-time control charts.

Results

The analysis indicated that the hurdle model was the most appropriate for predicting screwdriving defects, improving process quality and providing actionable insights for real-time monitoring.

Contact

For any questions, feel free to reach out to me at: leonardo.livi1@edu.unifi.it.