Optimizing energy, downtime, and throughput in footwear production through machine learning

Area of Science:

• Industrial Engineering
• Manufacturing Technology
• Data Science

Background:

• Footwear manufacturing faces challenges in optimizing production efficiency and reducing waste.
• Machine learning offers potential solutions for enhancing predictive capabilities in industrial settings.

Purpose of the Study:

• To investigate the impact of machine learning model optimization on footwear production efficiency.
• To compare the performance of an optimized logistic regression model against other supervised learning algorithms.

Main Methods:

• Systematic refinement of a logistic regression model for defect prediction.
• Benchmarking the optimized model against Support Vector Machines, Naïve Bayes, and Multinomial classifiers.
• Evaluating model performance using metrics such as accuracy, specificity, sensitivity, F1-score, and balanced accuracy.

Main Results:

• Optimized logistic regression model achieved 97.06% predictive accuracy and 100% specificity.
• The model outperformed other algorithms in sensitivity, F1-score, and balanced accuracy.
• Industrial performance improvements included a 7.2% increase in production throughput, a 9% reduction in equipment downtime, and a 5.3% decrease in energy consumption.

Conclusions:

• Tuned machine learning classification models integrated with real-time analytics can significantly enhance footwear manufacturing.
• Data-driven optimization frameworks are crucial for improving productivity, energy efficiency, and sustainability in intelligent production environments.