# Machine Learning Approaches in Agile Manufacturing with Recycled Materials for Sustainability

**Authors:** Aparna S. Varde, Jianyu Liang  
**arXiv:** [2303.08291](https://arxiv.org/abs/2303.08291)  
**Date:** 2023-03-15  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.08291v1/

## Key points

- Focus on intersection of AI and Materials Science for sustainability and environmental conservation
- Develop domain-specific tools using machine learning for decision support in agile manufacturing processes with recycled materials
- Utilize cutting-edge techniques like Artificial Neural Networks (ANN), Random Forests (RF), and Convolutional Neural Networks (CNN) for green technology in smart manufacturing
- Achieve impressive accuracy rates around 90% with ANN models, high efficiency, and cost-effectiveness
- Explore advanced computer vision models like VGG and ResNet to enhance predictive analytics accuracy, efficiency, and robustness
- Bridge the gap between AI and Materials Science to champion environmental sustainability through innovative technological solutions
- Repurpose waste materials into valuable products for sustainable manufacturing practices
- Assess impacts on sustainability through appropriate metrics considering use of recycled materials

## Summary

In this research, we focus on the intersection of AI and Materials Science to promote sustainability and contribute towards environmental conservation. Our goal is to develop domain-specific tools using machine learning for decision support in agile manufacturing processes that involve recycled materials. By leveraging cutting-edge techniques like Artificial Neural Networks (ANN), Random Forests (RF), and Convolutional Neural Networks (CNN), we aim to make significant strides towards green technology in smart manufacturing. Building upon our previous work on AutoDomainMine for computational estimation, we have delved deeper into the realm of agile sustainable manufacturing by utilizing recycled and waste materials in industrial processes. Our approach involves predictive analysis based on real data from lab experiments using machine learning models such as ANN, RF, and CNN. The results obtained from our ANN models showcase an impressive accuracy rate of around 90%, coupled with high efficiency and cost-effectiveness. Similarly, our CNN and RF models exhibit exceptional accuracy levels, enabling us to conduct in-depth analyses at a finer granularity level while ensuring comprehensibility for explainable AI. Moving forward, we aim to explore advanced computer vision models like VGG and ResNet to enhance accuracy, efficiency, and robustness in our predictive analytics. Through our research endeavors, we strive to bridge the gap between AI and Materials Science while championing environmental sustainability through innovative technological solutions. Drawing inspiration from successful AI applications across various scientific domains, we are committed to pioneering sustainable manufacturing practices that not only benefit the environment but also contribute positively towards defense operations by repurposing waste materials into valuable products. Our ultimate goal is to assess the impacts on sustainability through appropriate metrics that consider the use of recycled materials and the effectiveness of the resulting products. By integrating cutting-edge machine learning techniques with traditional manufacturing processes, we are paving the way for a greener future where recycled materials play a pivotal role in driving innovation and progress in industrial sectors.