EUBCE 2026 - Ines PÉREZ COUÑAGO - Multimodal and Transparent AI for Adaptive Process Optimization in Sustainable Manufacturing

Multimodal and Transparent AI for Adaptive Process Optimization in Sustainable Manufacturing

Short Introductive summary

This work presents a multimodal artificial intelligence framework for quality control in bio-based composite manufacturing, addressing high process variability and limited data availability. The proposed system integrates vision-based inspection, using deep learning models (Mask R-CNN and U-Net), with process-parameter prediction through a Random Forest model. Additionally, explainable AI techniques are employed to identify the root causes of defects and generate actionable process adjustments. The approach incorporates synthetic data generation and cross-validation to enhance robustness and reliability. Results demonstrate accurate defect detection and up to 87% prediction accuracy, enabling practical, constraint-aware process optimization. Overall, the system provides a scalable and industry-ready solution that not only detects defects but also supports real-time decision-making and process improvement, contributing to more efficient and sustainable manufacturing.

I. Pérez Couñago, AIMEN Technology Center, O Porriño, SPAIN
M. Lago López, AIMEN Technology Center, O Porriño, SPAIN
I. Gil Cortés, AIMEN Technology Center, O Porriño, SPAIN
M. Morandeira Domínguez, AIMEN Technology Center, O Porriño, SPAIN
S. Muíños Landín, AIMEN Technology Center, O Porriño, SPAIN
A. Marqués Paola, AITIIP Centro Tecnológico, Zaragoza, SPAIN
J. Vidal Navarro, Moses Productos, Zaragoza, SPAIN