Global manufacturing is entering an era of unprecedented variability in material properties, driven by sustainability goals and market volatility. The adoption of recycled steels introduces supplier-specific differences, while cost-reduction strategies and geopolitical disruptions (tariffs, trade barriers, and resource shortages) further amplify process scatter. These dynamics challenge conventional quality control in automotive sheet metal forming and demand intelligent, adaptable production systems.

This dissertation addresses how data-driven methods can strengthen process robustness without major infrastructure changes. Three guiding hypotheses are explored: (H1) machine learning can provide insights into the impact of input variations; (H2) synthetic data can supplement or replace operational data for model development; and (H3) existing sensor signals can be reinterpreted to reduce reliance on additional instrumentation.

A hybrid methodology combining finite element simulations, stochastic modeling, and industrial press shop data was developed. Key contributions include: (1) generation of synthetic datasets for predictive modeling of draw-in and cushion force; (2) application of unsupervised learning for early detection of anomalous material batches; and (3) a novel process monitoring metric, process work, derived from existing sensors to monitor process health.

The findings provide a framework for integrating intelligent data-driven tools into legacy systems, supporting the transition toward resilient and sustainable manufacturing practices.