Innovative solutions for process and quality control within the production system

The automotive industry is transitioning toward mass customisation and electrification, increasing variability in final assembly and the risk of non-conformities that affect costs, customer satisfaction, and environmental footprint. Quality assurance must therefore evolve from appraisal-oriented control to prevention, leveraging data and automation. Zero-Defect Manufacturing (ZDM) provides a unifying framework that uses Industry 4.0 technologies to detect, predict, prevent, and repair defects. This dissertation develops practical decision-support methods to operationalise ZDM in automotive final assembly, reducing defect propagation while balancing economic and sustainability objectives through three complementary research streams. First, inspection planning is improved through a cost-minimising mixed-integer linear programming (MILP) model that decides inspection placement and intensity and selects the best modality (human, automatic, or tablet-assisted) under takt-time constraints, detection performance, and learning/training investments. A Design Research Methodology study complements this with guidelines to strengthen human visual inspection. Second, the thesis supports the integration of Automatic Optical Inspection (AOI) from both investment and operations perspectives. A staged techno-economic assessment combines Multi-Criteria Decision-Making with a detailed cost model to compare image-acquisition configurations (fixed, cobot-mounted, handheld) and justify the investment. For hybrid human–robot inspection, a Fuzzy Inference System (FIS) quantifies task–agent suitability using quality criticality, ergonomics, and cognitive load, and a bi-objective MILP assigns and sequences inspection tasks by maximising suitability while minimising makespan. Third, defect prediction is made compatible with manual, highly variable work by extracting risk signals from operator behaviour. A wearable Human Activity Recognition (HAR) approach using wrist-worn inertial sensors identifies task execution and checks compliance with prescribed sequences; experiments demonstrate feasibility and highlight requirements for data acquisition, robustness, and integration into real-time routines. Overall, results show that sustained gains arise when AI and digitalisation augment human expertise, enabling hybrid quality systems that reduce scrap and rework and advance sustainability.