PI: Dr. Ahmed Azab
CoI: Dr. Moayad AlNammi

This proposal aims to revolutionize hybrid manufacturing (HM) process planning through the convergence of Smart Manufacturing, Digital Twinning (DT), and Generative AI, forming the core of an advanced Industry 4.0 (I4.0) framework addressing both additive and subtractive manufacturing, as well as the delicate interactions between the two. At the heart of this vision is a novel Generative Pretrained Transformer (GPT)-based Large Multimodal Model (LMM), developed for macro-level Computer-Aided Process Planning (macro-CAPP). The model integrates part geometry and process logic by encoding CAD data into token embeddings and decoding them into sequenced manufacturing operations through an autoregressive Plan Decoder. This architecture supports end-to-end planning, from feature recognition to process sequencing, enabling real-time, adaptive planning for HM systems combining metal additive and subtractive processes.

The LMM-GPT framework is supported by a hybrid methodology that fuses Operations Research, interpretable Machine Learning (ML), and Logical Analysis of Data (LAD). This hybrid engine generates pretraining data without relying on commercial datasets, allowing robust, domain-specific customization. The approach applies constrained clustering, rule mining, and decomposition methods to iteratively learn and improve process plans, effectively bridging predictive analytics and symbolic reasoning.

In parallel, Digital Twin technology plays a critical role across both macro- and micro-CAPP layers. At the micro-level, a generative AI-driven DT framework enables real-time monitoring and adaptation of process parameters, forming a self-healing loop based on live process signature feedback. This is a key enabler of Quality 4.0—minimizing defects and a step towards asset health management at the process-level.