SmartMan - Specific use of digitalization, AI and competence development in vehicle supplier production and quality management
Motivation
The transformation towards autonomous, electric driving poses significant challenges to the production chains of the German automotive industry. The demand for high-performance electronic components is growing rapidly. In addition to quality and customer satisfaction, output and margins must be optimized. Domestic supply chains face global competition. Electronic modules are progressively assembled using advanced packaging and interconnection techniques. Production errors often become critical only at later stages, propagating to the system level and significantly affecting quality and lifespan. For a more sustainable production with minimal waste, functional safety, and extended module lifetimes, robust manufacturing processes are indispensable.
Objective
SmartMan aims to achieve a digitalized, intelligent, and sustainable transformation of semiconductor package production to enable a zero-defect approach and resource-efficient electronics manufacturing. This goal will be accomplished through the use of a new generation of inline-capable error analysis modules that are sensitive to defects arising during production. This quality control system is embedded in an intelligent, big-data-supported production environment and control system, utilizing AI methods to provide real-time feedback and corrective decisions directly to the production line. The improved process robustness, reliability, and performance of current components and future developments ensure a significant competitive advantage for German electronics manufacturing. It allows an unprecedented level of process control capability and prepares the industry for the increasingly fierce competition in the multi-billion-dollar market for microelectronic, power, and optoelectronic semiconductor systems in the automotive industry.
Approach
Figure: Research focus areas of the iwb within the SmartMan project
In the initial phase, production processes are specified and validated by defining demonstrators, identifying relevant process parameters and error analysis methods, and creating digital twins. Inline-capable error analysis systems are then developed for various production stages, supplemented by AI-based automated evaluation, and integrated into the pilot line. The digital production line is optimized through real-time defect classification and flexible production planning. A Reinforcement Learning-based AI agent is trained using a simulation (digital twin) to enhance production control and optimize dispatching decisions. The methodology is tested on a production line and securely integrated into existing systems. In parallel, the impact of defects on process robustness and product lifespan is analyzed by introducing realistic pre-damages, conducting accelerated stress tests, and developing finite element models. These models enable a comprehensive assessment of process robustness and lifespan. Finally, the digitalization strategy is validated using industrial demonstrators. The methodology is evaluated under realistic conditions, including the investigation of defect propagation, production yield analysis, and sustainability. The results contribute to the derivation of process guidelines and support the implementation of the intended zero-defect strategy and resource-efficient electronics production.