I-Corps: Translation potential of a neurosymbolic intelligence platform for root cause identification and corrective action in semiconductor manufacturing

Award ID: 2634190

This I-Corps project is based on the development of a platform that identifies why manufacturing defects occur and determines how to correct them. Manufacturers across semiconductor, electronics, automotive, aerospace, and defense supply chains face costly production delays, material waste, reduced yield, and recurring quality problems because current diagnostic tools often detect defects without clearly explaining their causes. This technology combines production knowledge with data-driven analysis to deliver rapid, traceable diagnoses and recommended corrective actions, while preserving expertise that might otherwise be lost when experienced personnel leave an organization. Users may include electronics manufacturing service providers, semiconductor packaging and test companies, fabrication facilities, equipment suppliers, and manufacturers in other quality-critical sectors. This software platform may reduce investigation time, improve yield, accelerate new-product introduction, and support more consistent engineering decisions across different production environments. This I-Corps project utilizes experiential learning coupled with first-hand investigation of the industry ecosystem to assess the translation potential of a neurosymbolic artificial intelligence platform for explainable root-cause analysis and zero-defect manufacturing. This technology integrates formal ontologies and knowledge graphs with multimodal deep-learning models that analyze inspection images, time-series sensor signals, process parameters, equipment conditions, materials, and test outcomes. The hybrid reasoning framework combines the pattern-recognition capabilities of convolutional, recurrent, and graph neural networks with explicit representations of failure modes, causal relationships, and manufacturing constraints. Unlike black-box machine-learning systems or rule-based tools, the platform is designed to provide both predictive accuracy and traceable explanations connecting observed defects to probable causes and corrective actions. Previous research demonstrated methods for representing manufacturing processes and failure mechanisms in knowledge graphs and for coordinating multiple neural models with symbolic reasoning. Users may benefit from faster diagnosis, improved yield, retained organizational knowledge, and greater confidence in manufacturing decisions. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Program Director: Ruth Shuman