Physical Reservoir Computing for Smart Manufacturing of PCB Using Crystallinity‐Controlled HfO 2 Memristor for Reservoir

The demand for smart manufacturing in PCB fabrication has increased since the structure has evolved to multilayer stacking, thus anomaly detection in each layer becomes a crucial factor in the manufacturing process. Unlike CNN‐based methods, which require a high‐performance computing unit, reservoir computing (RC) models are emerging methods for defect detection. This work realizes physical RC via HfO 2 ‐based memristors by modulating the crystallinity of HfO 2 , which can minimize the ferroelectric effect while maintaining a high‐k insulating state. The electrical characteristics and short‐term memory of memristors have been evaluated, and applied to RC for PCB defect analysis, and achieved over 85% accuracy for all types of defects while using a simple ridge readout unit. The physical RC model can be applied as an edge device for each manufacturing process for image recognition, time‐series analysis with low‐power consumption, and two, which can realize autonomous manufacturing for various industries.