A Reinforcement Learning-Based Method and System for Adjusting the PVA Absorption Roller Wafer Drying Process

By combining reinforcement learning and real-time perception from multiple sensors with deep learning and expert knowledge graphs to optimize the PVA water-absorbing roller wafer drying system, the problem of existing systems being unable to adjust and adapt to the dynamic characteristics of PVA water-absorbing rollers in real time has been solved, achieving efficient and stable wafer drying results and low operation and maintenance costs.

CN122129865APending Publication Date: 2026-06-02SHANDONG SENGONG NEW MATERIAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHANDONG SENGONG NEW MATERIAL TECH CO LTD
Filing Date
2026-03-03
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing PVA absorbent roller wafer drying systems cannot detect dynamic interference during the drying process in real time, and cannot optimize parameters, resulting in incomplete drying or scratches on the wafer surface. Furthermore, they fail to consider the dynamic characteristics of the PVA absorbent roller throughout its entire life cycle, leading to a decrease in absorbency and a continuous decline in drying effect.

Method used

A PVA water-absorbing roller wafer drying process adjustment system based on reinforcement learning is adopted. The system uses multi-source cross-modal sensors to perceive the status of the roller, wafer and environment in real time. It combines deep reinforcement learning and expert knowledge graph to optimize parameters and achieve global optimal adjustment. It also uses digital twin technology for predictive compensation.

Benefits of technology

It achieves stability and consistency in wafer drying, reduces moisture residue and scratch defect rates, adapts to different production priority requirements, improves system anti-interference capability and production continuity, and reduces operation and maintenance costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122129865A_ABST
    Figure CN122129865A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of semiconductor manufacturing technology and discloses a method and system for adjusting the PVA water absorption roller wafer drying process based on reinforcement learning. The sensor fusion module builds a full-dimensional perception system, accurately capturing four-dimensional state data of the PVA roller, wafer, environment, and actuator through multiple types of sensors, and outputting a unified state vector after processing by a cross-modal fusion algorithm. The reinforcement decision module combines expert knowledge graphs and deep reinforcement learning, adopting a dual-drive architecture and a dynamic multi-objective reward function. It can accelerate model convergence by relying on expert experience, and dynamically balance drying quality, efficiency, energy consumption, and equipment wear, outputting the globally optimal combination of parameters such as roller pressure and speed, effectively reducing wafer moisture residue and scratch defect rate. The quality feedback module constructs a four-dimensional real-time evaluation system based on physical, optical, electrical, and temporal trends, quantifying drying quality from multiple dimensions such as contact angle, surface defects, and resistivity distribution, and providing real-time feedback.
Need to check novelty before this filing date? Find Prior Art