Artificial Intelligence-Based Dynamic Adaptive NAND Flash Reading Method, System, Media, and Device

CN121459893BActive Publication Date: 2026-06-26JIANGSU XINSHENG INTELLIGENT TECH CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
JIANGSU XINSHENG INTELLIGENT TECH CO LTD
Filing Date
2025-10-24
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing technologies cannot effectively and dynamically adapt to threshold voltage drift caused by factors such as temperature, device aging, and read interference in NAND Flash memory, resulting in high read latency, increased power consumption, and a high risk of data loss.

Method used

An AI-based dynamic adaptive NAND Flash read method is adopted, which predicts voltage offset through a neural network model and combines it with an online transfer learning mechanism to achieve accurate read voltage adjustment and early warning of bad blocks, thereby optimizing the LDPC decoding process.

Benefits of technology

It significantly reduces read latency and power consumption, improves system real-time performance and reliability, reduces the risk of data loss, and increases the yield of storage arrays.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121459893B_ABST
    Figure CN121459893B_ABST
Patent Text Reader

Abstract

The application discloses a kind of dynamic self-adapting NAND Flash reading method, system, medium and equipment based on artificial intelligence, it is related to solid-state storage technical field, the method includes: S1, initiates reading request, uses default voltage to read page data, it is decoded by LDPC algorithm, constructs feature vector and is preprocessed, then it is input into neural network model that is trained in advance, voltage prediction is carried out;S2, fine sampling is carried out to predicted voltage, reconstructs accurate threshold voltage distribution, calculates log-likelihood ratio;S3, design shared bottom layer trunk network, so that neural network model has independent output branch, predicted voltage simultaneously evaluates the long-term health status of block;S4, design online transfer learning mechanism, so that neural network model uses a small amount of new data to quickly adapt to new environment.The application greatly reduces the delay and significantly saves power consumption, through early warning of bad block, data loss is prevented in the bud, and dynamic read interference suppression reduces write amplification factor to 1.26.
Need to check novelty before this filing date? Find Prior Art