首页|通过PO-KELM的3D NAND FLASH寿命预测方法研究

通过PO-KELM的3D NAND FLASH寿命预测方法研究

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
随着半导体行业的快速发展,以及各种芯片国产化的趋势越来越明显,3D NAND FLASH作为当前存储器件的重要代表,其寿命预测对于保障系统可靠性至关重要.因此,通过硬件搭建现场可编程门阵列采集平台,对 3D NAND FLASH进行特性分析,在不同擦除/写入次数下模拟FLASH可能发生的不同误码情况,分析耐久性、数据保持特性以及读、写干扰特性的变化趋势.同时提出鹦鹉优化器改进的核极限机器学习机,由于核极限学习机参数寻优困难,鹦鹉优化器通过搜索位置提高参数寻优速度和准确度.采用将已使用的循环次数作为输出结果对FLASH进行寿命预测.实验结果表明,相比其他机器学习,采用鹦鹉优化的核极限学习机预测模型精度可以达到 98.5%,在提升训练速度和准确度中具有重要的现实意义.
Research on life prediction of 3D NAND FLASH memory based on PO-KELM
With the rapid development of the semiconductor industry and the increasing trend towards domestic production of various chips,3D NAND FLASH,as a significant representative of current storage devices,its lifespan prediction is crucial for ensuring system reliability.Therefore,a hardware-based FPGA acquisition platform was built to conduct characteristic analysis of 3D NAND FLASH.It simulated different error scenarios that may occur under different erase/write cycles,analyzed the trends in durability,data retention characteristics,and read/write disturbance characteristics.Additionally,PO-KELM was proposed.Due to the difficulty in optimizing KELM parameters,the Parrot Optimizer enhances parameter optimization speed and accuracy by searching positions.The number of cycles used was adopted as the output for predicting the lifespan of FLASH.Experimental results show that compared to other machine learning methods,using PO-optimized KELM achieves a prediction model accuracy of 98.5%,which is of significant practical importance in improving training speed and accuracy.

3D NAND FLASHfield programmable gate array(FPGA)machine learningparrot optimizer(PO)kernel based extreme learning machine(KELM)

卜柯方、李杰、秦丽

展开 >

中北大学电子测试技术国防科技重点实验室,山西太原 030051

中北大学仪器科学与动态测试教育部重点实验室,山西太原 030051

3D NAND FLASH 现场可编程门阵列(FPGA) 机器学习 鹦鹉优化器(PO) 核极限学习机(KELM)

山西省基础研究计划项目山西省基础研究计划项目

202103021224186202303021221114

2024

10.11857/j.issn.1674-5124.2024060114

中国测试
中国测试技术研究院

中国测试

CSTPCD北大核心
影响因子：0.446
ISSN：1674-5124
年,卷(期)：2024.50(9)
  • 1