基于改进长短记忆网络的智慧工厂中设备故障预测
Equipment Fault Prediction in Smart Factory Based on Improved Long and Short Memory Network
石碧波1
作者信息
- 1. 通鼎互联信息股份有限公司,江苏 苏州 215200
- 折叠
摘要
结合实时提取的工厂数据,对设备进行故障预测是智慧工厂建设的重要步骤.提出基于改进长短记忆网络的智能工厂设备故障预测方法.对采集的设备运行数据进行归一化和异常值处理,以确保数据统一范围并排除异常值的干扰.使用主成分分析(PCA)法对预处理后的数据进行降维,基于时序分析法从降维后的数据中提取并选择具有高信息增益的特征.设计一种改进的长短时记忆网络模型,用于智能工厂设备故障预测.该模型通过全连接层输出故障预测结果.实证结果表明:采用本研究提出的方法可以获得更高的故障预测精准度,并且其预测耗时更短,较对比方法具有明显优势.
Abstract
Combined with real-time extracted factory data,fault prediction of equipment is an important step in smart factory construction.Introduce the intelligent factory equipment fault prediction method based on the improved long and short memory network.The collected equipment operation data are normalized and outlier treated to ensure uniform range of data and eliminate interference of outliers.Principal component analysis(PCA)was used to reduce the preprocessed da-ta,and features with high information gain were extracted and selected from the reduced data based on timing analysis.Design an improved long-short memory network model for smart factory equipment fault prediction.The model outputs the fault prediction results through the fully connected layer.
关键词
智慧工厂/电气设备/故障预测/预测性维护/设备管理Key words
smart factory/electrical equipment/fault prediction/predictive maintenance/equipment management引用本文复制引用
出版年
2024
NETL
NSTL
万方数据