| 引用本文: | 赵晓月,何书睿,陈先中,侯庆文.强干扰环境下高炉雷达信号机器学习算法[J].控制理论与应用,2016,33(12):1667~1673.[点击复制] |
| ZHAO Xiao-yue,HE Shu-rui,CHEN Xian-zhong,HOU Qing-wen.Machine learning algorithm of blast furnace radar in strong interference environment[J].Control Theory and Technology,2016,33(12):1667~1673.[点击复制] |
|
| 强干扰环境下高炉雷达信号机器学习算法 |
| Machine learning algorithm of blast furnace radar in strong interference environment |
| 摘要点击 2423 全文点击 1827 投稿时间：2016-06-29 修订日期：2017-01-18 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2016.60458 |
| 2016,33(12):1667-1673 |
| 中文关键词 高炉雷达 机器学习 C4.5决策树 希尔伯特--黄变换 边际谱 |
| 英文关键词 blast furnace radar machine learning C4.5 algorithm Hilbert-Huang transform marginal spectrum |
| 基金项目 |
|
| 中文摘要 |
| 高炉料面属矿物--煤气--焦炭多元高温固体/熔体混杂共存的粗糙表面,其电磁反射特征包含非均匀和非平
稳的气固混合介质的表层电磁回波信息、布料溜槽引起的周期性遮蔽效应、十字测温等装置引起的固定干扰,以及
电磁辐射等环境因素引发的随机噪声. 本文研究了复杂环境下连续调频波(FMCW)提取的料面信号,采用瞬时频率
分析的希尔伯特--黄变换(HHT)方法代替传统FFT方法. 结合经验模态分解,将原始非平稳信号分解为若干个平稳的
内在模式函数; 并按照基于先验知识的决策树算法分类与学习, 获得各类分量权值并加权突出真实物料的电磁信
号; 通过Hilbert变换得到原始信号的时频域特征, 可以揭示流态化料面包含的丰富冶炼信息.同时该算法也有助于
提高料面成像的帧准确率和稳定性,为钢铁行业节能减排提供可靠的数据支撑. |
| 英文摘要 |
| The burden surface of blast furnace belongs to rough surface that composes of solid or molten mineral-
gas-coke mixture in high temperature. The electromagnetic re?ection features include non-uniform and non-stationary
gas-solid ?uidized material electromagnetic echoes, periodic shadowing effect caused by distributing chute, ?xed inter-
ferences caused by cross temperature measurement as well as random noises caused by environmental factors such as
electromagnetic radiation. In this paper, the Hilbert-Huang transform (HHT) method of instantaneous frequency analysis
is used in place of the traditional FFT method to process the burden surface signals extracted by continuous frequency
modulation (FMCW) in complex environment. Combined with empirical mode decomposition, the original non-stationary
signal is decomposed into several stationary intrinsic model functions. Then the method classi?es and learns in accordance
with the decision tree algorithm based on prior knowledge, getting various types of signal component weight and weighting
and highlighting the real material electromagnetic signal. The time-frequency characteristics of the decomposed signals are
obtained by Hilbert transform, which can reveal the rich smelting information contained in the ?uidized burden surface.
Meanwhile the algorithm can also improve the frame accuracy and stability of the burden surface imaging and provide
reliable data support for the energy saving and emission reduction of the steel industry. |
|
|
|
|
|