ICA Based Vibration Detection of Unknown Systems and Its Active Suppression via Online Learning

基于ICA的未知系统振动检测及其在线学习主动抑制

• 批准号: 15560379
• 负责人: SUGIMOTO Kenji
• 金额: $ 2.37万
• 依托单位: Nara Institute of Science and Technology
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 2003
• 资助国家: 日本
• 起止时间: 2003 至 2004
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-15560379/
• 关键词: Independent Component Analysis; System Identification; Vibration Detection; Unknown Input; Mechanical Vibration

The objective of this research lies in introducing Independent Component Analysis(ICA), a new framework of methods for multivariable analysis, to identification or design of control systems. At this moment, frequency domain approaches are mainly adopted in ICA for dynamical systems. In control systems, however, it is more desirable if we can process in time domain. In view of this, the basic idea of this research is 1)to confine ourselves to the vector AR model, 2)expand the discrete-time signal to time series, 3)express the input-output relations in terms of algebraic equation with fixed elements (which is extended state space representation), and then 4)apply ICA in a transparent way.We first completed this basic idea as an algorithm and proposed it as a blind system identification technique. Then we verified the effectiveness of the proposed method by means of both theoretical analysis and numerical simulation. The result was presented in domestic and international conferences, and then in a journal paper.We further apply this method to two practical issues, namely detection of structure change and rejection of unknown disturbance. Concerning detection of change, we fabricated an experimental system with flexible structure and performed detection of a change due to a virtual fault. The result was successful and we presented it in domestic and international conferences. We have also submitted a paper to a journal.Concerning disturbance rejection, we verified the effectiveness by means of simulation and the result was presented in domestic and international conferences. Now we are preparing a paper for a journal. However, we have not yes succeeded in on-line suppression of vibration, which was mentioned in our original research plan. This is due to lack of a real-time performance in our method. It is our current research theme to overcome this difficulty.

本研究的目的在于将独立成分分析（伊卡）这一新的多变量分析方法框架引入到控制系统的辨识或设计中。目前，动力系统的独立分量伊卡主要采用频域方法。然而，在控制系统中，如果我们能够在时域中处理，则更可取。有鉴于此，本文的基本思路是：1）将研究对象限定在向量AR模型中; 2）将离散时间信号扩展为时间序列; 3）将输入输出关系用固定元素的代数方程表示（这是扩展的状态空间表示），4）以透明的方式应用独立分量伊卡），我们首先将ICA的基本思想作为一种算法加以完善，并将其作为一种盲系统辨识技术提出。通过理论分析和数值仿真验证了该方法的有效性。该成果在国内外会议上发表，并发表在期刊论文中，我们进一步将该方法应用于结构变化检测和未知干扰抑制两个实际问题。关于变化的检测，我们制作了一个具有灵活结构的实验系统，并执行由于虚拟故障的变化的检测。结果是成功的，我们在国内和国际会议上介绍了它。在抗干扰方面，我们通过仿真验证了该方法的有效性，并在国内外会议上发表了成果。现在我们正在为一本杂志准备一篇论文。然而，我们还没有成功的在线抑制振动，这是我们原来的研究计划中提到的。这是由于我们的方法缺乏实时性能。克服这一困难是我们当前的研究主题。

项目成果

A.Satoh, K.Sugimoto: "Partial Eigenstructure Assignment Approach for Robust Flight Control"AIAA Journal of Guidance, Control and Dynamics. 27・1. 145-150 (2004)

A.Satoh、K.Sugimoto：“鲁棒飞行控制的部分特征结构分配方法”AIAA 制导、控制和动力学杂志 27・1（2004 年）。

独立成分分析による機械システムの異状検出

使用独立组件分析检测机械系统异常

• 发表时间: 2004
• 期刊: 第48回システム制御情報学会研究発表講演会
• 作者: 鈴木 新;新田 益大;杉本 謙二
• 通讯作者: 杉本 謙二

独立成分分析を用いた自己回帰モデルの未知入力同定法

基于独立成分分析的自回归模型未知输入识别方法

• 发表时间: 2005
• 期刊: 計測自動制御学会論文集 41
• 作者: 新田益大;杉本謙二;佐藤淳
• 通讯作者: 佐藤淳

K.Sugimoto, Y.Kikkawa: "Control Systems Analysis via Blind Source Deconvolution"4-th and International Symposium on Independent Component Analysis and Blind Signal Separation. (CD-ROM). (2003)

K.Sugimoto、Y.Kikkawa：“通过盲源反卷积进行控制系统分析”第四届独立分量分析和盲信号分离国际研讨会。

H infinity Design with Loose Eigenstructure Assignment--A rank-one LMI Approach

松散特征结构分配的H无穷设计--一种一级LMI方法

• 发表时间: 2005
• 期刊: Transactions of the Society of Instrument and Control Engineers Vol.41
• 作者: A.Satoh;K.Sugimoto;M.Sebek
• 通讯作者: M.Sebek