• 沒有找到結果。

應用T-S模糊模型之順滑模態可靠度控制研究

N/A
N/A
Protected

Academic year: 2021

Share "應用T-S模糊模型之順滑模態可靠度控制研究"

Copied!
17
0
0

加載中.... (立即查看全文)

全文

(1)

行政院國家科學委員會專題研究計畫 成果報告

應用 T-S 模糊模型之順滑模態可靠度控制研究

研究成果報告(精簡版)

計 畫 類 別 : 個別型

計 畫 編 號 : NSC 97-2221-E-009-087-

執 行 期 間 : 97 年 08 月 01 日至 98 年 09 月 30 日

執 行 單 位 : 國立交通大學電機與控制工程學系(所)

計 畫 主 持 人 : 梁耀文

計畫參與人員: 碩士班研究生-兼任助理人員:魏源廷

碩士班研究生-兼任助理人員:王世昕

碩士班研究生-兼任助理人員:丁立偉

碩士班研究生-兼任助理人員:吳家榮

碩士班研究生-兼任助理人員:鄭旭志

碩士班研究生-兼任助理人員:謝宜展

博士班研究生-兼任助理人員:徐勝均

報 告 附 件 : 出席國際會議研究心得報告及發表論文

處 理 方 式 : 本計畫可公開查詢

中 華 民 國 98 年 09 月 28 日

(2)

行政院國家科學委員會補助專題研究計畫成果報告

應用

T-S 模糊模型之順滑模態可靠度控制研究

(Study of reliable sliding mode control based on T-S fuzzy system

model)

計畫類別:

5

個別型計畫 □ 整合型計畫

計畫編號: NSC

97-2221-E-009-087

執行期間: 97 年 08 月 01 日 至 98 年 07 月 31 日

計畫主持人:

梁耀文 副教授

共同主持人:

計畫參與人員:

成果報告類型(依經費核定清單規定繳交):

5

精簡報告 □完整報告

本成果報告包括以下應繳交之附件:

□赴國外出差或研習心得報告一份

□赴大陸地區出差或研習心得報告一份

5

出席國際學術會議心得報告及發表之論文各一份

□國際合作研究計畫國外研究報告書一份

處理方式:除產學合作研究計畫、提升產業技術及人才培育研究計畫、列

管計畫及下列情形者外,得立即公開查詢

□涉及專利或其他智慧財產權,

5

一年□二年後可公開查詢

執行單位:

國立交通大學電機與控制工程學系

華 民 國 98 年 8 月 30 日

(3)

行政院國家科學委員會專題研究計畫成果報告

應用 T-S 模糊模型之順滑模態可靠度控制研究

(Study of reliable sliding mode control based on T-S fuzzy system

model)

計畫編號:NSC 97-2221-E-009-087-

執行期限:97 年 8 月 1 日至 98 年 7 月 31 日

主持人:梁耀文 國立交通大學電機與控制工程學系副教授

一、中文摘要

隨著人類科技不斷的快速進步及推陳

出新,使得大家的生活舒適程度及交通便

捷性不斷的提昇,各種系統之設備規模、

複雜程度、所投注的資金以及可能造成的

災害威脅也因而大幅提高。因此,人們對

於系統之安全性、可靠性及有效性的要求

也變得格外殷切。由於系統的複雜程度不

斷提昇,其所使用的控制策略、所需的計

算時間以及其有效性便顯得格外重要,且

這些因素往往決定了系統控制的品質與成

敗。因此本計畫研究採用

T-S 模糊模型

(T-S fuzzy model)結合順滑模態(Sliding

mode control)理論來進行可靠度控制器的

分析與設計工作。採用順滑模態控制的原

因 在 於 它 所 擁 有 的 高 度 穩 健 特 性

(Robustness)及快速的反應能力;而採用

T-S 模糊模型來進行分析與設計的主要原

因是它只需計算相關的線性系統模型資

料,這對於具有高度非線性的複雜系統而

言,可以有效的減少計算時間,達到有效

控制的目的。因此,在本研究計畫裡我們

進行

T-S 模糊模型及順滑模態控制策略結

合之可靠度控制器的研發。

關鍵詞:非線性控制系統、順滑模態控制、

TS 模糊模型、可靠度控制、穩定性分析、

軌跡追蹤。

Abstract

Due to the growing demands for system

reliability in a highly automated industrial

system and in aerospace missions, where

repair and maintenance often can not be

achieved immediately, the study of reliable

control has become paramount importance

and has attracted considerable attention. On

the other hand, since the modern control

systems are constructed more and more

complicated, the employed control strategy

and the time for controller implementation

have become extreme importance. In fact,

the two mentioned-factors have a strong

relation to the quality and the efficiency of

the control mission. In this project, we

combine the T-S fuzzy model approach and

the Sliding Mode Control (SMC) scheme for

alleviating the computational burden and

promoting associated system reliability

performances. The reason for adopting SMC

scheme comes from its own advantages

including responding rapidly and robustness

to uncertainties and disturbances, while T-S

fuzzy approach allows one to save lots of

on-line computational burden, which is

especially important for those systems with

highly nonlinear and complicated dynamics.

The combined scheme saves lots of on-line

(4)

computational burden while achieve

efficiently control objective.

Keywords: Nonlinear control systems,

sliding mode control, TS-fuzzy model,

reliable control, stabilizability analysis,

tracking performance.

二、緣由與目的

1、背景說明及計畫重要性

據美國軍用航空部門統計,美國在二

次大戰期間飛機因為故障而損失的數量

比起被擊落數量整整多了

1.5 倍,而運往

遠 東 的 設 備 經 過 運 輸 後 有

60%不能使

用,使用後的維修也成為重大問題,所以

他們首先體認到可靠度控制的重要性與

當系統缺乏可靠性的重大代價。有鑒於

此,美國國防部在戰後投注大筆經費進行

裝備可靠性的研究,開啟了錯誤偵測與診

斷及可靠度控制的研究領域。近年來由於

科技的發達及航太工業的突飛猛進,提供

給人類舒適的生活環境及快捷之交通便

利。各種系統之設備規模,複雜程度及所

投注的資金也因而大幅提昇。因此,人們

對於系統之安全性、可靠性及有效性的要

求也越來越殷切。尤其是航空、太空、核

電廠及化工廠等等具有高危險性的特殊

機具,更可能由於系統的不穩定而導致重

大災難。舉例來說:

1979 年美國三里島核

能電廠的意外事件、

1986 年 1 月美國挑戰

者號及

2003 年 2 月哥倫比亞號太空梭的

不幸事件、1998 年 8 月至 1999 年 5 月的

短短

10 個月之間,美國 3 種運載火箭“大

力神”、“雅典娜”、“德爾他”共發生了 5

次發射失敗,造成

30 多億美元的直接經

濟損失,美國的太空計劃也因此受到嚴重

打擊。美國在

2004 年秋天正式啟用位於

阿拉斯加葛瑞利堡的美國飛彈防禦系統

基地,根據調查報導,這項斥資超過

1000

億美元研發的飛彈防禦系統效用受到嚴

重地懷疑,在關鍵系統研發屢次的延遲

下,只有進行了基本的試射測試,由於相

關的可靠度控制和錯誤偵測與診斷的效

能也尚未完全,所以目前實際上的效用據

評估可能僅有

20%的防禦能力,而美國在

未來

5 年內,每年還將繼續投入 90 到 100

億美元在此飛彈防禦系統上。這些相關的

事件與分析都充分說明了錯誤偵測與診

斷和可靠度控制這研究主題的重要性。

由於這股研究熱潮方興未艾,在學術

界方面也持續的吸引了許多世界各地學

者專家高度的重視。目前已有許多關於此

方面的研究成果和理論不斷的被發表和

提出,而諸多的重要國際學術會議如美國

控制研討會(ACC)、IEEE 控制與決策會議

(CDC)及國際控制聯合大會(IFAC)也都將

此項研究主題列為重要的討論議題。美國

電機電子工程師學會

(IEEE)亦成立可靠

度學會(IEEE Reliability Society),出版關

於可靠度方面的期刊:IEEE Transactions

on Reliability,而在其他電機、電子、控

制和資訊等相關的期刊,也有許多專家學

者 專 門 研 究 探 討 關 於 可 靠 度 控 制 的 問

題,足見此方面研究主題的重要及迫切

性。

2、研究目的

誠如華衛二號衛星

(ROCSAT-2)在其

設計中運用了四個旋轉輪(reaction wheel)

來做姿態控制,其目的就是希望當有一具

旋轉輪發生故障時,衛星姿態仍能獲得有

效的控制,也就是說系統被要求能具有容

錯的能力。而這容錯的能力也正是可靠度

控制研究的主要目的:希望系統不論在正

常或部分故障時仍能維持特定的性能表

現。因為對於遠在太空的衛星系統而言,

一旦系統發生故障,其回收維修相當不容

易而且回收所需耗費的成本極高。而對於

一般系統而言,當系統發生故障時,所需

的零件及維修通常也無法及時有效的供

(5)

應。因此,對於控制系統可靠度的要求益

發引起人們的關注。一般而言,“故障”會

使系統表現出不希望的特性,而當動態系

統中出現部分功能喪失時,將會導致整個

系統性能的惡化,甚至引發系統的不穩

定。為了有效的提升及確保系統在正常操

作及部分故障時的性能,我們除了可以經

由 一 些 製 造 技 術 來 提 高 產 品 的 品 質 之

外,另一個重要而且可行的方式,則是透

過不斷監測系統運行的狀態並預測發展

之趨勢,盡可能把可能發生的故障消除在

剛開始發生的階段。而為了有效獲知系統

的 運 作 狀 況 ,

“ 故 障 偵 測 與 診 斷 ”(fault

detection and diagnosis)的學問乃因應而

生。此故障偵測與診斷機制的主要目的就

是希望當系統發生故障或異常現象時能

及時的發出警告訊號,並分離出異常狀況

的原因,來源及嚴重程度,提供給決策機

制採取最合理、最正確的處置以及採取最

適當有效的控制策略,以避免設備的損壞

及造成可能的不幸。由於任何一個控制系

統均無法避免故障或異常現象的發生,因

此如何有效的避免故障的發生,或者是如

何能適度地容忍輕微的故障發生,是一個

相當實際且重要的研究課題。因此。本計

劃之研究目的是希望能研發新一代的可

靠度控制技術使系統具有容錯能力及高

可靠性,並希望能將此研發技術應用至實

際系統上使其具有安全及容錯的目標。同

時,經由此研發的過程為國家培養相關之

電機控制研究人才。相信藉由研究訓練、

技術累積和研發成果,對於國內學術的研

究發展,國防科技的技術提升和工業的實

務應用都必定能有所助益。

三、結果與討論

考慮如下非線性二階系統:

2 1

x

x

&

=

x

&

2

=

f

(

x

)

+

B

u

+

d

(1)

其中

n T n n n T n

x

x

x

x

R

x

R

x

1

=

(

1

,

L

)

,

2

=

(

+1

,

L

2

)

為 系 統 狀 態 ,

為 控 制 輸 入 ,

為可能之模型不確定

性及外界干擾,

是平

滑函數,

T T T

,

)

(

x

1

x

2

x

=

m n T m n

u

u

+

+

=

R

u

(

1

,

L

,

)

n T n

d

d

R

d

=

(

1

,

L

,

)

n

R

x

f

(

)

G

(

x

)

R

n

( )

T

表示一向量或矩陣之轉置。在

本研究,我們假設

。在系統(1)中的

描述,我們已經假設系統有冗餘的控制輸

入。將制動器分為兩群 H 與 F,代表的是

健康的制動器與允許損壞的制動器。系統

(1)可以重寫成:

0

0

f

(

)

=

2 1

x

x

&

=

x

&

2

=

f

(

x

)

+

B

H

u

H

+

B

F

u

F

+

d

.

(2)

我們假設預先選取之健康的制動器滿足

為非奇異矩陣。

n H

R

u

B

H

∈ R

n×n

本計劃的目標是組織一個適當的控制

使得即使當所有的或部分的制

動器在 F 集合中發生損壞的狀況時,閉迴

路系統的原點仍然漸近穩定。

H

u

u

F

A. T-S 模糊系統模型

T-S 模糊模型是藉由多個線性系統之

組合而成的系統來近似原有的非線性系統

(1)。假設有第 i 個(

i

=

1

,

2

,

L

,

p

)T-S 模糊模

型對系統(2)的規則描述如下:

如果

ς

1

q i

M

1

,

L

,

ς

M

,

i

1

,

,

p

,

i q

=

L

2 1

x

x

&

=

x

&

2

=

f

(

x

)

+

B

H

u

H

+

B

F

u

F

(3)

其中

ς

1

,

L

,

ς

q

為假定變數,

對假定變數之隸屬函數,p 和 q 標示為規

則與假定變數的個數,及

。T-S

模糊則根據各別線性系統(4)系統狀態的權

重而被建立,如下所示:

i i

M

q

M

1

,

L

,

n n i

A

∈ R

× 2 1

x

x

&

=

p H H F F

(4)

i i i

x

A

x

B

u

B

u

x

=

+

+

=1 2

α

(

)

&

B. SMC 可靠度控制器設計

藉著與

T-S 模糊模型的關聯,系統(2)

可以改寫為

x

&

1

=

x

2

(5)

x

=

x

x

+

f

+

u

+

u

+

d

=1

(

)

(

)

2 H H F F p i i i

A

B

B

α

&

(6)

其中

=

p= i 1 i

A

i

(

)

)

(

:

f

x

x

f

α

因為(1)為一個二階系統的集合,我們可以

假設順滑面為

(6)

)

(

)

(

)

,

,

,

(

:

s

1

s

2

s

n T

x

2

t

M

x

1

t

s

=

L

=

+

(7)

其中

n n

M

∈ R

×

為一個正定矩陣。清楚的,

假設系統狀態仍然在順滑平面上時,則想

要之

之穩定化表現可以指數性的

逼近。為了補償干擾或不確定的效果,我

們引入假設

0

x

(t

)

假設

1:存在非負函數

ρ

j

(

x

,

t

),

j

=

1

,

L

,

n

,

使得

,其中

可能數值及

)

,

(

|

|

|

)

(

|

|

)

(

|

B

F

u

*F j

+

f

j

+

d

j

ρ

j

x

t

* F

u

u

F

(⋅ 標示為第 j 列的向

)

j

量。

遵循

SMC 設計的流程[2],我們選擇

+

+

Λ

=

= −

(

)

sgn(

)

2 1 1

s

x

x

x

u

i H p i i H H

B

α

A

M

(8)

其中

)

t)

,

(

,

,

t)

,

(

diag(

ρ

1

+

η

1

ρ

n

+

η

n

=

Λ

H

x

L

x

j

=

1 L

,

,

n

η

j

>

0

標示為訊號函

數且

,即在控

制器

下,它遵循(5)-(7)及假設 1 使得

。這個不等式使得系統狀

態無論當在

F 內的控制器健康或者損壞時

都可以在有限時間內接觸到順滑平面 [2]。

)

sgn(⋅

T n

s

s

),

,

sgn(

))

(sgn(

:

)

sgn(

s

=

1

L

H

u

=

n j j j T

s

1

η

|

|

s

s

&

除了

的設計外,當部分或者全部在

F 中的制動器為健康時,我們也將設計

來提升整個系統的表現。從(5)-(6) 及(8),

則 我 們 有

清楚的,使得系統狀態接近順滑平面比

來的快的一個

的選擇是

H

u

F

u

F F T T

B u

s

s

s

&

nj=1

η

j

|

s

j

|

.

0

u

F

=

u

F

)

sgn(

s

u

F

=

Λ

F

B

TF

(9)

其中

Λ

F

=

diag

(

η

n+1

,

L

,

η

n+m

)

η

n+ι

0

所有

ι

=

1

,

L

,

m

.

。這些推導證明了控制增益

ι

η

n+

,

ι

=

1 L

,

,

m

,的大小對在 F 集合中的

制動器保證可穩定化的表現,可以從零到

可允許的最大控制輸入的變化範圍。這允

許了在 F 中的制動器可以完全損壞,部分

損壞,吸收或放大在任何階及任何組合的

狀況下,因此有了以下定理。

定理

1:如果假設 1 滿足,在(8)與(9)的控

制律之下,即使當部分或全部的制動器為

不正常操作,系統(2)的原點為局部漸近穩

定。

C. 應用於衛星系統

一個在圓形軌道上的衛星姿態控制可

以描述成(1)的形式,其中 n=3 [11]。六個

狀態標示為三個尤拉角

(

φ

,

θ

,

ϕ

)

與它們的

微分。為了簡化,我們假設在本研究中,

火箭推進器只能用控制力及有一個冗餘的

制 動 器 來 表 現 可 靠 度 任 務 。 藉 著 使

T

)

,

,

,

,

,

(

φ

θ

ϕ

φ

&

θ

&

ϕ

&

=

x

f

(

x

)

=

(

f

1

(

x

),

f

2

(

x

),

f

3

(

x

))

T

,整個系統動態可以描述如[2]

其中

.

28

.

0

28

.

0

28

.

0

28

.

0

69

.

0

69

.

0

69

.

0

69

.

0

67

.

0

67

.

0

67

.

0

67

.

0

=

B

這裡,

為對應三個體座標軸的慣

性矩,

y x

I

I ,

I

z 0

ω

標示為常數軌道率,c 與 s 標示

cos 與 sin 函數。

為了獲得適當的

T-S 模型來近似原始

非線性動態,我們首先令

。一

項的集合有下列的形式:

x

x

x

f

(

)

=

A

(

)

)

(x

A

=

1 1 3 2 2 0 1 , 1

2

)

2

(

))

(

(

x

x

s

x

c

I

I

I

x

A

x z y

ω

,

2

)

2

(

3

2

)

2

(

1 1 2 2 2 0 1 1 3 2 2 2 2 0

x

x

s

x

c

x

x

s

x

s

x

s

ω

ω

=

2 2 1 2 3 2 0 2 , 1

4

(

2

)

1

))

(

(

x

sx

x

c

x

s

I

I

I

x

A

x z y

ω

,

)

2

(

4

1

1 2 2 2 3 2 0

s

x

x

sx

x

s

ω

2 1 2 3 3 2 0 3 , 1

)

2

(

2

1

))

(

(

c

x

s

x

x

x

s

I

I

I

x

A

x z y

=

ω

,

2

)

2

(

2

1

1 2 2 3 3 2 0

sx

s

x

x

x

s

ω

,

0

))

(

(

A

x

1,4

=

x z y

I

I

I

sx

sx

x

A

(

))

1,5

=

0 3 2

+

(

ω

,

2

1

1 3 0 2 3 1 0 6

⎥⎦

⎢⎣

+

+

x

ω

cx

sx

sx

ω

cx

sx

1 3 0 5 2 3 0 6 , 1

2

1

))

(

(

x

cx

cx

I

I

I

cx

cx

x

A

x z y

ω

ω

+

⎢⎣

+

=

]

,

1 2 3 0

sx

sx

sx

ω

,

)

2

(

4

1

))

(

(

3 1 1 2 2 0 1 , 2

=

s

x

x

sx

cx

I

I

I

x

A

y x z

ω

=

2 3 1 2 2 2 0 2 , 2

2

)

2

(

))

(

(

s

x

cx

x

x

s

I

I

I

x

A

y x z

ω

(7)

,

2

)

2

(

3

1 2 2 2 0

cx

x

x

s

ω

,

2

)

2

(

2

1

))

(

(

3 3 1 2 2 0 3 , 2

=

x

x

s

sx

cx

I

I

I

x

A

y x z

ω

y x z

I

I

I

sx

sx

sx

cx

cx

x

A

(

))

2,4

=

0 3 1

+

0 3 2 1

+

(

ω

ω

,

2

1

1 3 0 2 3 1 0 6

⎥⎦

⎢⎣

+

+

x

ω

cx

sx

sx

ω

cx

sx

,

))

(

(

A

x

2,5

=

ω

0

sx

3

cx

2

sx

1 1 2 3 0 1 3 0 6 , 2

))

(

(

A

x

=

ω

sx

cx

+

ω

cx

sx

sx

,

2

1

2 3 0 4

⎥⎦

⎢⎣

+

x

sx

cx

I

I

I

y x z

ω

,

)

2

(

4

3

))

(

(

1 1 2 2 0 1 , 3

=

x

sx

x

s

I

I

I

x

A

z y x

ω

=

2 2 1 3 2 2 0 2 , 3

2

)

2

(

))

(

(

x

x

s

x

s

x

s

I

I

I

x

A

z y x

ω

,

2

)

2

(

2

3

1 2 2 2 0

sx

x

x

s

ω

,

2

)

2

(

))

(

(

2 1 3 3 2 0 3 , 3

=

cx

cx

x

x

s

I

I

I

x

A

z y x

ω

z y x

I

I

I

cx

cx

sx

sx

sx

x

A

(

))

3,4

=

0 3 2 1

0 3 1

+

(

ω

ω

,

2

1

1 2 3 0 1 3 0 5

⎥⎦

⎢⎣

+

x

ω

cx

cx

ω

sx

sx

sx

,

))

(

(

A

x

3,5

=

ω

0

sx

3

cx

2

cx

1

,

2

1

2 3 0 4

⎥⎦

⎢⎣

+

x

sx

cx

I

I

I

y x z

ω

1 3 0 13 1 2 0 6 , 3

))

(

(

A

x

=

ω

sx

cx

cx

+

ω

sx

sx

其中

( x

A

(

))

i,j

標示為矩陣

A

(x

)

之(i,j)-項。

為了建構相對應的

T-S 模型。我們將

從可能的工作區間來選擇適當的操作點,

使得衛星的動態可以被此

T-S 模型很好的

近 似 。 在 這 個 例 子 中 , 我 們 假 設

, 同 時 我 們 假 設

,

2000

N

m

s

2

I

I

x

=

z

=

I

y

=

400

N

m

s

2

,

s

rad /

10

0312

.

1

3 0 −

×

=

ω

]

2

/

,

2

/

[

1

π

π

x

x

2

[

π

,

π

]

]

2

/

,

2

/

[

3

π

π

x

。為了觀察前鑑部變數

(premise variable)的個數的效果,我們考慮

下列兩種方案。

案例 A:(考慮三個角度為前鑑部變數)

在 本 案 例 , 操 作 點 選 擇 為

} , , 1 , , , 1 , , , 1 | ) 0 , 0 , 0 , , , ( {xi,j,k= x1,i x2,j x3,k T i= Ln1 j= Ln2k= Ln3

其 中

}

,

,

{

1 , 1 1 , 1

x

n

x

L

{

,

,

}

2 , 2 1 , 2

x

n

x

L

}

,

,

{

x

3,1

L

x

3,n3

[

π

/

2

,

π

/

2

]

[

π

,

π

]

]

2

/

,

2

/

[

π

π

,為相對應三個所選擇的切割

點。在本案例,我們選擇

n

1

=

n

2

=

n

3

=

5

引入三角隸屬函數,如同圖

1 所描述。

1 案例 A 之三角隸屬函數

案例 B:(考慮六個前鑑部變數)

在本案例,操作點選成

=

6 , 5 , 4 , 3 , 2 , 1

{

x

i i i i i i j j T i i i i i i

x

x

x

x

x

i

n

x

,

,

,

,

,

)

|

1

(

1,1 2,2 3,3 4,4 5,5 6,6

為可能的整數,其中

。在本

例中,我們選擇

,並且

也使用了三角隸屬函數,如圖

2 所示

j

n

j

=

1 L

,

,

6

}

2

=

j

n

j

=

1 L

,

,

6

2 案例 B 之三角隸屬函數

模擬結果總結在圖

3-5。我們使用了下

列三種控制方式:SMC 可靠度設計[2](標

示為

SMC),T-S 模型為基礎之可靠度方法

對 應 於 不 同 的 前 鑑 部 變 數 個 數

( 標 示 為

Case A 與 Case B)。SMC 可靠度設計之參

數為

M

=

2

I

3

,

η

j

=

0

.

5

對所有

η

j

Λ

H

F

Λ

對所有

的 j,訊號函數替換為飽和函數,飽和函數

的邊界寬度為

0.05 來減輕訊號函數所產生

T

t

t

t

),

0

.

1

cos(

),

0

.

1

cos(

5

))

sin(

1

.

0

(

=

d

T

)

2

.

0

,

3

.

1

,

3

.

0

,

5

.

1

,

07

.

0

,

7

.

0

(

)

0

(

=

x

|

u

j

|

1

(8)

的切跳現象。此外,我們選擇

當做可能

損壞的控制器,也就是

在 t = 2 時損壞。可以從圖 3

觀察在上述的三種控制方式中,穩定化表

現都能成功的被達成。然而,因為

T-S 模

型在案例

B 中非常接近原始非線性模型狀

態軌跡,順滑變數與案例

B 的控制曲線與

SMC 可靠度設計非常接近,這可以從圖 3-5

了解。透過直接計算消耗能量與平方性

能,有如下之關係:

2

u

}

,

,

{

u

1

u

3

u

4

H

=

}

{

u

2

F

=

u

2 314 . 8 ) ( 930 . 3 ) ( 917 . 3 ) ( ≈ ≤ ≈ ≤ T CaseACaseB T SMC T u u u u u u

256

.

5

)

(

254

.

5

)

(

T SMC

CaseB T

x

x

x

x

602

.

5

)

(

T CaseA

x

x

清楚的,

SMC 與案例 B 在如下之兩種性能

指標

都非常接近,而案例

A

消耗較更多能量與較大的

。值得一提

的是從圖

5(b)可看出

在兩秒後損壞且

0.75 秒附近改變正負號,

符號的

改變可以從

的正負號改變來確認,這與

(9)相符合。此外,由於在模擬中使用飽和

函數,案例

A 之

的值在

u

T

u

x

T

x

x

T

x

2

u

2

u

u

2

s

2

B

2

u

0

.

75

≤ t

2

時是

0.5 , 這 並 不 與 (9) 相 違 背 因 為 此 時

。最後,當重複計算控制器

次後,T-S 型設計(包含隸屬權重決定),

CPU 運算時間較傳統 SMC 為少,各控制

方法所耗費的計算時間有著以下關係:

1

|

|

B

FT

s

>

5

×

10

4

453

.

7

)

CPU

(

087

.

5

)

CPU

(

CaseA

CaseB

313

.

10

)

CPU

(

SMC

。從這些模擬,可以

發現

SMC 與案例 B 的性能彼此非常接

近,且比案例

A 好。然而,案例 A 消耗較

少的時間在控制器實現上,因為它只有使

用三個角度當作前提變數。此外,所提出

T-S 型方法不只減輕線上運算負擔,也

可以有效的完成穩定化任務,如同

SMC 設

計一般,且以

T-S 模型為基礎的方法在前

鑑部變數之分割變得更細的時候不會產生

額外線上運算負擔,。

3 系統之六個狀態

4 系統之順滑變數

5 控制器輸入

本計畫所獲得的研究成果已整理並在

期刊及研討會發表(詳計畫成果自評及附

件)。

(9)

四、計畫成果自評

本計畫的主要目的在於探討如何結合

TS-模糊模型及變結構控制策略進行穩定

性及軌跡追蹤的可靠度控制任務。針對本

計畫之研究主題,我們在這一年內已完成

下列工作項目:

1. 提出 T-S 模糊模型為基礎的可靠度

設計,即使當可能損壞的制動器發生

部分損壞或者全部損壞時依然可以

達到穩定化之性能表現。

2. 提供的方法可以大量減輕線上運算

負擔,因為所使用來近似原始非線性

系統的

T-S 模糊系統模型之系統參數

大部分可以離線取得。

3. 所提供的方法有著快速響應以及穩

健的特性,因為採納了順滑模控制技

術補償了額外干擾及在非線性與

T-S

模糊系統之間模型的不確定性。

4. 即使增加模糊法則的數量也不會產

生額外的線上運算負擔,衛星系統的

模擬例子清楚的展示了所提方案之

效率與好處。

就計畫而言,我們已經達到了預期的

成果。這些研究成果有些已發表於國外著

名期刊及研討會,包括

[1] Y.-W. Liang, S.-D. Xu, D.-C. Liaw, and

C.-C. Chen, “A Study of T-S

Model-Based SMC Scheme With

Application to Robot Control,” IEEE

Transactions on Industrial Electronics,

Vol. 55, No. 11, pp. 3964-3971, 2008.

[2] Y.-W. Liang, S.-D. Xu, and L.-W. Ting,

“T-S Model-Based SMC Reliable

Design for a Class of Nonlinear

Systems,” IEEE Transactions on

Industrial Electronics, accepted for

publication, 2009.

[3] Y.-W. Liang, S.-D. Xu, and C.-C. Chen,

“Study of a combination of the T-S

fuzzy and the SMC approaches,” CACS

Automatic Control Conference (

中華民

國自動控制研討會

), Tainan, Taiwan,

Nov. 21-23, 2008.

五、參考文獻

[1] R. A. Decarlo, S. H. Zak, and G. P.

Matthews, “Variable structure control of

nonlinear multivariable systems: a

tutorial,” proc. IEEE, vol. 76, no. 3, pp.

212-232, 1988.

[2] Y.-W. Liang , S.-D. Xu, and C.-L. Tsai,

“Study of VSC reliable design with

application to spacecraft attitude

stabilization,” IEEE Trans. Control

Systems Technology, vol. 15, no. 2, pp.

(10)

出席國際學術會議心得報告

計畫編號 NSC

97-2221-E-009-087

計畫名稱

應用 TS 模糊模型之順滑模態可靠度控制研究

出國人員姓名

服務機關及職稱

梁耀文,國立交通大學電機與控制系,副教授

會議時間地點 August 18-21, 2009,日本,福岡(Fukuoka)

會議名稱 ICCAS-SICE

2009

發表論文題目 SMC Reliable Design for T-S Model-Based Systems

一、參加會議經過

此次國際學術研討會議名稱為 ICROS-SICE International Joint Conference 2009

(ICCAS-SICE 2009),是由 The Society of Instrument and Control Engineers (SICE)及

The Institute of Control, Robotics and Systems (ICROS)主辦,協辦單位包括有 IEEE

Industrial Electronics Society、IEEE Robotics and Automation Society、IEEE Control

Systems Society、The Instrumentation, Systems, and Automation Society (ISA) 、 Asian

Control Association (ACA) 、China Instrument and Control Society (CIS)、Chinese

Association of Automation (CAA)、Chinese Automatic Control Society (CACS) 、

International Measurement Confederation (IMEKO) 、 IEEE Japan Council 、 IFAC

NMO-Japan、The Institute of Electrical Engineers of Japan。會議期間為西元 2009 年八

月 十 八 日 至 八 月 二 十 一 日 , 地 點 為 日 本 福 岡 之 Fukuoka International Congress

Center。我們在八月十八日由新竹出發至中正機場,搭機飛往日本福岡機場。本人

之論文報告日期被安排在八月十九日週三上午九點第一位上台報告,會場有多位學

者專家對本人之研究成果甚感興趣,除了聽取這些學者專家的意見與建議外,本人

也積極的與他們分享研究成果,自覺收穫豐碩。除此之外,會議期間我們也聆聽了

諸多與自己研究領域相關之最新的研究報告,對於專家學者之認真與專注及求甚解

之精神印象深刻,自覺收穫良多。會議後我們也遊覽福岡及其附近之旅遊景點,包

括有將台灣割讓給日本之馬關條約簽約現場,感觸良多。並於八月二十二日返回台

灣。

二、與會心得

由於參加此次研討會的學者專家甚多,因此這次的學術研討會議包含的研究主

(11)

題也相當廣泛。除了有許多的學術理論研究成果外,也包含有許多工業應用的應用

成果展現。其中有量測、控制、系統資訊、系統判別、計算機及工業電力電子等相

關應用之議題。這次的會議主要是以兩種形式呈現,一種是以海報的方式展現研究

成果,另一種則是以口頭報告的形式呈現研究心得。我是以口頭報告的形式參與發

表研究成果。會議期間觀摩各地學者專家呈現研究成果的方式,並互相交換研究心

得。在控制領域方面,研究成果包括有控制理論、非線性系統控制、智慧型控制、

模糊理論、類神經網路、基因演算法、電力系統、電力電子、工業資訊學、電腦與

控制技術、感應器與致動器等等。此次會議之主要目的在於探討目前控制理論之趨

勢及其應用的最新發展,與會人士除了地主國日本外,還有許多來自世界各地之學

者專家。大家齊聚一堂共同研討及分享彼此之研究成果與研究心得。參加此次會議

不但能親自目睹及體會著名學者的風範及研究執著精神,萌生見賢思齊的成長動

力。同時能增進國際會議發表論文的膽識及獲致國際學術界最新的研究動態。在會

議期間許多控制界的前輩所提出的問題及意見不但具體而深入,且往往能點出問題

之所在及提供進一步研究方向與可能解答之輪廓,個人自覺收穫良多。也希望自己

能有多些機會參與類似的國際會議。在這次的會議後,我們攜帶回來了大會的光碟

資料,使我們可以盡覽此次會議的所有成果。在此願再一次的感謝國科會的經費補

助,使我能順利的參與這次的研究成果發表饗宴。

(12)

SMC Reliable Design for T-S Model-Based Systems

Yew-Wen Liang, Sheng-Dong Xu, Der-Cherng Liaw, Cheng-Chang Chen, and Li-Wei Ting

Department of Electrical and Control Engineering, National Chiao Tung University, Taiwan (Tel: +886-3-5712121 Ext: 31669; E-mail: [email protected])

Abstract: This paper studies the robust reliable control issues based on the Takagi-Sugeno (T-S) fuzzy system modeling

method and the sliding mode control (SMC) technique. The combined scheme is shown to have the merits of both approaches. It not only alleviates the on-line computational burden by using the T-S fuzzy system model to approximate the original nonlinear one (since most of the system parameters of the T-S model can be computed off-line) but it also preserves the advantages of rapid response and robustness of the SMC schemes. Moreover, the combined scheme does not require on-line computation of any nonlinear term of the original dynamics and the increase in the partition number of the region of premise variables does not create extra on-line computational burdens for the scheme. Under the design, the control mission can continue safely without prompt external support even when the susceptible actuators fail to operate. The proposed analytical results are also applied to the attitude control of a spacecraft. Simulation results demonstrate the benefits of the proposed scheme.

Keywords: Sliding mode control, reliable control, nonlinear control systems, T-S fuzzy system model.

1. INTRODUCTION

Recently, the study of reliable (or fault-tolerance) con-trol has attracted considerable attention (see, e.g., [4], [9], [10], [11], [12], [17], [19]). The objective of reli-able control is to design an appropriate controller such that the closed-loop system can tolerate the abnormal op-erations of specific control components and retain the overall system stability with acceptable system perfor-mance. Among the existing reliable control studies, sev-eral approaches have been presented. These approach-es include the linear matrix inequality (LMI)-based ap-proach [12], the algebraic Riccatti equation (ARE)-based approach [17], the coprime factorization approach [18], the Hamilton-Jacobi (HJ)-based approach [9], [19], and the sliding mode control (SMC)-based approach [4], [10], [11]. Among the above-mentioned reliable con-trol studies, only the HJ-based and the SMC-based ap-proaches deal with reliability issues for nonlinear sys-tems. However, the reliable controller of the HJ-based approach explicitly depends on the solution of an asso-ciated Hamilton-Jacobi equation which is in general dif-ficult to solve. Though a power series method [8] may alleviate the difficulty through computer calculation, the obtained solution is only approximate and the computa-tion load grows quickly when the system is complicated. In contrast, the SMC reliable controllers do not require the solution of any HJ equation, while retain the tages of conventional SMC designs [11]. Those advan-tages include rapid response, robustness, and ease of im-plementation [11], [13], [14], [15].

On the other hand, because of the conceptual simplic-ity and the fact that most of the system parameters can be computed off-line, the Takagi-Sugeno (T-S) modeling scheme has become a popular and powerful fuzzy sys-tem modeling approach (see, e.g., [1], [13], [16]). The basic idea of the T-S approach is first to decompose a nonlinear system into several linear models according to different cases where the associated linear models best fit

the nonlinear one, and then to aggregate each individual linear model into a single nonlinear one in terms of each model’s membership functions. Though the concept is simple, the T-S fuzzy system model has been theoretical-ly justified as a universal approximator which makes the T-S fuzzy system model become particularly useful, es-pecially when the nonlinear model is complicated. In or-der to compensate for the additional uncertainties result-ing from the difference between the original and the T-S models, a combined scheme incorporated with the SM-C technique was recently proposed (see, e.g., [13]). The combined scheme not only alleviates the on-line compu-tational burden since the T-S fuzzy system model is uti-lized to approximate the original nonlinear one, but it also preserves the advantages of rapid response and robust-ness of the SMC schemes. In light of those remarkable benefits, this paper will investigate the reliability issues from the combined scheme viewpoint.

2. PROBLEM STATEMENT

Consider a class of 2nd-order nonlinear control sys-tems   and     (1) where          ,              and        

is the system states,

          

is the control inputs,

      

  

denote possible model un-certainties and/or external disturbances,  



is a smooth function, and



denotes the transpose of a vec-tor or a matrix. In this study, we assume that. It

is important to note that in the description of the system given by Eq. (1) we have assumed that the system has control input redundancy. We divide the actuators into

two groupsand, within which we assume that

al-l of the actuators inare healthy while those in are

(13)

rewritten as  and          (2)

Since the nonsingularity assumption of is necessary

for the existence of equivalent control in SMC design when all the actuators infail to operate [6], we assume

that the pre-selected healthy actuators satisfy  

,

and 



is a nonsingular matrix.

The objective of this study is to organize an

appropri-ate and

so that the origin of the closed-loop

sys-tem is asymptotically stable even when all or some of the actuators in the setfail to operate.

3. T-S MODEL-BASED SMC RELIABLE

DESIGN

In light of the advantages of the T-S modeling and SM-C approaches as stated above, this study will combine the two schemes for the design of reliable controllers.

3.1 T-S Fuzzy Model Description

It is known that a nonlinear system can be approximat-ed by a T-S fuzzy model ([13], [16]), which is describapproximat-ed by a combination of several linear models with suitable weighting. Theth (  ) rule of the T-S fuzzy

model for System (2) has the following form:

If is   is    then   and          (3) where , ,

 are premise variables,

, ,



are membership functions for premise variables, and

denote the number of rules and premise variables,

re-spectively, and 

 

. The T-S fuzzy model is then constructed according to the weight of the system state on each linear model as (4) below:

 and               (4)

where the weightings



   0 for all  and

     . 3.2 SMC Reliable Design

By incorporating with the T-S fuzzy model, System (2) can be rewritten as    (5) and                (6) where          . Since System (1)

contains a set of 2nd-order systems, we may assume the sliding surface to be          ¾  ½  (7) where  

is a positive definite matrix. Clearly, if the system state remains on the sliding surface, then

the desired stabilization performance of   can

be exponentially achieved. To compensate for the effects

of disturbances and/or uncertainties, we impose the next assumption:

Assumption 1: There exist nonnegative scalar

func-tions    ,     such that                 , where  

describes the possi-ble values of

and 

denotes the

th entry of a vector.

Following the SMC design procedure [11], we select

              sgn  (8)

where diag       with

 

 for    , sgndenotes the sign

func-tion, and sgn sgn  sgn  

. Under the control , it follows from (5)-(7) and Assumption 1 that         

. This inequality implies that the

sys-tem states will reach the sliding surface in a finite amoun-t of amoun-time [11] no maamoun-tamoun-ter wheamoun-ther amoun-the acamoun-tuaamoun-tors in are

healthy or not.

In addition to the design of as discussed above, we

now investigate the design of

 to promote the overall

system performance when some or all of the actuators

in are healthy. From (5)-(6) and (8), we have

             

. Clearly, one of the choices

of

to make system states approach the sliding surface

faster than in the case of  is     sgn     (9) where   diag     and   for

all   . These derivations show that the

mag-nitude of control gains 

 , for the

actua-tors in the setthat guarantee stabilization performance

may vary from to the allowable maximum control input

magnitude. That is, it allows the situation of actuators in

to be total failure, partial failure, attenuation or

ampli-fication in any order and in any combination. From the derivations above, we then have the following result.

Theorem 1: Suppose that Assumption 1 holds.

Then the origin of System (2) is locally asymptotically stable under the control law given by (8) and (9) even when some or all of the actuators inexperience

abnor-mal operation.

4. APPLICATION TO SPACECRAFT

ATTITUDE STABILIZATION

An attitude model for a spacecraft in a circular orbit

can be described in the form of (1) with   [11].

The six state variables denote the three Euler’s angles

 and their derivatives. For simplicity, we

as-sume in this study that the thruster is the only applied control force and there is an actuator redundancy to per-form the reliable task. By letting  

 and            

, the overall system dynamics are described as follows [11]:

                                                   

(14)

                                                                                                                                                                                                                                                                                                                                 Here,  ,  , and 

are the inertia with respect to the

three body coordinate axes,

denotes the constant

or-bital rate, andanddenote theandfunctions,

respectively.

To derive an appropriate T-S model to approximate the original nonlinear dynamics, we first express    . A set of entries of  have the following

form-s:                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                          where      denotes the

-entry of the matrix

 . Next, a set of operating points will be

select-ed for the construction of the associatselect-ed linear model-s. These operating points are selected from the possible workspace, so that the motion of the spacecraft can be well approximated by using a convex combination of the associated linear models. For demonstration, we assume that        ,       ,       

!", and the angular positions

are constrained to be  #" #" , 

 ##,

and



 #" #" . To investigate the effects of the

number of premise variables, we consider the following two cases: The first in which the three angles are chosen as the premise variables, while in the second case all six states are included.

(15)

4.1 Case for Three Premise Variables

In this case, the operating points are chosen in

the form of                            $     , where    ,        and      

are three selected partitions of#" #" , ##, and#" #" , respectively. In this case, we

select   

  



 and employ the triangular

membership functions. Under these settings, we have

 

  operating points. The associated 125 linear

models can then be easily obtained. Two of them are list-ed below:                               where      

. After determining the 

lin-ear models, the T-S fuzzy system model can be easily determined when the angular positions of the spacecraft

are available. Define the region %

                            

 & . The upper bounds

of       Ü 

  over the

re-gion%

  can be computed off-line, and it is found that

the maximum value of

  

among all of the

re-gions %   is         . Since

the T-S type controller only uses three premise variables with triangular membership functions, it therefore trig-gers at most eight rules (i.e., at most



linear models) at each time instant. Thus, it does not create an extra on-line computational burden if the partition for the region-s of , 

 and 

 are made finer. However, since the

maximum value of a function over a smaller sub-region is smaller than or equal to that of the same function over the whole region, it follows that a finer partition for the region of ,

 and 

 will result in a smaller

magni-tude of 

 as stated in Assumption 1. Thus, the

con-trol magnitude will be smaller so that the physical concon-trol magnitude constraint is easier to fulfill for practical appli-cations if the partition of ,

and 

are made finer.

4.2 Case for Six Premise Variables

The operating points in this case are chosen in the for-m of                              and 

are positive integers for  

  . In this example, we select



 for  

 and also employ the triangular membership

func-tions. Under these settings, we have



 

operat-ing points and linear models which are determined from the relation

       

      .

De-tails are omitted. The T-S model can then be easily deter-mined when all of the angular positions and velocities of the spacecraft are available. Since the T-S type controller

uses six premise variables for this case, it triggers 64 rules (i.e.,



linear models) at each time instant. Furthermore, it does not create an extra on-line computational burden if the partition for the regions of the system states is made finer, as seen in the previous case. Moreover, it is found that



 , which can be computed off-line.

This implies that the difference between the T-S model and the original dynamics for this case is much small-er than that of Case A, though this case consumes more time (since it triggers 64 rules at every time instant) to evaluate the T-S model than that of Case A (only triggers 8 rules at each time instant).

Numerical results are summarized in Figs. 1-3. A-mong these, we use the following three control schemes: One is the SMC reliable design [11] (labeled by SMC), and the other two are the T-S model-based SMC reliable scheme with a different number of premise variables as s-tated in Cases 4.1 and 4.2 above (Labeled by Case A and Case B, respectively). The parameters of these SMC

reli-able designs are set to be 

,    for all in and ,        ,          ,   for

all, and the sign function is replaced by the saturation

function with a boundary layer width of   to

allevi-ate the chattering produced by the sign function. In ad-dition, we select

 as the susceptible actuator, that is,       and   

, and assume that 

fails at . It is observed from Fig. 1 that the

stabiliza-tion performance is, as expected, achieved for all of the three control schemes. However, since the T-S model for Case B is very close to the original nonlinear model, the state curves, the sliding variables and the control curves for Case B and the SMC reliable design are also very close to each other, which can be recognized from Figs.

1-3. By direct calculation, the consumed energy and

the quadratic performance have the following relations:

                and                .

Clear-ly, the two performance indices

  and  of the two control schemes SMC and Case B are found to be close to each other, while Case A consumes more energy and experiences a larger value of



than the other

two schemes. It is worth noting from Fig. 3(b) that



fails after  and changes sign around  for

all of the three schemes. The sign change of

is

ver-ified by the sign change of 

, which agrees with Eq.

(9). Furthermore, owing to the use of the saturation func-tion for simulafunc-tion, the magnitude of

 for Case A is

seen to be less than rather than equal to  during

the time period   , which does not contradict

Eq. (9) when  

'. Finally, when repeatedly

com-puting the controllers

times, the T-S type design (including the determination of membership weightings) consumes less CPU time than the classic SMC design in the relation of!"#    !"#   !"#

 . Based upon these

數據

Fig. 1 Time history of the six system states.

參考文獻

相關文件

 name common laboratory apparatus (e.g., beaker, test tube, test-tube rack, glass rod, dropper, spatula, measuring cylinder, Bunsen burner, tripod, wire gauze and heat-proof

專案執 行團隊

Microphone and 600 ohm line conduits shall be mechanically and electrically connected to receptacle boxes and electrically grounded to the audio system ground point.. Lines in

• A delta-gamma hedge is a delta hedge that maintains zero portfolio gamma; it is gamma neutral.. • To meet this extra condition, one more security needs to be

了⼀一個方案,用以尋找滿足 Calabi 方程的空 間,這些空間現在通稱為 Calabi-Yau 空間。.

We do it by reducing the first order system to a vectorial Schr¨ odinger type equation containing conductivity coefficient in matrix potential coefficient as in [3], [13] and use

 Promote project learning, mathematical modeling, and problem-based learning to strengthen the ability to integrate and apply knowledge and skills, and make. calculated

• Examples of items NOT recognised for fee calculation*: staff gathering/ welfare/ meal allowances, expenses related to event celebrations without student participation,