外文翻譯現(xiàn)代控制理論的發(fā)展

1、<p><b>　　現(xiàn)代控制理論的發(fā)展</b></p><p>　　1.智能控制(Intelligent Control) </p><p>　　智能控制是人工智能和自動控制的結(jié)合物,是一類無需人的干預(yù)就能夠獨立地驅(qū)動智能機器,實現(xiàn)其目標的自動控制。智能控制的注意力并不放在對數(shù)學(xué)公式的表達、計算和處理上,而放在對任務(wù)和模型的描述,符號和環(huán)境的識別以及知識庫和

2、推理機的設(shè)計開發(fā)上。智能控制用于生產(chǎn)過程,讓計算機系統(tǒng)模仿專家或熟練操作人員的經(jīng)驗,建立起以知識為基礎(chǔ)的廣義模型,采用符號信息處理、啟發(fā)式程序設(shè)計、知識表示和自學(xué)習(xí)、推理與決策等智能化技術(shù),對外界環(huán)境和系統(tǒng)過程進行理解、判斷、預(yù)測和規(guī)劃,使被控對象按一定要求達到預(yù)定的目的。 </p><p>　　智能控制的理論基礎(chǔ)是人工智能,控制論,運籌學(xué)和系統(tǒng)學(xué)等學(xué)科的交叉,它的主要特點是: </p><p

3、>　　(1)同時具有以知識表示的非數(shù)學(xué)廣義模型和以數(shù)學(xué)模型表示的混合控制過程; </p><p>　　(2)智能控制的核心在高層控制,即組織級,它的主要任務(wù)在于對實際環(huán)境或過程進行組織; </p><p>　　(3)系統(tǒng)獲取的信息不僅是數(shù)學(xué)信息,更重要的是文字符號、圖像、圖形、聲音等各種信息。 </p><p>　　智能控制正處于發(fā)展過程中,還存在許多有待研究

4、的問題: </p><p>　　(1)探討新的智能控制理論; </p><p>　　(2)采用語音控制; </p><p>　　(3)提高系統(tǒng)的學(xué)習(xí)能力和自主能力; </p><p>　　(4)利用現(xiàn)有的非線性技術(shù)分析閉環(huán)系統(tǒng)的特性; </p><p>　　(5)智能控制的實現(xiàn)問題。 </p><p

5、>　　2.非線性控制(Nonlinear Control) </p><p>　　非線性控制是復(fù)雜控制理論中一個重要的基本問題,也是一個難點課題,它的發(fā)展幾乎與線性系統(tǒng)平行。非線性系統(tǒng)的發(fā)展,數(shù)學(xué)工具是一個相當(dāng)困難的問題,泰勒級數(shù)展開對有些情況是不能適用的。古典理論中的“相平面”法只適用于二階系統(tǒng),適用于含有一個非線性元件的高階系統(tǒng)的“描述函數(shù)”法也是一種近似方法。由于非線性系統(tǒng)的研究缺乏系統(tǒng)的、一般性的理

6、論及方法,于是綜合方法得到較大的發(fā)展,主要有: </p><p>　　(1)李雅普諾夫方法:它是迄今為止最完善、最一般的非線性方法,但是由于它的一般性,在用來分析穩(wěn)定性或用來鎮(zhèn)定綜合時都欠缺構(gòu)造性。 </p><p>　　(2)變結(jié)構(gòu)控制:由于其滑動模態(tài)具有對干擾與攝動的不變性,到80年代受到重視,是一種實用的非線性控制的綜合方法。 </p><p>　　(3)微分

7、幾何法:在過去的的20年中,微分幾何法一直是非線性控制系統(tǒng)研究的主流,它對非線性系統(tǒng)的結(jié)構(gòu)分析、分解以及與結(jié)構(gòu)有關(guān)的控制設(shè)計帶來極大方便.用微分幾何法研究非線性系統(tǒng)是現(xiàn)代數(shù)學(xué)發(fā)展的必然產(chǎn)物,正如意大利教授Isidori指出:“用微分幾何法研究非線性系統(tǒng)所取得的成績,就象50年代用拉氏變換及復(fù)變函數(shù)理論對單輸入單輸出系統(tǒng)的研究,或用線性代數(shù)對多變量系統(tǒng)的研究?！钡@種方法也有它的缺點,體現(xiàn)在它的復(fù)雜性、無層次性、準線性控制以及空間測度被破

8、壞等。因此最近又有學(xué)者提出引入新的、更深刻的數(shù)學(xué)工具去開拓新的方向,例如:微分動力學(xué)、微分拓撲與代數(shù)拓撲、代數(shù)幾何等。 </p><p>　　3.自適應(yīng)控制(Adaptive Control) </p><p>　　自適應(yīng)控制系統(tǒng)通過不斷地測量系統(tǒng)的輸入、狀態(tài)、輸出或性能參數(shù),逐漸了解和掌握對象,然后根據(jù)所得的信息按一定的設(shè)計方法,作出決策去更新控制器的結(jié)構(gòu)和參數(shù)以適應(yīng)環(huán)境的變化,達到所要

9、求的控制性能指標。 </p><p>　　自適應(yīng)控制系統(tǒng)應(yīng)具有三個基本功能: </p><p>　　(1)辨識對象的結(jié)構(gòu)和參數(shù),以便精確地建立被控對象的數(shù)學(xué)模型; </p><p>　　(2)給出一種控制律以使被控系統(tǒng)達到期望的性能指標; </p><p>　　(3)自動修正控制器的參數(shù)。因此自適應(yīng)控制系統(tǒng)主要用于過程模型未知或過程模型結(jié)構(gòu)已

10、知但參數(shù)未知且隨機的系統(tǒng)。 </p><p>　　自適應(yīng)控制系統(tǒng)的類型主要有自校正控制系統(tǒng),模型參考自適應(yīng)控制系統(tǒng),自尋最優(yōu)控制系統(tǒng),學(xué)習(xí)控制系統(tǒng)等。最近,非線性系統(tǒng)的自適應(yīng)控制,基于神經(jīng)網(wǎng)絡(luò)的自適應(yīng)控制又得到重視,提出一些新的方法。 </p><p>　　4.魯棒控制(Robust Control) </p><p>　　過程控制中面臨的一個重要問題就是模型不確定

11、性,魯棒控制主要解決模型的不確定性問題,但在處理方法上與自適應(yīng)控制有所不同。自適應(yīng)控制的基本思想是進行模型參數(shù)的辯識,進而設(shè)計控制器。控制器參數(shù)的調(diào)整依賴于模型參數(shù)的更新,不能預(yù)先把可能出現(xiàn)的不確定性考慮進去。而魯棒控制在設(shè)計控制器時盡量利用不確定性信息來設(shè)計一個控制器,使得不確定參數(shù)出現(xiàn)時仍能滿足性能指標要求。 </p><p>　　魯棒控制認為系統(tǒng)的不確定性可用模型集來描述,系統(tǒng)的模型并不唯一,可以是模型集里

12、的任一元素,但在所設(shè)計的控制器下,都能使模型集里的元素滿足要求。魯棒控制的一個主要問題就是魯棒穩(wěn)定性,目前常用的有三種方法: </p><p>　　(1)當(dāng)被研究的系統(tǒng)用狀態(tài)矩陣或特征多項式描述時一般采用代數(shù)方法,其中心問題是討論多項式或矩陣組的穩(wěn)定性問題; </p><p>　　(2)李雅普諾夫方法,對不確定性以狀態(tài)空間模式出現(xiàn)時是一種有利工具; </p><p>

13、;　　(3)頻域法從傳遞函數(shù)出發(fā)研究問題,有代表性的是Hoo控制,它用作魯棒性分析的有效性體現(xiàn)在外部擾動不再假設(shè)為固定的,而只要求能量有界即可。這種方法已被用于工程設(shè)計中,如Hoo最優(yōu)靈敏度控制器設(shè)計。 </p><p>　　5.模糊控制(Fuzzy Control) </p><p>　　模糊控制借助模糊數(shù)學(xué)模擬人的思維方法,將工藝操作人員的經(jīng)驗加以總結(jié),運用語言變量和模糊邏輯理論進行推

14、理和決策,對復(fù)雜對象進行控制。模糊控制既不是指被控過程是模糊的,也不意味控制器是不確定的,它是表示知識和概念上的模糊性,它完成的工作是完全確定的。 </p><p>　　1974年英國工程師E.H.Mamdam首次把Fuzzy集合理論用于鍋爐和蒸氣機的控制以來,開辟了Fuzzy控制的新領(lǐng)域,特別是對于大時滯、非線性等難以建立精確數(shù)學(xué)模型的復(fù)雜系統(tǒng),通過計算機實現(xiàn)模糊控制往往能取得很好的結(jié)果。 </p>

15、<p>　　模糊控制的類型有: </p><p>　　(1)基本模糊控制器,一旦模糊控制表確定之后,控制規(guī)則就固定不變了; </p><p>　　(2)自適應(yīng)模糊控制器,在運行中自動修改、完善和調(diào)整規(guī)則,使被控過程的控制效果不斷提高,達到預(yù)期的效果; </p><p>　　(3)智能模糊控制器,它把人、人工智能和神經(jīng)網(wǎng)絡(luò)三者聯(lián)系起來,實現(xiàn)綜合信息處理,

16、使系統(tǒng)既具有靈活的推理機制、啟發(fā)性知識與產(chǎn)生式規(guī)則表示,又具有多種層次、多種類型的控制規(guī)律選擇。 </p><p>　　模糊控制的特點是不需要精確的數(shù)學(xué)模型,魯棒性強,控制效果好,容易克服非線性因素的影響,控制方法易于掌握。最近有人提出神經(jīng)——模糊Inter3融合控制模型,即把融合結(jié)構(gòu)、融合算法及控制合為一體進行設(shè)計。又有人提出利用同倫BP網(wǎng)絡(luò)記憶模糊規(guī)則,以“聯(lián)想方式”使用這些經(jīng)驗。 </p>&

17、lt;p>　　模糊控制有待進一步研究的問題:模糊控制系統(tǒng)的功能、穩(wěn)定性、最優(yōu)化問題的評價;非線性復(fù)雜系統(tǒng)的模糊建模,模糊規(guī)則的建立和模糊推理算法的研究;找出可遵循的一般設(shè)計原則。 </p><p>　　6.神經(jīng)網(wǎng)絡(luò)控制(Neural Network Control) </p><p>　　神經(jīng)網(wǎng)絡(luò)是由所謂神經(jīng)元的簡單單元按并行結(jié)構(gòu)經(jīng)過可調(diào)的連接權(quán)構(gòu)成的網(wǎng)絡(luò)。神經(jīng)網(wǎng)絡(luò)的種類很多,控制中

18、常用的有多層前向BP網(wǎng)絡(luò),RBF網(wǎng)絡(luò),Hopfield網(wǎng)絡(luò)以及自適應(yīng)共振理論模型(ART)等。 </p><p>　　神經(jīng)網(wǎng)絡(luò)控制就是利用神經(jīng)網(wǎng)絡(luò)這種工具從機理上對人腦進行簡單結(jié)構(gòu)模擬的新型控制和辨識方法。神經(jīng)網(wǎng)絡(luò)在控制系統(tǒng)中可充當(dāng)對象的模型,還可充當(dāng)控制器。常見的神經(jīng)網(wǎng)絡(luò)控制結(jié)構(gòu)有: </p><p>　　(1)參數(shù)估計自適應(yīng)控制系統(tǒng); </p><p>　　(2

19、)內(nèi)?？刂葡到y(tǒng); </p><p>　　(3)預(yù)測控制系統(tǒng); </p><p>　　(4)模型參考自適應(yīng)系統(tǒng); </p><p>　　(5)變結(jié)構(gòu)控制系統(tǒng)。 </p><p>　　神經(jīng)網(wǎng)絡(luò)控制的主要特點是:可以描述任意非線性系統(tǒng);用于非線性系統(tǒng)的辨識和估計;對于復(fù)雜不確定性問題具有自適應(yīng)能力;快速優(yōu)化計算能力;具有分布式儲存能力,可實現(xiàn)在線、

20、離線學(xué)習(xí)。 </p><p>　　最近有人提出以Hopfield網(wǎng)絡(luò)實現(xiàn)一種多分辨率體視協(xié)同算法,該算法以逐級融合的方式自動完成由粗到細,直至全分辨率的匹配和建立。又有人提出一種網(wǎng)絡(luò)自組織控制器,采用變斜率的最速梯度下降學(xué)習(xí)算法,應(yīng)用在非線性跟蹤控制中。今后需進一步探討的問題是提高網(wǎng)絡(luò)的學(xué)習(xí)速度,提出新的網(wǎng)絡(luò)結(jié)構(gòu),創(chuàng)造出更適用于控制的專用神經(jīng)網(wǎng)絡(luò)。</p><p><b>　　外

21、文資料翻譯</b></p><p>　　The development of modern control theory</p><p>　　1. The Intelligent Control (Intelligent Control)</p><p>　　Intelligent control is the artificial intelligen

22、ce and automatic control combined, is a kind of people without intervention can be independently drive intelligent machine, realize the goal of automatic control. Intelligent control attention is not on the expression of

23、 mathematical formula, calculation and processing, and on to the task and the description of the model, symbols and environment of the recognition and the knowledge base and reasoning machine design and development. Inte

24、lligent cont</p><p>　　Intelligent control is based on the theory of artificial intelligence, cybernetics, operations research and systematics of subjects such as cross, its main features are:</p><

25、p>　　(1) And has knowledge of mathematical model of generalized said and mathematical model of the said in hybrid control process;</p><p>　　(2) The core of intelligent control in high-rise control, namely

26、the organization level, its main task is to the actual environment or processes organizations;</p><p>　　(3) System getting information is not only the mathematical information, more important is to text symb

27、ols, images, graphics, sounds, etc, all kinds of information.</p><p>　　Intelligent control is in the process of development, there still exist many problems of further researches:</p><p>　　(1) E

28、xplore a new intelligent control theory;</p><p>　　(2) Using voice control;</p><p>　　(3) Improve the system of independent learning ability and ability;</p><p>　　(4) Use of the exist

29、ing nonlinear technical analysis of the closed-loop system characteristics;</p><p>　　(5) Of the intelligent control system.</p><p>　　2. Nonlinear Control (Nonlinear Control)</p><p>

30、　　Nonlinear control is the theory of complex control of an important basic problem, but also a difficult subject, the development of linear system with almost parallel. The development of the nonlinear system, mathematic

31、al tools is a very difficult question, Taylor series development for some situation is not applicable. The classical theory of the "phase plane" law only applies to the second order systems, apply to contain a

32、non-linear element of high order system "describing function" method is </p><p>　　(1) The lyapunov method: it is by far the most perfect, the most general nonlinear method, but because of its gener

33、al, used to analysis the stability or used to calm when lack structure are comprehensive sex.</p><p>　　(2) Variable structure control: because its sliding mode has to interfere with the perturbation, j, to t

34、he 80's attention, is a kind of practical non-linear control of the synthesis method.</p><p>　　(3) Differential geometry method: in the past 20 years, differential geometry method, nonlinear control syst

35、ems research has been the mainstream, it to nonlinear system structure analysis, decomposition, and structure of the relevant control design bring great convenience. Use the method of differential geometry nonlinear syst

36、em is the inevitable outcome of the development of modern mathematics, as Italy isidori professor pointed out: "with differential geometry nonlinear system method to study </p><p>　　3. Adaptive Control

37、(Adaptive Control)</p><p>　　Adaptive control system by constantly measuring system of input, the state, output or performance parameters, and gradually to understand and grasp object, and then according to t

38、he information obtained according to certain design method, make decisions to update the controller's structure and parameters to adapt to the change of environment, to reach the required control performance indicato

39、rs. Adaptive control system should have three basic functions:</p><p>　　(1) Identification of the object structure and parameters so as to accurately established the mathematical model of the controlled obje

40、cts;</p><p>　　(2) Given a control law in order to make the controlled system reach the expected performance index;</p><p>　　(3) To be automatic correction controller parameters. So the adaptive

41、control system is mainly used to process model unknown or process model structure known but unknown parameters and the random system.</p><p>　　Adaptive control system of main type is used for the control sys

42、tem, the model reference adaptive control system, the optimal control system for, learn to control system, etc. Recently, nonlinear adaptive control system based on neural network adaptive control and get attention, and

43、put forward some new methods.</p><p>　　4. Robust Control (Robust Control)</p><p>　　Process control is an important issue facing is model uncertainty, robust control mainly to solve the model unc

44、ertainty problem, but in processing method and adaptive control different. Adaptive control of basic idea is to carry out the model parameter of identify, and then to design the controller. Controller parameters adjustme

45、nt depends on the renewal of the model parameters, not in possible uncertainty into consideration. And robust control in design as far as possible when the controller us</p><p>　　Robust control think of the

46、uncertainty of the system can be used to describe the model collection system, the model is not only, can be a model of any one of the elements in the collection, but in the controller design, can make the model collecti

47、on of elements meet the requirements. Robust control of a major problem is robust stability, now commonly used in three ways:</p><p>　　(1) When the system is studied with state matrix or features polynomial

48、algebra in description generally USES the method, the center is to discuss the problem or matrix of the stability of polynomial problem;</p><p>　　(2) Iyapunov method of uncertainty to state space model appea

49、r is a useful tool; (3) frequency domain method from the transfer function of the research question, typically Hoo control, it used as a robust effectiveness of analysis reflected in external disturbance that no longer f

50、or fixed, and only request energy bounded can. This method has been used in engineering design, such as Hoo optimal sensitivity controller design.</p><p>　　5. Fuzzy Control (Fuzzy Control)</p><p&g

51、t;　　Fuzzy control with fuzzy mathematical simulation of people thinking method, the process of the operator experience summarized, use the language variables and fuzzy logic theory reasoning and decision-making, complex

52、object to control. Fuzzy control is not mean to be accused of process is fuzzy, that doesn't mean the controller is not set, it is the concept of knowledge and said the fuzziness, it completed work is entirely sure.&

53、lt;/p><p>　　In 1974 the British engineer for the first time E.H.M amdam the Fuzzy set theory for the boiler and steam machine control since, opened up a new field of Fuzzy control, especially for large delay an

54、d nonlinearity is hard to develop practical mathematical model of complex system, through the computer to realize Fuzzy control often can obtain very good results.</p><p>　　The type of fuzzy control are:<

55、/p><p>　　(1) Basic fuzzy controller, once the fuzzy control table was determined, control rule is fixed; (2) the adaptive fuzzy controller, in operation of automatic modification, perfect and adjust the rules,

56、that was accused of process control effect is improved, and achieve the desired effect;</p><p>　　(3) Intelligent fuzzy controller, which, artificial intelligence and neural network linked to three, to realiz

57、e comprehensive information processing, make the system not only has the flexible reasoning mechanism, heuristic knowledge and production rules said, and has many layers, various types of control laws of choice.</p>

58、;<p>　　The characteristics of the fuzzy control is not to need to accurate mathematical model, robust, the control effect is good, easy to overcome the influence of nonlinear factors, the control method is easy to

59、 grasp. Recently, someone put forward fuzzy neural-Inter3 fusion control model, that is, the fusion structure, fusion algorithm and control is an organic whole to carry on the design. Again someone proposed by homotopy B

60、P network memory fuzzy rules to "lenovo way" use these experience. Fuzzy c</p><p>　　6. The Neural Network Control (Neural Network Control)</p><p>　　Neural network is the simple unit by

61、 so-called neurons in parallel structure after adjustable connection power consists of the network. There are many kinds of neural network, control the commonly used have multilayer feedforward BP network, the RBF networ

62、k, the Hopfield network.the computer and adaptive resonance theory model (ART), etc.</p><p>　　The neural network control by using neural network is the tool from mechanism on the simple structure of the new

63、simulation control and identification methods. Neural network in the control system can serve as a object model, also can serve as a controller. Common neural network control structure are:</p><p>　　(1) Para

64、meter estimation adaptive control system;</p><p>　　(2) Internal model control system;</p><p>　　(3) Predictive control system;</p><p>　　(4) Model reference adaptive system;</p>

65、<p>　　(5) Variable structure control systems.</p><p>　　The neural network control of key features are: can describe any nonlinear system; Used for nonlinear system identification and estimate; For a c

66、omplicated uncertainty of the adaptive; Quick optimization calculation capacity; With a distributed storage capacity, which can realize the online and off-line learning.</p><p>　　Recently, someone to put for

67、ward the Hopfield network.the computer implementation of a multiresolution stered coordination algorithm, based on the level of the fusion of automatic way finish by coarse to fine until the full resolution matching and

68、established. Again someone put forward a kind of self-organization network controller, the change of the steepest slope gradient descent learning algorithm, the application of the nonlinear tracking control in. In the fu