采煤機相關(guān)英文文獻翻譯

1、<p><b>　　英文原文：</b></p><p>　　Control strategy for an intelligent shearer height adjusting system</p><p>　　FAN Qigao*, LI Wei, WANG Yuqiao, ZHOU Lijuan, YANG Xuefeng, YE Guo Schoo

2、l of Mechanical & Electrical Engineering, China University of Mining & Technology, Xuzhou , China</p><p>　　Abstract: An intelligent shearer height adjusting system is a key technology for mining at a

3、 man-less working face. A control strategy for a shearer height adjusting system based on a mathematical model of the height adjusting mechanism is proposed. It considers the non-linearity and time variations in the cont

4、rol process and uses Dynamic Fuzzy Neural Networks (D-FNN). The inverse characteristics of the system are studied. An adaptive on-line learning and error compensation mechanism guarantees </p><p>　　Keywords:

5、 shearer; height adjusting system; dynamic fuzzy neural network</p><p>　　1 Introduction </p><p>　　The shearer and its control system are main components for coal mining. The shearing process in

6、cludes drum lifting and traction control. Domestic shear drum lifting now uses manual adjustments after artificial observation or a geometric track cutting-memory method after trial manual adjustments from test cuttings.

7、 The installation of sensors on the shearer that could identify coal-rock has been proposed. Information from the sensors would be used to achieve drum height control directly by automa</p><p>　　achieve self

8、-adaptive learning and control. Setting up such a system involves considerable uncertainty and a great deal of time.</p><p>　　Considering the factors and the need for improving product quality and resource r

9、ecovery by automatic control of the drum height we propose a new method called the shearer intelligent height adjusting system control method. It is based on Dynamic Fuzzy Neural Networks (D-FNN). D-FNN are neural networ

10、ks that have the characteristics of powerful on-line learning, fast learning and good generalization. D-FNN give real-time control and improve dynamic characteristics of a shearer height adjusting sy</p><p>

11、　　2 Analysis of a shearer height adjusting system </p><p>　　2.1 Structure of the shearer height adjusting system </p><p>　　The shearer height adjusting mechanism uses a hydraulic servo system

12、having good dynamic performance. Fig. 1 diagrams a drum shearer. The electro-hydraulic servo system controls extension of the hydraulic cylinder and moves the rocker arm to set the height. The adjusting mechanism is a pl

13、anar open chain consisting of a series of connected rod structures and corresponding kinematic pairs. A descripion of the relative motion of the parts shows how height adjustment occurs. A detailed motion analys</p>

14、;<p>　　1) All components are rigid and elastic deformation is ignored; </p><p>　　2) Gaps between all mechanisms are ignored. </p><p>　　2.2 Mathematical analysis of the shearer height ad

15、justment system</p><p>　　Fig. 2 shows the initial position of the hydraulic cylinder as , the end position as , the long arm of the rocker arm is L, short arm is , the draw bar between the height adjustment

16、 cylinder and the rocker arm is , the distance between the height adjustment cylinder and the rocker pivot is D and the angle between the long arm and the short arm is. </p><p>　　Definition 1. Shearer mining

17、 height H: </p><p>　　H=L (1)</p><p>　　End position is given by allowing the displacement of the hydraulic cylinder, , to be established.

18、 </p><p>　　Definition 2. Displacement of the hydraulic cylinder, , is: </p><p><b>　　(2)</b></p><p><b>　　where </b></p><p>　　We write: </p&g

19、t;<p><b>　　(3)</b></p><p><b>　　where </b></p><p>　　Substitution gives as: </p><p><b>　　(4)</b></p><p>　　Since b is given

20、by can be expressed as a function of rocker-height to angle: </p><p><b>　　(5)</b></p><p>　　Kinetic analysis of the model shearer height adjusting system shows it is a third order

21、 system. The system transfer function is: </p><p><b>　　(6)</b></p><p>　　where K is the system gain, ζ is the system damping ratio, w is the natural frequency of the system, F (s)

22、 the Laplace transform of the servo mechanism, the Laplace transform of (in Eq.(5)), is derived from Eq.(6), the swing angle, θ , of the rocker arm is from Eq.(5) and θ controls the feedback. </p><p

23、>　　Since the height adjusting system is non-linear and a time-varying dynamic system a traditional PID controller cannot provide satisfactory control. D-FNN are proposed as meeting the requirements of reliability and

24、real time performance. </p><p>　　3 Dynamic fuzzy neural networks </p><p>　　D-FNN are based on the expansion of Radial Basis Function (RBF) neural networks. The prominent characteristics of th

25、is learning algorithm are the simultaneous adjustment of parameters and the identification of an appropriate structure. This provides rapid learning suitable for real-time control and for modeling of the shearer height a

26、djusting system The structure of a dynamic fuzzy neural network is shown in Fig. 3. </p><p>　　In Fig. 3 , , …, are the system input variables, y is the system output, is the membership function, j, of the

27、 input variable, i, is the fuzzy rule of membership function j, is the normalized node of j, is the connection weight of rule j and u is the whole system rule number. </p><p>　　The swing angle, θ , of

28、the rocker arm was chosen as the system input variable that controls expansion of the hydraulic cylinder. A Gaussian function, Eq.(7), is used for the membership function. </p><p><b>　　(7)</b><

29、;/p><p>　　where i ranges from 1 to r, j ranges from 1 to u, is the membership function, j, of , is the center of the Gaussian membership function, j, of , is the width of the Gaussian membership function, j,

30、 of , r is the input variable number and u is the number of the membership function as well as the whole system rule number. </p><p>　　The output of , rule j, is obtained from: </p><p><b>

31、　　(8)</b></p><p>　　where X is given by: and the center of RBF neural network j is given by: </p><p>　　This gives the D-FNN model as:</p><p><b>　　(9)</b><

32、/p><p>　　where α is the connection weight of rule i. </p><p>　　4 D-FNN control strategy </p><p>　　The D-FNN control scheme is shown in Fig. 4. The basic idea is obtaining the invers

33、e characteristic of the shearer height adjusting system and then producing a compensation signal from this inverse dynamic model. There are two dynamic fuzzy neural networks here: A and B. Network A is for system weight

34、training while network B is a copy of the trained A network that is used for producing the control signal. </p><p>　　The control algorithm is: </p><p><b>　　(10)</b></p><p&

35、gt;　　where x Δ is the expected displacement of the height adjusting hydraulic cylinder; PD Δx the actual displacement of the cylinder produced by the PD controller and DFNNB Δx the actual displacement of the cylin

36、der produced by network B. </p><p>　　The PD controller is for faster and more accurate tracking performance. The key to the D-FNN control </p><p>　　system is the training of D-FNN B to minimize

37、the squared error between expected and actual displacements produced by network B:</p><p><b>　　(11)</b></p><p>　　A gradient descent method is used for the weight adjusting algorithm:

38、 </p><p><b>　　(12)</b></p><p>　　where λ is the learning rate and λ >0. λ has a large influence on the convergence rate. Increasing of λcan speed up the convergence rate, whic

39、h is more suitable for time-varying system modeling and control. At the same time the anti-interference performance of the system declines. A decrease in λ slows down convergence but produces a system less sensitive to

40、 interference. A self-adjusting learning rate method is proposed herein, the principle being that when the new error exceeds the last</p><p><b>　　(13)</b></p><p>　　Tests show that D-

41、FNN using the self-adjusting learning rate method requires much less training time than systems using a fixed learning rate.</p><p>　　5 System simulation </p><p>　　The mathematical model and a

42、D-FNN control algorithm may be used in a model shearer height ad-</p><p>　　justing system built using Matlab/Simulink[. The actual parameters are from a German Eickhoff SL500 machine. The shearer maximum cut

43、ting height is 5.50 m and the foot wall is 1.08 m. The angle of the rocker arm is –21.3°~+55°. The draw bar, LG, is 2.05 m, the short arm, LR, is 1.20 m, D is 0.9 m and the angle </p><p>　　5.1 Sim

44、ulation of a D-FNN controller </p><p>　　Suppose the rocker arm moves within a range of –21.3°~+55°. The D-FNN control strategy traces the trajectory of the rocker arm and the trajectory tracing err

45、or are shown in Fig. 5. In Fig. 5b the maximum trajectory tracing error of the rocker arm is 0.65°, which occurs early in the training stage. At this point the D-FNN is undergoing on-line learning, namely learning

46、the proper inverse model of the shearer height adjusting system. So in the early stage network B has insufficient accuracy to co</p><p>　　5.2 Simulation of a PID controller</p><p>　　The traject

47、ory of the rocker arm, and the corresponding tracing error, are shown in Fig. 6 for the traditional PID controller. </p><p>　　As shown in Fig. 6b, the maximum trajectory error is 5.8°; this is unaccepta

48、ble for the whole system. The simulation results show that the D-FNN controller is more robust and adjusts faster. </p><p>　　6 Conclusions </p><p>　　1) A mathematical analysis of the shearer he

49、ight adjusting structure was used to build a mathematical model. The constraints between the control and feedback variables of the shearer height adjusting system were determined from the model. </p><p>　　2)

50、 The combined advantages of fuzzy control and neural network control used in the D-FNN control strategy to adjust shearer height were described. A proposed control scheme of the system, having the desired inverse charact

51、eristic, is derived. By adjusting the weights and compensating for accuracy the control scheme satisfactorily met the needs of a height adjusting system.</p><p>　　3) A simulated D-FNN controller system using

52、 parameters from an Eickhoff SL500 shearer was compared to a traditional PID controller: the D-FNN controller was more accurate. The D-FNN algorithm overcomes limitations of traditional network optimization algorithms an

53、d avoids falling into local minimum points. Self adaptive, on-line learning greatly improves the training speed. The system stability and accuracy meet the requirements for a shearer height adjusting system</p>&l

54、t;p>　　Acknowledgements </p><p>　　Financial support for this work, provided by the National High Technology Research and Development Program of China (No.2008AA), and China University of Mining & Techn

55、ology Scaling Program, are gratefully acknowledged. </p><p>　　References </p><p>　　[1] Zhang J M, Fan X, Zhao X S. Automatic horizon control system of coal mining machine. Journal of China Uni

56、versity of Mining & Technology, 2002, 31(4): 415-418. (In Chinese) </p><p>　　[2] Liang Y W, Xiong S B. Neuarl network and PID hybrid adaptive control for horizontal control of shearer. In: Proceeding of

57、 the 7th International Conference on Control,Automation, Robotics and Vision IEEE. Singapore, 2002: 671-674. (In Chinese) </p><p>　　[3] Lei Y Y, Yin Z X, Qian H. Study on hydraulic automatic ranging cutting

58、 height of shearer. Journal of Chongqing University, 1994, 17(1): 52-58. (In Chinese) </p><p>　　[4] Er M J, Wu S Q. A fast learning algorithm for parsimonious fuzzy neural systems. Fuzzy Sets and Systems,

59、2002, 126(3): 337-351. </p><p>　　[5] Gao Y, Er M J, Yang S. Adaptive fuzzy neural control of robot manipulators. IEEE Trans Ind Electron, 2001, 48: 1274-1278. </p><p>　　[6] Chang Y C. Adaptiv

60、e fuzzy-based tracking control for nonlinear SISO systems via VSS and H approaches. IEEE Trans Fuzzy Syst, 2001(9): 278-292.</p><p>　　[7] Li C, Lee C Y. Self-organizing neuro-fuzzy system for control of unk

61、nown plants. IEEE Transactions on Fuzzy Systems, 2003, 11(1): 135-150. </p><p>　　[8] Er M J, Low C B, Nah K H, Lim M H, Ng S Y. Real-time implementation of a dynamic fuzzy neural networks controller for SC

62、ARA. Microprocessors and Microsystems, 2002, 26(9/10): 449-461. </p><p>　　[9] Juang C F, Lin C T. Noisy speech processing by recurrently adaptive fuzzy filters. IEEE Transactions on Fuzzy Systems, 2001, 9(1

63、): 139-152.</p><p>　　[10] Esposito A, Marinaro M, Oricchio D, Scarpetta S. Approximation of continuous and discontinuous mappings by a growing neural RBF-based algorithm. Neural Networks, 2000, 13(6): 651-6

64、65. </p><p>　　[11] Magee D P. Matlab extensions for the development, testing and verification of real-time DSP software. In: Proceedings of 42nd Annual Conf Design Automation. California, 2005: 603-606. <

65、/p><p>　　[12] Bhatt T M, McCain D. Matlab as a development environment for FPGA design. In: Proceedings of 42nd Annual Conf Design Automation. California, 2005: 607-610. </p><p>　　[13] Yang Y J, D

66、eng H Y, Li X. Simulation of screening process based on MATLAB/Simulink. Journal of China University of Mining & Technology, 2006, 16(3): 330- 332. </p><p>　　[14] Liu S Y, Du C L, Cui X X, Cheng X. Model

67、 test of the cutting properties of a shearer drum. Mining Science and Technology, 2009, 19(1): 74-78. </p><p>　　[15] Fang X Q, Zhao J J, Hu Y. Tests and error analysis of a self-positioning shearer operati

68、ng at a manless working face. Mining Science and Technology, 2010, 20(1): 53- 58. </p><p><b>　　中文翻譯： </b></p><p>　　采煤機高度智能調(diào)節(jié)系統(tǒng)控制方案</p><p>　　范啟高，周麗娟，李偉，王玉橋，楊學(xué)鋒，葉國安</p&

69、gt;<p>　　機電工程學(xué)院，中國礦業(yè)大學(xué)，徐州，中國</p><p>　　摘要：一種采煤機高度智能調(diào)節(jié)系統(tǒng)是在無人工作面開采的關(guān)鍵技術(shù)。高度調(diào)節(jié)機構(gòu)的策略是在控制采煤機高度調(diào)節(jié)系統(tǒng)的數(shù)學(xué)模型的基礎(chǔ)上提出的。在控制過程和使用動態(tài)模糊神經(jīng)網(wǎng)絡(luò)（D-FNN）中的非線性和時間的變化，對該系統(tǒng)的逆特性進行了研究。自適應(yīng)在線學(xué)習(xí)和誤差補償機制保證系統(tǒng)溫度的實時性能和可靠性。Matlab / Simulink

70、環(huán)境模擬高度調(diào)節(jié)控制系統(tǒng)采用了來自德國的艾柯夫SL500采煤機的參數(shù)。仿真結(jié)果表明，一個D-FNN控制器的跟蹤誤差小于一個PID控制器。此外，D-FNN控制方案具有良好的通用性和跟蹤性能，使得它可以滿足采煤機高度調(diào)節(jié)系統(tǒng)的需求。</p><p>　　關(guān)鍵詞：采煤機; 高度調(diào)節(jié)系統(tǒng)；動態(tài)模糊神經(jīng)網(wǎng)絡(luò)</p><p><b>　　1引言</b></p>&l

71、t;p>　　采煤機及其控制系統(tǒng)是煤炭開采的主要組成部分。采煤過程包括汽包吊裝和牽引力控制系統(tǒng)?，F(xiàn)在國內(nèi)采煤汽包吊裝使用手動調(diào)整后，試用手動調(diào)整通過人工觀測或幾何的軌道記憶切割的方法后測試扦插。安裝在采煤機上的傳感器可以識別已提議的煤巖。來自傳感器的信息將被用來直接實現(xiàn)滾筒的高度控制，自動升降采煤機。由于煤層結(jié)構(gòu)的復(fù)雜性，這項基于簡單的鼓高度反饋的技術(shù)沒有被廣泛地應(yīng)用，煤巖界面識別以及煤炭頂，煤底的相關(guān)技術(shù)問題，如此全面的要求采煤的

72、機械化。其他已提出了智能采煤機高度調(diào)節(jié)系統(tǒng)基于一種自適應(yīng)PID神經(jīng)網(wǎng)絡(luò)控制的方法。這需要操作采煤機高度調(diào)節(jié)系統(tǒng)通過從神經(jīng)網(wǎng)絡(luò)的數(shù)據(jù)樣本參數(shù)仔細選擇和調(diào)整算法。然后將通過檢查對測試樣品的性能決定該系統(tǒng)的適用性。進行了測定后的神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)和參數(shù)，可以應(yīng)用到實際系統(tǒng)。而系統(tǒng)運行中實現(xiàn)自適應(yīng)學(xué)習(xí)和控制的參數(shù)可以進一步調(diào)整。建立這樣一個系統(tǒng)涉及到相當(dāng)大的不確定性同樣需要大量的時間。</p><p>　　考慮到需要提高產(chǎn)品

73、質(zhì)量和資源回收鼓的高度自動化控制的因素，我們提出了一種新的方法叫采煤機智能高度調(diào)節(jié)系統(tǒng)控制方法。它基于動態(tài)模糊神經(jīng)網(wǎng)絡(luò)（D-FNN）。D-FNN神經(jīng)網(wǎng)絡(luò)具有強大的在線學(xué)習(xí)，快速學(xué)習(xí)和良好的概括能力的特點。設(shè)計一個智能的高度調(diào)整控制系統(tǒng)的采煤機，D-FNN提供實時控制，提高采煤機高度調(diào)節(jié)系統(tǒng)的動態(tài)特性，并提供了理論基礎(chǔ)。</p><p>　　2采煤機高度調(diào)節(jié)系統(tǒng)的分析</p><p>　　2

74、.1采煤機高度調(diào)節(jié)系統(tǒng)的結(jié)構(gòu)</p><p>　　采煤機高度調(diào)節(jié)機構(gòu)采用了液壓伺服系統(tǒng)具有良好的動態(tài)性能。</p><p>　　圖1是一個滾筒采煤機。電動液壓伺服系統(tǒng)是控制液壓缸的延伸并且使搖臂移動到設(shè)定的高度。該調(diào)整機構(gòu)是由一系列的連桿機構(gòu)和相應(yīng)的運動鏈組成的。通過描述各部件的相對運動來說明高度調(diào)節(jié)是如何發(fā)生的。運動的詳細分析如下。</p><p><b&g

75、t;　　假設(shè)：</b></p><p>　　1）所有組件的剛性和彈性變形被忽略;</p><p>　　2）所有機制之間的差距將被忽略。</p><p>　　圖1。采煤機高度調(diào)節(jié)機構(gòu)</p><p>　　2.2采煤機高度調(diào)節(jié)系統(tǒng)的數(shù)學(xué)分析</p><p>　　圖2示出液壓缸的初始位置為La，終點位置為Lb，搖

76、臂的臂長為L，短臂為Lr，高度調(diào)整缸和搖臂之間的拉桿長度為Lg，高度調(diào)整缸和搖臂主軸之間的距離為D，長臂和和短臂之間的角度是 θ 。</p><p>　　圖2。采煤機高度調(diào)節(jié)系統(tǒng)的數(shù)學(xué)模型</p><p>　　定義1。采煤機開采高度H：</p><p>　　結(jié)束位置Lb由下式給出 .xΔ為允許被建立液壓缸的位移。</p><p>　　定

77、義2。在液壓缸中的位移是x Δ ： </p><p><b>　　其中：</b></p><p><b>　　得出：</b></p><p><b>　　其中：</b></p><p><b>　　x Δ替換為：</b></p><p&

78、gt;　　B由下式給出，搖臂高度角的函數(shù)可表示為：</p><p>　　采煤機高度調(diào)節(jié)系統(tǒng)的模型動力學(xué)分析表明，它是一個三階系統(tǒng)。系統(tǒng)傳遞函數(shù)為：</p><p>　　其中，K是系統(tǒng)的增益，ζ是系統(tǒng)的阻尼比，w是系統(tǒng)的固有振動頻率，是拉普拉斯變換的伺服機構(gòu)，xΔ（在式（5）中）的拉普拉斯變換是來自式（6），來自式5中搖臂的擺角θ控制著反饋。</p><p>　　由的

79、高度調(diào)節(jié)系統(tǒng)是非線性的，并且隨時間變化，所以一個傳統(tǒng)的PID控制器不能提供滿意的控制。D-FNN的目的則是為滿足可靠性和實時性的要求。</p><p><b>　　3動態(tài)模糊神經(jīng)網(wǎng)絡(luò)</b></p><p>　　D-FNN基于徑向基函數(shù)（RBF）神經(jīng)網(wǎng)絡(luò)的擴張。這種學(xué)習(xí)算法的突出特點是同步調(diào)整適當(dāng)?shù)慕Y(jié)構(gòu)參數(shù)和識別。提供了適合于快速學(xué)習(xí)的實時控制和采煤機高度調(diào)節(jié)系統(tǒng)建模

80、。動態(tài)模糊神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)如圖3。</p><p>　　在圖3中 x1, x2, …, xr的系統(tǒng)的輸入變量，y是系統(tǒng)的輸出，是輸入變量的隸屬函數(shù)，Rj為Nj表示的模糊規(guī)則，Nj是j的歸節(jié)點。ω的規(guī)則j的連接權(quán)和u是整個系統(tǒng)的規(guī)則編號。</p><p>　　圖3動態(tài)模糊神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)</p><p>　　在搖臂的擺動角θ，被選擇作為系統(tǒng)的輸入變量，控制液壓缸的膨脹。

81、高斯函數(shù)，式（7），用于為隸屬函數(shù)。</p><p>　　i的范圍是從1到R，J的取值范圍從1到u，Uij是xi的隸屬函數(shù)，Cij是高斯隸屬函數(shù)的中心，σJ是高斯隸屬函數(shù)的寬度，Xi，Ri是輸入變量數(shù)，u是隸屬函數(shù)，以及整個系統(tǒng)的規(guī)則編號。</p><p>　　Rj和J的輸出規(guī)則通過下式獲得：</p><p>　　其中X： 神經(jīng)網(wǎng)絡(luò)RBF中心J：</p>

82、;<p>　　這給出了D-FNN模型為：</p><p>　　其中α是i的重量連接規(guī)則。</p><p>　　4 .D-FNN控制計劃 </p><p>　　如圖4。D-FNN的控制方案?；舅悸肥谦@得采煤機高度調(diào)整系統(tǒng)的逆特性，然后從這個逆動力學(xué)模型產(chǎn)生的補償信號。這里有兩個動態(tài)模糊神經(jīng)網(wǎng)絡(luò)：A和B。A是網(wǎng)絡(luò)系統(tǒng)的重量訓(xùn)練，而網(wǎng)絡(luò)B是一個訓(xùn)練有素的

83、網(wǎng)絡(luò)，用于產(chǎn)生控制信號的副本。</p><p>　　圖4。D-FNN控制計劃</p><p><b>　　控制算法是：</b></p><p>　　其中Δx是預(yù)期的高度調(diào)整液壓缸位移;是由PD控制器產(chǎn)生的氣缸的實際位移。是網(wǎng)絡(luò)B所產(chǎn)生的氣缸的實際位移。</p><p>　　PD控制器是更快，更準確的跟蹤性能。D-FNN控

84、制系統(tǒng)的關(guān)鍵是訓(xùn)練D-FNN乙的預(yù)期和實際之間的位移所產(chǎn)生的網(wǎng)絡(luò)B，以盡量減少誤差平方：</p><p>　　A的權(quán)重調(diào)整算法中使用的梯度下降方法：</p><p>　　其中λ是學(xué)習(xí)速率，λ> 0。λ的收斂速度上有很大的影響。λ的增加，加快了收斂速度，這是更適合于時變系統(tǒng)建模和控制。在同一時間的抗干擾系統(tǒng)的性能下降。λ的減少減慢收斂，但產(chǎn)生的系統(tǒng)干擾不太敏感。自我調(diào)節(jié)學(xué)習(xí)率這里提出的

85、方法，其原理是，當(dāng)新的錯誤超過最后一個錯誤已發(fā)生過沖和λ應(yīng)該減少。如果是新的錯誤小于最后一個錯誤的權(quán)重的調(diào)整是有效的和λ應(yīng)增加。如果誤差是恒定的，則λ保持相同。這可以寫成：</p><p>　　測試表明，D-FNN使用自我調(diào)節(jié)學(xué)習(xí)率的方法使用一個固定的學(xué)習(xí)速率的系統(tǒng)相比，需要更少的訓(xùn)練時間。</p><p><b>　　5系統(tǒng)仿真</b></p><

86、;p>　　用于模型中的采煤機高度調(diào)節(jié)系統(tǒng)，利用Matlab / Simulink構(gòu)建的數(shù)學(xué)模型和一個D-FNN控制算法。實際參數(shù)來自德國艾柯夫SL500型采煤機。采煤機最大切割高度為5.50</p><p>　　米，英尺高的墻是1.08米。搖臂的角度為-21.3°?55°。拉桿Lg，2.05米，短臂，Lr，1.20米，D為0.9米，角度θ是67°。</p><

87、;p>　　5.1模擬一個D-FNN控制器</p><p>　　假設(shè)在搖臂內(nèi)移動的范圍為-21.3°?55°。對D-FNN控制策略的痕跡在搖臂的運動軌跡和軌跡跟蹤誤差示于圖 5。</p><p>　　在圖5b搖臂最大的軌跡跟蹤誤差為0.65°，早在訓(xùn)練階段發(fā)生。此時的D-FNN進行在線學(xué)習(xí)，即學(xué)習(xí)正確的逆模型采煤機高度調(diào)節(jié)系統(tǒng)。因此，在早期階段網(wǎng)絡(luò)B具有足

88、夠的準確度，以補償誤差的控制信號。但作為訓(xùn)練所得的平均誤差下降，直到在最后階段，它已降低到±0.1°，滿足系統(tǒng)要求。</p><p>　　5.2模擬PID控制器</p><p>　　搖臂，以及相應(yīng)的跟蹤誤差的軌跡，如圖所示6的傳統(tǒng)的PID控制器。</p><p><b>　　（a）搖臂軌跡</b></p>&l

89、t;p><b>　?。╞）軌跡誤差</b></p><p>　　圖5搖臂和跟蹤誤差軌跡</p><p><b>　?。╝）搖臂軌跡</b></p><p><b>　?。╞）軌跡誤差</b></p><p>　　圖6搖臂的軌跡，跟蹤誤差</p><p&

90、gt;　　如該圖所示。6b中，整個系統(tǒng)中不能接受的是最大的軌跡誤差為5.8°，為。仿真結(jié)果表明，D-FNN控制器是更加穩(wěn)健和調(diào)整快。</p><p><b>　　6結(jié)論</b></p><p>　　采煤機高度調(diào)整結(jié)構(gòu)的數(shù)學(xué)分析被用來建立了一個數(shù)學(xué)模型。采煤機高度調(diào)節(jié)系統(tǒng)控制和反饋變量之間的約束由模型確定的。</p><p>　　模糊控

91、制和神經(jīng)網(wǎng)絡(luò)控制的D-FNN控制策略調(diào)整采煤機高度的綜合優(yōu)勢進行了描述。擬議的系統(tǒng)的控制方案由所需的反時限特性推導(dǎo)，。通過調(diào)整權(quán)重和補償?shù)臏蚀_性控制方案，圓滿地完成了高度調(diào)節(jié)系統(tǒng)的需求。</p><p>　　3）一個模擬D-FNN控制器用艾柯夫SL500采煤機的系統(tǒng)參數(shù)與傳統(tǒng)的PID控制器</p><p>　　進行了比較：D-FNN控制器是更準確的。D-FNN算法克服??了傳統(tǒng)網(wǎng)絡(luò)的局限性

92、優(yōu)化</p><p>　　的zation算法和避免落入局部極小點。在線自學(xué)，大大提高了訓(xùn)練速度。系統(tǒng)的穩(wěn)定性和精度滿足采煤機高度調(diào)節(jié)系統(tǒng)的要求。</p><p><b>　　致謝</b></p><p>　　這項工作，由中國國家高技術(shù)研究發(fā)展計劃（No.2008AA），中國礦業(yè)大學(xué)科技攀登計劃，提供財務(wù)支持表示感謝。</p>&l

93、t;p><b>　　參考文獻</b></p><p>　　[1] 張俊明，樊X，趙效松自動地平線采煤機控制系統(tǒng)。[中國礦業(yè)大學(xué)學(xué)報，2002，31（4）：415-418。（中國）</p><p>　　[2]梁YW，熊紹柏,Neuarl網(wǎng)絡(luò)和PID混合自適應(yīng)控制水平控制采煤機。：控制，自動化，機器人和視覺IEEE第七屆國際會議。新加坡，2002：671-674。

94、（中國）</p><p>　　[3]雷YY，尹周勛，錢H.研究采煤機液壓自動測距切割高度。[重慶大學(xué)學(xué)報，1994，17（1）：52-58。 （中國）</p><p>　　[4]而MJ，吳森泉吝嗇模糊神經(jīng)系統(tǒng)的快速算法。模糊系統(tǒng)與數(shù)學(xué)，2002，126（3）：337-351。</p><p>　　[5]高宇，楊S，Er MJ 自適應(yīng)模糊神經(jīng)網(wǎng)絡(luò)控制機器人。IEEE跨

95、工業(yè)電子，2001，48：1274-1278。</p><p>　　[6] 張 Y C. 基于自適應(yīng)模糊跟蹤控制非線性SISO系統(tǒng)通過VSS和H方法。IEEE跨模糊SYST，2001（9）：278-292。</p><p>　　[7]李才，李椿鏞自組織神經(jīng)模糊系統(tǒng)控制不知名的植物。IEEE模糊系統(tǒng)，2003，11（1）：135-150。</p><p>　　[8]

96、 Er M J, Low C B, Nah K H, Lim M H, Ng S Y.，動態(tài)模糊神經(jīng)網(wǎng)絡(luò)的SCARA機器人的控制器的實時實現(xiàn)。微處理器和微，2002，26（9/10）：449-461。</p><p>　　[9]莊CF，林CT . 嘈雜的語音處理反復(fù)地模糊自適應(yīng)濾波器。IEEE模糊系統(tǒng)應(yīng)用，2001，9（1）：139-152。</p><p>　　[10]Esposito

97、 A, Marinaro M, Oricchio D, Scarpetta S.連續(xù)和非連續(xù)映射的逼近越來越基于RBF神經(jīng)算法。神經(jīng)網(wǎng)絡(luò)，2000，13（6）：651-665。</p><p>　　[11]馬吉坪。的Matlab的擴展的實時DSP軟件的開發(fā)，測試和驗證。：法律程序的第42屆聯(lián)盟設(shè)計自動化。加利福尼亞州，2005：603-606。</p><p>　　[12]哈特TM，麥凱恩

98、D. Matlab作為開發(fā)環(huán)境為FPGA設(shè)計中。：法律程序的第42屆聯(lián)盟設(shè)計自動化。加利福尼亞州，2005年：607-610。</p><p>　　[13]楊YJ，鄧曄，李X 仿真基于MATLAB/ Simulink的篩選過程。[中國礦業(yè)大學(xué)學(xué)報，2006，16（3）：330 - 332。</p><p>　　[14]劉SY，杜CL，崔XX，程十型號的采煤機滾筒切割性能測試。采礦科學(xué)與技術(shù)

99、，2009，19（1）：74-78。</p><p>　　[15]方修琦，胡華，趙JJ 測試和誤差分析的自我定位在無人工作面采煤機操作。采礦科學(xué)與技術(shù)，2010，20（1）：53 - 58。</p><p><b>　　致謝</b></p><p>　　本次畢業(yè)設(shè)計是在王義亮老師的悉心指導(dǎo)下完成的，從設(shè)計的選題，、設(shè)計過程到論文的撰寫，他都投入

100、了大量的精力與心血。</p><p>　　此次畢業(yè)設(shè)計中，我所在的小組設(shè)計的題目是采煤機牽引部的設(shè)計，在整個過程中，指導(dǎo)老師王義亮老師付出了很大的心血，從日程表的制定到監(jiān)督，從圖紙的繪制到說明書的整理與審閱，都在王老師的指導(dǎo)下有條不紊的進行。</p><p>　　在整個設(shè)計過程當(dāng)中，王義亮老師給予了我以及其他組員們無微不至的關(guān)懷和幫助，他認真嚴謹?shù)闹螌W(xué)態(tài)度和淵博的知識都深深的感染了我，他還

101、幾次從出差地專門趕回來給我們指導(dǎo)。由于王義亮老師的這些努力，才保證了我們畢業(yè)設(shè)計的順利進行。王老師具有淵博的專業(yè)技術(shù)知識，他嚴謹務(wù)實的學(xué)術(shù)精神也在深深激勵著我們。為我們今后的生活和工作提供了很好的楷模。在此，我要向王義亮老師致以最衷心的感謝！</p><p>　　在此次畢業(yè)設(shè)計中，還有好多熱心的同學(xué)以及同組的組員給與我我大量的幫助，在此也向他們表示感謝。</p><p>　　感謝各位學(xué)者、