自適應(yīng)神經(jīng)網(wǎng)絡(luò)控制器在主動(dòng)隔振控制中的應(yīng)用研究

目  錄
摘要 III
Abstract V
第 1 章 緒論 1
1.1 引言 1
1.2 主動(dòng)隔振控制技術(shù)國(guó)內(nèi)外研究現(xiàn)狀與發(fā)展 1
1.2.1 控制器的研究方面 3
第2 章 隔振平臺(tái)設(shè)計(jì)及其動(dòng)力學(xué)分析 5
2.1隔振機(jī)理 5
2.2 主動(dòng)隔振系統(tǒng)設(shè)計(jì) 12
2.3 兩級(jí)主動(dòng)隔振系統(tǒng)的動(dòng)力學(xué)模型 23
2.4 兩級(jí)隔振系統(tǒng)主動(dòng)控制原理 24
2.5 本章小結(jié) 25
第3 章 直接自適應(yīng)神經(jīng)網(wǎng)絡(luò)主動(dòng)隔振控制技術(shù)研究 26
3.1 神經(jīng)網(wǎng)絡(luò)概述 26
3.2 神經(jīng)網(wǎng)絡(luò)基本理論 26
3.2.1 神經(jīng)元的基本模型 26
3.2.2 神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)及其算法 27
3.3 直接自適應(yīng)神經(jīng)網(wǎng)絡(luò)振動(dòng)主動(dòng)控制技術(shù)研究 33
3.3.1控制系統(tǒng)結(jié)構(gòu) 33
3.3.2直接自適應(yīng)神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)及學(xué)習(xí)算法研究 34
3.4 基于Matlab 的自適應(yīng)神經(jīng)網(wǎng)絡(luò)控制器設(shè)計(jì)及仿真 36
3.4.1利用GUI設(shè)計(jì)自適應(yīng)神經(jīng)網(wǎng)絡(luò)控制器的拓 36
3.4.2數(shù)字仿真及結(jié)果分析 36
3.5 本章小結(jié) 38
結(jié)論 39
參考文獻(xiàn) 41
致 謝 43

自適應(yīng)神經(jīng)網(wǎng)絡(luò)控制器在主動(dòng)隔振控制中的應(yīng)用研究
摘要
振動(dòng)給人類和機(jī)械帶來(lái)的危害是不言而喻的，從而主動(dòng)隔振變得相當(dāng)重要。本文將預(yù)測(cè)與神經(jīng)網(wǎng)絡(luò)相結(jié)合，構(gòu)造了一種振動(dòng)主動(dòng)控制系統(tǒng), 運(yùn)用 3 層BP(Back Propagation)網(wǎng)絡(luò)對(duì)系統(tǒng)進(jìn)行了試驗(yàn)研究。
本文分析了致動(dòng)器在雙層隔振系統(tǒng)中不同安裝方式下的主動(dòng)控制力與隔振系統(tǒng)參數(shù)之間的關(guān)系。在該兩級(jí)主動(dòng)隔振系統(tǒng)中，主動(dòng)隔振致動(dòng)器可以有三種不同的安裝方式，采用仿真的方法選定了最優(yōu)的一種安裝方式使各種激振源產(chǎn)生最佳的減振效果佳的?；谌龑覤P 網(wǎng)絡(luò)，構(gòu)造了直接自適應(yīng)神經(jīng)網(wǎng)絡(luò)控制器，并對(duì)隔振系統(tǒng)進(jìn)行了振動(dòng)主動(dòng)控制，避免了以前神經(jīng)網(wǎng)絡(luò)振動(dòng)主動(dòng)控制的雙網(wǎng)絡(luò)結(jié)構(gòu)形式，該方法采用系統(tǒng)的輸出作為神經(jīng)網(wǎng)絡(luò)的學(xué)習(xí)輸入，不需要預(yù)先或在線辨識(shí)系統(tǒng)，能實(shí)時(shí)調(diào)整控制率找到最優(yōu)的控制力，具有良好的自適應(yīng)性。利用GUI 設(shè)計(jì)自適應(yīng)神經(jīng)網(wǎng)絡(luò)控制器的拓?fù)浣Y(jié)構(gòu)，基于Matlab 設(shè)計(jì)了自適應(yīng)神經(jīng)網(wǎng)絡(luò)控制器。研究結(jié)果表明，采用該法設(shè)計(jì)的控制器能夠?qū)崿F(xiàn)從頻率為0.1Hz 到50Hz 的隔振能力，而且隨著輸入層和隱含層神經(jīng)元個(gè)數(shù)的增加，低頻隔振效果也有一些改善，與此同時(shí)，神經(jīng)網(wǎng)絡(luò)控制器的結(jié)構(gòu)也復(fù)雜了，網(wǎng)絡(luò)規(guī)模增大，計(jì)算和訓(xùn)練的速度變慢了，難以滿足振動(dòng)控制系統(tǒng)的實(shí)時(shí)性要求。

關(guān)鍵詞：神經(jīng)網(wǎng)絡(luò)，主動(dòng)隔振，執(zhí)行器，動(dòng)力學(xué)，MATLAB
 

Adaptive neural network controller in active vibration isolation control of the Applied Research
Abstract
Human and mechanical vibration to the harm is self-evident, active vibration isolation has become very important. This article will neural network technology in isolation system, a tectonic active vibration control systems, use of 3-BP (Back Propagation) network to the system were studied, forecast and neural network integration, in the structure of the two isolation , Active vibration isolation actuator is installed on the dynamics of isolation system have a greater impact .
This paper analyses the actuator in the double-isolation system installed in different ways under the control and active vibration isolation system parameters of the relationship between. In the two active vibration isolation system, active vibration isolation actuators can have three different installation methods, the use of simulation has been selected the best way to install a variety of exciting ways to produce the best source of the vibration Excellent results. BP based on the three-tier network, constructed directly adaptive neural network controller and isolation system for the active vibration control and avoid the previous neural network active vibration control of dual-network structure, the method used the system as a neural network output Enter the study, or do not need to advance on-line identification system capable of real-time adjustments to find the optimal control of the controlling power, has a good adaptability. GUI design using adaptive neural network controller topology, the design based on Matlab adaptive neural network controller. The results show that the use of the design of the controller can be realized from the frequency of 0.1 Hz to 50 Hz in isolation, but also with the input of the hidden layer and the increase in the number of neurons, the effect of low-frequency vibration isolation have some improvement, and Meanwhile, the neural network controller is also complicated by the structure, size of network increases, computing and slow down the pace of training, vibration control systems to meet the real-time requirements

Key words: Neural networks,Active vibration isolation, the Executive, Dynamics, MATLAB