没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
第 卷第 期
智能系统学报
Vol
年 月
CAAI Transactions on Intelligent Systems
Oct
非 线 性 时 滞 系 统 的 稳 定 自 适 应 控 制
任雪梅
北京理工大学 信息科学技术学院北京
摘要针对模型不确定性的连续时间时滞系统提出了一种新的神经网络自适应控制 系统的辨识模型是由神经
网络和系统的已知信息组合构成在此基础上建立时滞系统的预测模型 基于神经网络预测模型的自适应控制器
能够实现期望轨线的跟踪理论上证明了闭环系统的稳定性 连续搅拌釜式反应器仿真结果表明了该控制方案的
有效性
关键词自适应控制时滞系统神经网络 系统辨识
中图分类号TP文献标识码A文章编号
Stable adaptive neural network con trol of
nonlinea r time delay systems
REN Xuemei
Department of Automatic Control Beijing Institute of Technology Beijing China
AbstractIn this paper a new adaptive neural network controller is presented for a class of continuoustime nonlin
ear time delay systems subject to modeling uncertaintyThe neural network model requires a priori knowledge about
plant dynamics to provide prediction models for time delay systemsAn adaptive controller based on neural networks
was developed to produce the desired tracking performance in uncertain conditionsStability of the closedloop sys
tem is proved by the Lyapunov methodThe effectiveness of the proposed scheme was demonstrated through its ap
plication to the control of a continuous stirred tank reactor
Keyword sadaptive control time delay systems neural networks system identification
收稿日期
基金项目National Natural Science Foundation of China
The dynamic behaviors of many practical systems
are characterized by time delay due to transportation
lags and measurement delays which put severe limita
tions on control performanceA variety of time delay
compensation techniques have been presented for linear
dynamic modelsThe most noteworthy is the Smith
predictor
which is used to improve the control per
formance of linear time delay systemsFor nonlinear
systems with time delay some nonlinear extensions of
the Smith predictor have been proposed by Kravaris
and Wright
and Wong and Seborg
Although the
Smith predictor can help to deal with time delay prob
lems it requires that the model of the controlled
process be known accurately
Neural networks NNs have gained more atten
tion in the control of nonlinear systems due to their
learning capability and nonlinear approximation proper
tyRecent NNbased control schemes focus on using
NNs to design stable adaptive controllers based on the
Lyapunov stability theory
It should be noted that
most stable NNbased adaptive control strategies are
best suited to nonlinear systems free from time delay
For the control of time delay nonlinear systems Huang
and Lewis
provided a new structure of neural network
feedback controller for dynamical system in canonical
form with time delays caused by communication chan
nelsRivals and Personnaz
proposed a design proce
dure of neural internal model control system for stable
processes with delay in which the nonadaptive indirect
control system was designed from the inverse of the
modelGe and Hong et al
presented an adaptive
neural controller for a class of strictfeedback nonlinear
systems with unknown time delays and unknown virtual
control coefficientsHo and LI et al
discussed an
adaptive neural controller based on wavelet neural net
works for a class of nonlinear systems with state delay
Feedback linearization techniques have proved to
be a useful tool in solving control problems for nonlin
ear systems with input delay
However these meth
ods require exact mathematical models of plant dynam
icsObtaining an accurate model is not easy due to the
inherent complexity of nonlinear systemsIn general
partial knowledge of the plant can be obtained from pri
or operating dataIf a plant model is available this in
formation can be incorporated into the control scheme
to improve control performanceBased on these con
siderations this paper incorporates neural networks
with partly known plant dynamics to design an identifi
cation model for nonlinear systems with input delayA
nonlinear adaptive controller is developed based on a
neural network to provide time delay compensationIt
is shown that the proposed controller makes the system
output track a desired trajectory and its closedloop sta
bility is guaranteed
Neural network identification
model
Let denote the norm and
F
be
the Frobenius norm ie givenA a
ij
R
m n
A
F
trA
T
A
ij
a
ij
Moreover we have
AxA
F
xConsider a class of SISO non
linear systems with input delay
i
i
i r
r
a but
q
y
where R
r
R
n r
are the state variables u R
and yR are the controloutputrespectivelyis the
known time delayThe mappings a b
and q are unknown smooth vector fields defined
on a compact set
R
n
Assume that a
b
q
are
precisely known dynamics which are available from
previous experimental data a
b
q
are unknown nonlinear dynamics representing
system uncertainties and disturbancesThen the system
can be written as
y
r
a
y a
y b
y
b
yut
q
y q
y
where y y
yy
r
T
Remark 1 The system can be regarded as
the normal form of the general nonlinear system de
scribed by
x fx gxut
y hx
where x R
n
is the state variableIf the system has
relative degree r there exists a diffeomorphism
T
T
T
x defined by
i
i
x L
i
f
hxi
r and
k
r k
xkn r with L
g
r k
x
such that the system can be transformed into the
normal form described by
The control objective is stated as follows given a
desired trajectory y
d
find a control ut such that the
system output y tracks the desired trajectory y
d
while all
remaining signals of the system are boundedFor
this control purpose we make the following assump
tions
Assumption 1 The desired trajectory y
d
and its
derivatives
y
d
y
r
d
are continuous and bounded
Assumption 2 The sign of b
y b
i
yis
positive ie there exists a known constant b
such that b
y b
yb
The neural network identification model for is
described by
y
r
r
i
i
y
r i
a
y w
T
y
b
y w
T
yut
A
q
y w
T
y
where
yR and
R
n r
are the network output esti
mates of y and respectively
i
i rare cho
sen such that the polynomial s
r
s
r
r
is
Hurwitz A
is a n r n r stable matrixw
R
L
w
R
L
and w
R
L
n r
are weight matrixes
y R
L
y R
L
and
y R
L
are basis function vectors
Remark 2 Since the weights appear linearly in
the identification model and we regard the
functionsas w
T
i
i
yi as the linearly pa
rameterized approximatorsSeveral neural network ba
sis functions can be applied for this purpose such as
第 期任雪梅非线性时滞系统的稳定自适应控制
剩余6页未读,继续阅读
资源评论
weixin_38624519
- 粉丝: 5
- 资源: 899
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
最新资源
- 基于Ngrok内网穿刺实现web端控制树莓派IO口。全部资料+详细文档+高分项目.zip
- 基于tc与树莓派实现的弱网测试工具全部资料+详细文档+高分项目.zip
- 机械设计插片检测设备sw17可编辑全套设计资料100%好用.zip
- 基于树莓派、opencv、stm32、ebox的大平板小球平衡项目全部资料+详细文档+高分项目.zip
- 基于树莓派(debian系统)实现DIY的电子相册服务全部资料+详细文档+高分项目.zip
- 基于树莓派3b的口罩识别全部资料+详细文档+高分项目.zip
- 基于树莓派4B和OPENCV 的人脸识别全部资料+详细文档+高分项目.zip
- 基于树莓派zero的背单词小工具全部资料+详细文档+高分项目.zip
- 基于树莓派的0-5V示波器和信号发生器全部资料+详细文档+高分项目.zip
- 基于树莓派打造的环境信息采集平台全部资料+详细文档+高分项目.zip
- 基于树莓派的ROS机器人操作系统移植和应用研究全部资料+详细文档+高分项目.zip
- 基于树莓派的带屏智能音箱全部资料+详细文档+高分项目.zip
- 基于树莓派的计算机视觉框架部署全部资料+详细文档+高分项目.zip
- 机械设计半自动人工装箱设备sw18可编辑全套设计资料100%好用.zip
- 基于树莓派的光固化3D打印机助手,让你随时随地可以访问打印工作。全部资料+详细文档+高分项目.zip
- 基于树莓派的人脸识别和语音提醒全部资料+详细文档+高分项目.zip
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功