Professional Documents
Culture Documents
Adaptive PID and Model Reference Adaptive Control Switch
Adaptive PID and Model Reference Adaptive Control Switch
Research Article
Adaptive PID and Model Reference Adaptive Control Switch
Controller for Nonlinear Hydraulic Actuator
Received 4 September 2016; Revised 6 June 2017; Accepted 22 August 2017; Published 15 October 2017
Copyright © 2017 Xin Zuo et al. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Nonlinear systems are modeled as piecewise linear systems at multiple operating points, where the operating points are modeled as
switches between constituent linearized systems. In this paper, adaptive piecewise linear switch controller is proposed for improving
the response time and tracking performance of the hydraulic actuator control system, which is essentially piecewise linear. The
controller composed of PID and Model Reference Adaptive Control (MRAC) adaptively chooses the proportion of these two
components and makes the designed system have faster response time at the transient phase and better tracking performance,
simultaneously. Then, their stability and tracking performance are analyzed and evaluated by the hydraulic actuator control system,
the hydraulic actuator is controlled by the electrohydraulic system, and its model is built, which has piecewise linear characteristic.
Then the controller results are compared between PID and MRAC and the switch controller designed in this paper is applied
to the hydraulic actuator; it is obvious that adaptive switch controller has better effects both on response time and on tracking
performance.
that ⋃𝑙𝑖=1 𝑆𝑖 = 𝑆 with 𝑆 ∈ R𝑛 , and 𝑆𝑗 ∩ 𝑆𝑘 = 0 for all 𝑗 ≠ 𝑘. 𝐴 𝑖 Lemma 4. The system 𝑧(𝑡) ̇ = 𝐴 𝑚 (𝑡)𝑧(𝑡) is exponentially stable
and 𝑏𝑖 are supposed to be unknown. with decay rate 𝜎 ∈ (0, 1/2𝛼) if 𝑇0 satisfies
The problem is to find a controller 𝑢(𝑡) to ensure that the
state variables of the plant track asymptotically the reference 𝛼 𝑎2
states, say 𝑥𝑚 (𝑡). 𝑇0 ≥ ln (1 + 𝜇Δ𝐴 𝑚 ) , 𝜇 = 𝑚 max 𝑃𝑚𝑖 , (4)
1 − 2𝜎𝛼 𝜆 𝑚 𝛽 𝑖∈M
Here, we assume that the piecewise linear reference model
is given by where Δ𝐴 𝑚 = max𝑖,𝑗∈M ‖𝐴 𝑚𝑖 − 𝐴 𝑚𝑗 ‖ and 𝑎𝑚𝑖 , 𝜆 𝑚𝑖 > 0
such that ‖𝑒𝐴 𝑚𝑖 𝑡 ‖ ≤ 𝑎𝑚𝑖 𝑒−𝜆 𝑚𝑖 𝑡 , with 𝑎𝑚 = max𝑖∈M 𝑎𝑚𝑖 ,
̇ (𝑡) = 𝐴 𝑚𝑖 𝑥 (𝑡) + 𝑏𝑚𝑖 𝑟 (𝑡) ,
𝑥𝑚 (2) 𝜆 𝑚 = min𝑖∈M 𝜆 𝑚𝑖 , 𝛼 = max𝑖∈M 𝜆 𝑚𝑎𝑥 [𝑃𝑚𝑖 ], and 𝛽 =
min𝑖∈M 𝜆 𝑚𝑖𝑛 [𝑃𝑚𝑖 ], where 𝜆 𝑚𝑖𝑛 and 𝜆 𝑚𝑎𝑥 denote the minimum
where 𝐴 𝑚𝑖 , 𝑖 ∈ M, are stable and 𝑟(𝑡) ∈ 𝑅 is a bounded and maximum eigenvalues of a matrix.
reference input signal, when (𝐴 𝑚𝑖 , 𝑏𝑚𝑖 ) is active, as indicated
by 𝑥(𝑡) ∈ 𝑆𝑖 . 3.1. Adaptive Controller Structure. In this section, state feed-
back adaptive controller is proposed for the piecewise linear
Remark 1. The control systems with such piecewise com- system (1) to make its closed-loop system stable and state
ponents have become an important area of research. As tracking. And the following controller is applied:
discussed in Section 1, many results have been investigated.
For the nonlinear hydraulic actuator [29, 30] in subsea oil
𝑢 (𝑡) = 𝐹 (𝑡) 𝑥 (𝑡) + 𝐾 (𝑡) 𝑟 (𝑡) , (5)
and gas production, fast response time and better tracking
performance are pursued.
where 𝐾 ∈ R is the gain scalar and 𝐹 ∈ R1×𝑛 is the gain
In this paper, the control objective is to develop a switch matrix.
controller such that the closed-loop system is stable and 𝑥(𝑡) As shown in Figure 2 the 𝑖th region of the system adaptive
asymptotically tracks 𝑥𝑚 (𝑡), which has faster response time at control input is
the transient phase and better tracking performance.
𝑢 (𝑡) = 𝐹𝑖 𝑥 (𝑡) + 𝐾𝑖 𝑟 (𝑡) , 𝑖 ∈ M, (6)
3. Adaptive Switch Controller Design with (1) and (6), and we get the closed-loop system
In this section, a switch controller is proposed for the piece-
wise linear system (1) to achieve closed-loop stability and state 𝑥̇ (𝑡) = 𝐴 𝑖 𝑥 (𝑡) + 𝑏𝑖 𝐹𝑖 𝑥 (𝑡) + 𝑏𝑖 𝐾𝑖 𝑟 (𝑡) ,
tracking. The controller consists of PID and MRAC control (7)
𝑥 (𝑡) ∈ 𝑆𝑖 , 𝑖 ∈ M,
stage.
To simplify our discussion, we make the following
where the nominal parameters 𝐹0𝑖 , 𝐾0𝑖 exist, such that
assumptions about the piecewise linear system.
Assumption 2. In this note, assume that all the plant’s state 𝐴 𝑚𝑖 = 𝐴 𝑖 + 𝑏𝑖 𝐹0𝑖 ,
(8)
variables can be observed. 𝑏𝑚𝑖 = 𝑏𝑖 𝐾0𝑖 .
Assumption 3. For every linear subsystem, assume that the
PID and MRAC controller will switch at the switching point, The state tracking error is
and the switching point is decided by the norm of the error
between the state output and the state estimated during PID 𝑒 (𝑡) = 𝑥𝑚 (𝑡) − 𝑥 (𝑡) ,
(9)
control stage, which will be studied in the following. 𝑒 ̇ (𝑡) = 𝑥𝑚
̇ (𝑡) − 𝑥̇ (𝑡) .
The bounded 𝑥𝑚 (𝑡) is generated by the time-invariant
parameter set (𝐴 𝑚𝑖 , 𝑏𝑚𝑖 ). For the stability of piecewise linear In view of (7)–(9), we have
system (2), exponential stability of the homogeneous system
̇
𝑧(𝑡) = 𝐴 𝑚 (𝑡)𝑧(𝑡) is necessary and sufficient for stability of 𝑒 ̇ (𝑡) = 𝑥𝑚
̇ (𝑡) − 𝑥̇ (𝑡)
system (2) [7, 31, 32]. = 𝐴 𝑚𝑖 𝑒 (𝑡) + (𝐴 𝑚𝑖 − 𝐴 𝑖 − 𝑏𝑖 𝐹𝑖 ) 𝑥 (𝑡)
As studied in [7], let 𝑇0 = min𝑘∈Z+ {𝑡𝑘 − 𝑡𝑘−1 }, where Z+ (10)
stands for all possible integers, and let 𝑃𝑚𝑖 , 𝑄𝑚𝑖 ∈ R𝑛×𝑛 be + (𝑏𝑚𝑖 − 𝑏𝑖 𝐾𝑖 ) 𝑟 (𝑡)
symmetric, positive definite satisfying
= 𝐴 𝑚𝑖 𝑒 + 𝑏𝑚𝑖 𝐾0𝑖−1 𝐹̃𝑖 𝑥 (𝑡) + 𝑏𝑚𝑖 𝐾0𝑖−1 𝐾
̃𝑖 𝑟 (𝑡) ,
𝐴𝑇𝑚𝑖 𝑃𝑚𝑖 + 𝑃𝑚𝑖 𝐴 𝑚𝑖 = −𝑄𝑚𝑖 , 𝑖 ∈ M. (3)
where 𝐹̃𝑖 = 𝐹0𝑖 − 𝐹 and 𝐾
̃𝑖 = 𝐾0𝑖 − 𝐾.
And the following lemma ensures exponential stability of
homogeneous system and thus the stability of (2), which has 3.2. Adaptive Laws. For updating the parameters 𝐹𝑖 , 𝐾𝑖 ,
been proofed by Sang and Tao in [7]. we need to develop adaptive laws based on 𝑒(𝑡), with the
4 Mathematical Problems in Engineering
r (t) x (t)
K Plant
F
+
−
Reference model
−
r (t) + x (t)
Reference model PID control Plant
x(t) −
−1
+
K F
knowledge of lower and upper bounds of the parameters in 3.3. PID Controller Design. PID controller has been widely
𝐹0𝑖 , 𝐾0𝑖 ; the parameter projection adaptive laws researched; its basic structure is shown in Figure 3. For system
(1), the control objective is to make system (1) asymptotic
𝑇 𝑇
{𝑅𝑖1 𝑏𝑚𝑖 𝑃𝑚𝑖 𝑒𝑖 𝑥𝑖 , 𝐹𝑖 (𝑡) ∈ Ξ𝐹 , tracking 𝑥𝑚 (𝑡) with less time.
𝐹𝑖̇ = { Continuous-time PID controller itself is defined by
0, otherwise, several different algorithms. The common version of its
{
(11) algorithm defined
𝑇
{𝑅𝑖2 𝑏𝑚𝑖 𝑃𝑚𝑖 𝑒𝑖 𝑟, 𝐾𝑖 (𝑡) ∈ Ξ𝐾 ,
𝐾̇ 𝑖 = { 𝑡
0, otherwise, 𝑢PID (𝑡) = 𝐾𝑝𝑖 𝑒 (𝑡) + 𝐾𝑖𝑖 ∫ 𝑒 (𝜏) 𝑑𝜏 + 𝐾𝑑𝑖 𝑒 ̇ (𝑡) , (13)
{ 0
where Ξ𝐹 and Ξ𝐾 are the known bound regions of 𝐹𝑖 (𝑡) and with
𝐾𝑖 (𝑡) and 𝑅𝑖1 , 𝑅𝑖2 ∈ R, 𝑖 ∈ M, and satisfy
𝑒 (𝑡) = 𝑥𝑚 (𝑡) − 𝑥 (𝑡) , (14)
𝑅𝑖1 = 𝑃𝐹𝑖 𝐾0𝑖−𝑇 ,
which is the state tracking error, where 𝑥𝑚 (𝑡) = (𝑥𝑚1 , 𝑥𝑚2 , . . . ,
𝑅𝑖2 = 𝑃𝐾𝑖 𝐾0𝑖−𝑇 , (12) 𝑥𝑚𝑛 )𝑇 , 𝑥(𝑡) = (𝑥1 , 𝑥2 , . . . , 𝑥𝑛 )𝑇 , and 𝑒(𝑡) = (𝑒1 , 𝑒2 , . . . , 𝑒𝑛 )𝑇 .
The control objective is to guarantee that the plant state 𝑥(𝑡)
𝑖 ∈ M, tracks the reference trajectory 𝑥𝑚 (𝑡); that is, lim𝑡→∞ 𝑒(𝑡) = 0.
The control variable consists of three parts: proportional
with 𝑃𝐹𝑖 , 𝑃𝐾𝑖 being symmetric positive definite matrix. one, integral one, and derivative one. 𝐾𝑝𝑖 is the proportional
Remark 5. The parameter projection adaptive laws for piece- gain, 𝐾𝑖𝑖 is the integral time, and 𝐾𝑑𝑖 is the derivative time for
wise linear state tracking case in [7] are based on the the 𝑖th linear subsystem, which can be tuned to achieve the
knowledge of known sign 𝐾0𝑖 ; in this paper, the parameter asymptotic tracking of 𝑥𝑚 (𝑡) by 𝑥(𝑡) and takes less time.
projection laws need not have the knowledge of sign 𝐾0𝑖 .
3.4. Switch Controller Design. In this section, a new controller
Remark 6. Without the knowledge of 𝐾0𝑖 , it is difficult to get structure is proposed for the piecewise linear system (1) to
the parameters 𝑅𝑖1 and 𝑅𝑖2 ; however, as 𝑃𝐹𝑖 and 𝑃𝐾𝑖 are defined achieve the asymptotic tracking of 𝑥𝑚 (𝑡) by 𝑥(𝑡) with less
matrices, in this paper, 𝑅𝑖1 and 𝑅𝑖2 are defined matrices. time.
Mathematical Problems in Engineering 5
PID control
− +
x(t)
Ke0)$ Fe0)$
−
u0)$
r (t) + PID control
Reference model x (t)
Plant
−
Switch
u-2!#
Ke-2!# Fe-2!#
MRAC control
Notes. 𝐾𝑒PID , 𝐹𝑒PID , 𝐾𝑒MRAC , and 𝐹𝑒MRAC are all updated based ̂ cannot be got
Under the system with 𝑛 > 1, however, 𝑥(𝑡)
on the projection laws (11); the differences are that 𝐾𝑒PID and only by solving (17), as the inverse matrix of 𝐹 does not exist
𝐹𝑒PID are updated at the PID control stage and 𝐾𝑒MRAC and in many systems, and (15) can be written
𝐹𝑒MRAC are updated at the MRAC control stage.
𝑖 𝑖
We design the switching controller consisting of 2 parts: 𝐹𝑒PID 𝑥̂ (𝑡) = 𝑢PID − 𝐾𝑒PID 𝑟 (𝑡) . (19)
PID and MRAC, as shown in Figure 4. At the first stage,
PID controller is used to guarantee the error convergence to Then we use the following equation to describe the vector
a certain extent. At the same time, adaptive laws 𝐾𝑒PID and of output discrepancies
𝐹𝑒PID (11) are updated by the error between the state output 𝑖 𝑖
𝜉 = 𝑢PID − 𝐾𝑒PID 𝑟 (𝑡) − 𝐹𝑒PID 𝑥̂ (𝑡) . (20)
and the reference, with the plant controlled by the PID; we
have the equation Hence, the loss function can be written as
𝑖 𝑖
𝑢PID = 𝐾𝑒PID 𝑟 (𝑡) + 𝐹𝑒PID 𝑥̂ (𝑡) , ̂ 𝑆) = 𝜉2 = (𝑢PID − 𝐾𝑒PID
(15) 2 𝑖 𝑖 𝑇
𝐿 (𝑥, 𝑟 (𝑡) − 𝐹𝑒PID 𝑥̂ (𝑡))
̂ is the estimated states. (21)
where 𝑥(𝑡) 𝑖 𝑖
In the first stage, PID control laws are used to guarantee ⋅ (𝑢PID − 𝐾𝑒PID 𝑟 (𝑡) − 𝐹𝑒PID 𝑥̂ (𝑡)) ,
boundedness of asymptotic tracking, and MRAC laws are
used to achieve asymptotic tracking of 𝑥𝑚 (𝑡) by 𝑥(𝑡), with the where 𝐿(𝑥,̂ 𝑆) = 𝜉2 denotes the squared error or loss of 𝑥(𝑡)
̂
𝑖
assumption that 𝜀 > 0 exist; we set the switching “condition” ̂ 𝑢PID − 𝐾𝑒PID
on (𝑥, ̂ 𝑆) denotes collective loss of
𝑟(𝑡)) and 𝐿(𝑥,
a function on the training set 𝑆.
‖𝑥̂ (𝑡) − 𝑥 (𝑡)‖2 ≤ 𝜀. (16) ̂ by taking the derivatives of
We can seek the optimal 𝑥(𝑡)
̂ which satisfies the
the loss with respect to the parameters 𝑥(𝑡)
Remark 7. As described in Figure 4, 𝑥(𝑡) ̂ is the estimated following equation:
states to evaluate the MRAC controller during PID control
̂ − 𝑥(𝑡)‖2 > 𝜀, 𝐾 and 𝐹 are
stage. Under the condition ‖𝑥(𝑡) ̂ 𝑆)
𝜕𝐿 (𝑥, 𝑖 𝑇 𝑖
= −2 (𝐹𝑒PID ) (𝑢PID − 𝐾𝑒PID 𝑟 (𝑡))
updated by the error 𝑒(𝑡) = 𝑥𝑚 (𝑡) − 𝑥(𝑡) controlled by PID 𝜕𝑥̂ (22)
controller. When 𝐾 and 𝐹 make the estimated state 𝑥(𝑡) ̂ get 𝑇 𝑖
𝑖
the condition (16), MRAC controller is used to control the + 2 (𝐹𝑒PID ) 𝐹𝑒PID ̂
𝑥.
plant because it has better tracking performance and smaller
𝑖
error. With the inverse of (𝐹𝑒PID )𝑇 𝐹𝑒PID
𝑖
existing, we can get
With (15), there are two conditions during the process of −1
𝑇 𝑇
̂ the first is 𝑛 = 1 and the other is 𝑛 > 1. If 𝑛 = 1,
getting 𝑥(𝑡); 𝑖
𝑥̂ = ((𝐹𝑒PID 𝑖
) 𝐹𝑒PID 𝑖
) (𝐹𝑒PID 𝑖
) (𝑢PID − 𝐾𝑒PID 𝑟 (𝑡)) . (23)
from the equation
𝑖
𝑢PID = 𝐾𝑒PID 𝑖
𝑟 (𝑡) + 𝐹𝑒PID 𝑥̂ (𝑡) , (17) Then, we can get that 𝐹̂𝑒PID
𝑖
in Figure 4 is
𝑖
we can get {𝐹𝑒PID ,
{ 𝑛 = 1,
𝐹̂𝑒PID
𝑖
={ (24)
𝑖 −1 𝑖 {((𝐹𝑖 )𝑇 𝐹𝑖 )−1 (𝐹𝑖 )𝑇 , 𝑛 > 1.
𝑥̂ (𝑡) = (𝐹𝑒PID ) (𝑢PID − 𝐾𝑒PID 𝑟 (𝑡)) . (18)
{ 𝑒PID 𝑒PID 𝑒PID
6 Mathematical Problems in Engineering
For the piecewise linear system (1), which consists of 𝑙 structure is proposed to achieve closed-loop stability (signal
regions linear subsystems, a new adaptive switch controller boundedness) and state tracking.
𝑡
{ { {
{ 𝑢PID = 𝐾𝑝1 𝑒 (𝑡) + 𝐾𝑖1 ∫0 𝑒 (𝑡) 𝑑 (𝑡) + 𝐾𝑑1 𝑒 ̇ (𝑡)
{
{ {
{ {
{
{
{ {
{
{
{ {PID stage: {𝑢PID = 𝐾𝑒PID 𝑖 𝑖
𝑟 (𝑡) + 𝐹𝑒PID 𝑥̂ (𝑡)
{
{ 𝑢 = {
{
{
{ 𝑆1 {
{ {
{
{ {
{
{
{
{
{
{ {‖𝑥̂ (𝑡) − 𝑥 (𝑡)‖2 ≥ 𝜀
{
{ MRAC control: 1
𝑢MRAC = 𝐹𝑒MRAC 1
𝑥 (𝑡) + 𝐾𝑒MRAC 𝑟 (𝑡)
{
{ {
{
{ .
{
{ ..
{
{
{
{
{
{ 𝑡
{
{ { {
{ 𝑢PID = 𝐾𝑝𝑖 𝑒 (𝑡) + 𝐾𝑖𝑖 ∫0 𝑒 (𝑡) 𝑑 (𝑡) + 𝐾𝑑𝑖 𝑒 ̇ (𝑡)
{
{ {
{ {
{
{
{ {
{
{ {PID stage: {𝑢PID = 𝐾𝑒PID
𝑖 𝑖
𝑟 (𝑡) + 𝐹𝑒PID 𝑥̂ (𝑡)
𝑢𝑆 = {𝑢𝑆𝑖 = { {
{ (25)
{
{ {
{ {
{
{ {
{ {‖𝑥̂ (𝑡) − 𝑥 (𝑡)‖2 ≥ 𝜀
{
{ {
{ 𝑖 𝑖
{
{
{ {MRAC control: 𝑢MRAC = 𝐹𝑒MRAC 𝑥 (𝑡) + 𝐾𝑒MRAC 𝑟 (𝑡)
{
{ .
{
{ ..
{
{
{
{
{
{ 𝑙 𝑙 𝑡 𝑙
{
{ { {𝑢PID = 𝐾𝑝 𝑒 (𝑡) + 𝐾𝑖 ∫0 𝑒 (𝑡) 𝑑 (𝑡) + 𝐾𝑑 𝑒 ̇ (𝑡)
{
{
{ {
{ {
{
{
{ {PID stage: {𝑢 = 𝐾𝑙 𝑟 + 𝐹𝑙 𝑥̂
{
{
{
{ 𝑢 = {
{ PID 𝑒PID (𝑡) 𝑒PID (𝑡)
{
{ 𝑆𝑙 { {
{
{
{ {
{
{
{ {
{ {‖𝑥̂ (𝑡) − 𝑥 (𝑡)‖2 ≥ 𝜀
{ { 𝑙 𝑙
{ { MRAC control: 𝑢MRAC = 𝐹𝑒MRAC 𝑥 (𝑡) + 𝐾𝑒MRAC 𝑟 (𝑡) .
Remark 8. At the PID control stage, the estimated state 𝑥(𝑡) ̂ controllers are designed by Ziegler-Nichols Method, which
is got by the controller 𝑢PID , and the controller switches at the can guarantee the stability of the system. So we mainly study
̂ − 𝑥(𝑡)‖2 ≤ 𝜀.
condition ‖𝑥(𝑡) the stability and tracking properties controlled by MRAC
controller.
Remark 9. This novel developed control strategy combines Let {𝑡𝑘 }∞
𝑘=1 denote the time instants at which (1) switches
the advantages of the PID controller and MRAC controller, mode. With the definitions of 𝑎𝑚 , 𝜆 𝑚 , 𝛼, 𝛽, 𝜇, Δ𝐴 𝑚 in
the two major approaches for piecewise linear system, and Lemma 4, we have the following stability and tracking proper-
leads to a flexible controller, allowing us to exploit, maximally, ties.
the benefits of two control algorithms. The MRAC technique
is designed based on the model; however, as the plant Theorem 10. Consider the system consisting of the piecewise
model is not known and the MRAC controller cannot get linear system (1), the adaptive switch controller (6), and
good tracking performance at this stage, PID controller is adaptive laws (11). If
used to guarantee the closed-loop system’s tracking response
time and stability at this stage. The proposed controller is 𝑇0 ≥ 𝑇𝑑 = 𝛼 (1 + 𝑘) ln (1 + 𝜇Δ𝐴 𝑚 ) , 𝑘 > 0, (26)
composed of two components; one component is the PID
controller, which has faster tracking response time at the
transient phase, and the other is the MRAC, which has better the transient tracking error performance is given by
tracking performance when the plant’s model is identified
by the adaptive laws. In other words, we will design a 𝑇 𝑔
‖𝑒‖22 ≤ 𝑓 + + 𝑓, (27)
switch controller, which is integrated with PID and MRAC 𝑇0 𝛽
controller, adaptively chooses the proportion of these two
components, and makes the designed system have faster where 𝑓 = 𝜇Δ𝐴 𝑚 𝑔 and 𝑔 = max{𝑐(1+𝜇Δ𝐴 𝑚 )(1+𝑘)/𝑘, 𝑉(𝑡0 )}.
tracking response time and better tracking performance,
simultaneously. Therefore, we need to design the integrated Proof. For the PID controller, it has been much used in the
controller and develop a state feedback control law 𝑢(𝑡) so control loops of industrial process. Parameters 𝐾𝑝 , 𝑇𝑖 , and 𝑇𝑑
that the closed-loop system is stable and 𝑥(𝑡) asymptotically can be tuned to make the system stable and take less time for
tracks 𝑥𝑚 (𝑡), where 𝑥𝑚 (𝑡) is the reference model. 𝑥(𝑡) asymptotic tracks 𝑥𝑚 (𝑡).
For the MRAC controller, due to the fact that 𝑇𝑑 =
3.5. Stability and Tracking Properties. For system (1), every 𝛼(1 + 𝑘) ln(1 + 𝜇Δ𝐴 𝑚 ), it follows from Lemma 4 that 𝑇0 ≥
linear subsystem has two phases as shown in Figure 4, PID 𝑇𝑑 ensures the stability of (2). For analyzing the stability
control stage and MRAC control stage. The parameters of PID and tracking properties about closed-loop system consisting
Mathematical Problems in Engineering 7
of piecewise linear system (1), there, build the piecewise (i) 𝑇 ≤ 𝑇0 , 𝑡𝑘−1 ≤ 𝑡 ≤ 𝑡 + 𝑇 < 𝑡𝑘 . There is no mode switch
Lyapunov function over [𝑡, 𝑡 + 𝑇], since 𝑉 is nonincreasing; we have
𝑡+𝑇
𝑉 = 𝑒𝑇 𝑃𝑚𝑖 𝑒 + 𝐹̃𝑖𝑇 𝑃𝐹𝑖 𝐹̃𝑖 + 𝐾𝑖
̃𝑖 ,
̃𝑇 𝑃𝐾𝑖 𝐾 (28) ‖𝑒‖22 = ∫ |𝑒 (𝜏)|2 𝑑𝜏 ≤ (𝑉 (𝑡) − 𝑉 (𝑡 + 𝑇)) ≤ 𝑉 (𝑡)
𝑡 (35)
where 𝑃𝐹𝑖 , 𝑃𝐾𝑖 are symmetric positive definite matrices.
= 𝑔.
The derivative of 𝑉𝑖 by applying the properties of matrix
eigenvalues, along (8)–(12), is (ii) 𝑇 ≤ 𝑇0 , 𝑡 < 𝑡𝑘 ≤ 𝑡 + 𝑇. There is one and only one
switch over [𝑡, 𝑡 + 𝑇]. We have
𝑇
̃̇ 𝑃𝐹𝑖 𝐹̃𝑖 + 𝐹̃𝑇 𝑃𝐹𝑖 𝐹
𝑉̇ = 𝑒𝑇̇ 𝑃𝑚𝑖 𝑒 + 𝑒𝑇 𝑃𝑚𝑖 𝑒 ̇ + 𝑡𝑟 (𝐹 ̃̇ 𝑖 )
𝑖 𝑖 ‖𝑒‖22 ≤ (𝑉 (𝑡) − 𝑉 (𝑡 + 𝑇)) + 𝑒𝑇 (𝑡𝑘 ) Δ𝑃𝑚(𝑘) 𝑒 (𝑡𝑘 )
(36)
𝑇
̃̇ 𝑃𝐹𝑖 𝐾
+ 𝑡𝑟 (𝐾 ̃𝑖 + 𝐾 ̃̇ 𝑖 )
̃𝑇 𝑃𝐹𝑖 𝐾 ≤ 𝑔 + 𝜇Δ𝐴 𝑚 𝑔.
𝑖 𝑖
Motor
Subsea
Open
Controller
Port A chamber
Spring
Port B
chamber
Detector
Displacement sensor
𝑑2 𝑦 128𝜇𝑙𝑝 𝑄
=𝑚 + (𝜌𝑔𝐻 + ) (𝑆 − 𝑆2 ) (40)
𝑑𝑡2 𝜋𝑑𝑝4 With the different flow in the pipe line, piecewise linear
resistance characteristics exist during the hydraulic actuator
+ 𝑘 (𝐿 0 + 𝑦) + 𝑓𝑖 (𝑦) , in opening/closing process.
𝑓𝑖 (𝑦) = 𝛾𝑖 ∗ 𝑦, (42)
𝑑𝑦
𝑄 = (𝑆 − 𝑆1 ) . (41)
𝑑𝑡 where the variable description is shown in Table 1.
Mathematical Problems in Engineering 9
Figure 5 shows the electrohydraulic actuator control Table 2: Parameters of the hydraulic actuator in simulation.
system, which consists of power unit (Tank 1, Tank 2), electro-
hydraulic proportional control valve, hydraulic actuator, and 𝑘 (kN/m) 221
the control unit. As shown in (41) and (42), 𝑓𝑖 (𝑦) changes 𝑆 (mm2 ) 10
with 𝑦 in practical operation, which is piecewise linear. Let 𝑆2 (mm2 ) 4
𝜌𝑔𝐻(𝐴−𝐴 2 )+𝑘𝐿 0 ≐ 𝜌𝑔𝐻𝐴 1 ; for the piecewise linear model, 𝜇 (mm2 /s) 20.4
the system can be further expressed by the following integer 𝛾2 −223
program: 𝑙𝑝 40
𝑆1 (mm2 ) 2
𝑥1̇ = 𝑥2 ,
𝑑𝑝 (mm) 10
𝛾1 −229
1 1
𝑥2̇ = (𝑆 − 𝑆1 ) 𝑝 − (𝑘𝑥1 𝑚 (kg) 1
𝑚 𝑚 (43)
128𝜇𝑙𝑝
+ 𝛾𝑖 (𝑆 − 𝑆2 ) (𝑆 − 𝑆1 ) 𝑥2 − 𝑘𝑖 𝑥1 ) , signal, while the initial conditions are set to 𝑥𝑚 = [0 0]𝑇 .
𝜋𝑑𝑝4 With assumption 1 fulfilled, let
where 𝑥1 = 𝑦, 𝑥2 = 𝑑𝑦/𝑑𝑡, and the piecewise linear function 0 1
𝐴 𝑚1 = [ ],
𝑓 (𝑦) = 𝛾𝑖 ∗ 𝑦, −10 −5
{𝑎, 𝑥1 ∈ Ω1 , (44) 0
𝛾𝑖 = { 𝑏𝑚1 = [ ] ,
2
𝑏, 𝑥1 ∈ Ω2 ,
{ (47)
0 1
where Ω1 and Ω2 are the fluid element distribution in the 𝐴 𝑚2 = [ ],
pipeline, which is determined by the detector shown in −5 −5
Figure 6; in this paper, Ω1 is the region of 𝑥1 ≤ 0.1; Ω2
0
is the region of 𝑥1 > 0.1. For the servo valve, there exist 𝑏𝑚1 = [ ] .
𝑝 = 𝐾𝑢 ∙ 𝑝in and 𝐾𝑢 = 𝑘pv ∙ 𝑢 (where 𝑘pv represents the 2
function of electrohydraulic proportional control valve and
𝑝in is the input pressure of the electrohydraulic proportional There are two stages for every region of the nonlinear
control valve). Then the system can be written system; the first one is PID controller which plays a role in the
process of 𝑥𝑚𝑖 track to 𝑥𝑝𝑖 ; the second is the MRAC controller
𝑥1̇ used.
[ ]
𝑥2̇ At the PID control stage, we have
0 1 𝑡
[
[
] 𝑥1
][ ] 𝑢 = 𝑘𝑝𝑖 𝑒 (𝑡) + 𝑘𝑖𝑖 ∫ 𝑒 (𝜏) 𝑑𝜏 + 𝑘𝑑𝑖 𝑒 ̇ (𝑡) , 𝑖 = 1, 2, (48)
=[ 1 1 128𝜇𝑙𝑝
(𝑆 − 𝑆2 ) (𝑆 − 𝑆1 ))] 𝑥2 (45)
0
− (𝑘 + 𝛾𝑖 ) − (𝛾𝑖 4
𝑚 𝑚 𝜋𝑑𝑝
[ ] where 𝑒(𝑡) = 𝑥𝑚 (𝑡) − 𝑥(𝑡). And the PID controller parameters
0 𝑘𝑝𝑖 = [𝑘𝑝1
𝑖 𝑖
𝑘𝑝2 ], 𝑘𝑖𝑖 = [𝑘𝑖1
𝑖 𝑖
𝑘𝑖2 ], and 𝑘𝑑𝑖 = [𝑘𝑑1
𝑖 𝑖
𝑘𝑑2 ] are
+[1 ] 𝑢.
determined based on Ziegler-Nichols’ methods in practice
(𝑆 − 𝑆1 ) 𝑘pv ∙ 𝑝in
[𝑚 ] operation. In this example, we control the plant by PI
controller with the following parameters:
For simplicity of presentation, we choose 𝑘pv = 1, 𝑝in = 1,
with the elements being the state variables to control the
piston displacement, with (44) and (45) state equation of 𝑘𝑝1 = [20 1] ,
hydraulic actuator with parameters in Table 2.
𝑘𝑝2 = [15 3] ,
(49)
4.2. Switch Controller Synthesis. For the electrohydraulic 𝑘𝑖1 = [10 1] ,
actuator control system, we first construct the reference
model. 𝑘𝑖2 = [2 2] .
̇
𝑥𝑚1 𝐴 𝑚11 𝐴 𝑚12 𝑥𝑚1 (𝑡) 𝑏𝑚11
[ ]=[ ][ ]+[ ] 𝑟 (𝑡) , (46) For the model reference adaptive laws, the parameters
̇
𝑥𝑚2 𝐴 𝑚21 𝐴 𝑚22 𝑥𝑚2 (𝑡) 𝑏𝑚21
𝑅11 = 𝑅21 = 1,
where 𝑥𝑚1 = 𝑦, ̂ 𝑥𝑚2 = 𝑑𝑦/𝑑𝑡̂ are the reference trajectory
of 𝑥1 and 𝑥2 . 𝑟(𝑡) is the bounded continuous reference input 𝑅12 = 𝑅22 = 1,
10 Mathematical Problems in Engineering
0.4 1
xm1 (t), xp1 (t)
(a) (b)
0.4 0.5
xm1 (t), xp1 (t)
(a) (b)
0.4 0.4
xm2 (t), xp2 (t)
0.2 0.2
0 0
−0.2 −0.2
−0.4 −0.4
−0.6 −0.6
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time (s) Time (s)
(a) (b)
0.6
MRAC
PID
Switch
2
xm2 (t)-xp2 (t)
1
0
−1
−2
0 5 10 15 20 25 30 35 40
Time (s)
MRAC
PID
Switch
Figure 10: Errors controlled by switch controller, MRAC controller, and PID controller.
0.25 0.4
0.2 0.2
xm1 (t), xp1 (t)
xm1 (t), xp1 (t)
0.15 0
0.1 −0.2
0.05 −0.4
0 −0.6
0 0.5 1 1.5 2 2.5 3 20 21 22 23 24 25
Time (s) Time (s)
(a)
0.3 0.5
0.2
xm2 (t), xp2 (t)
xm2 (t), xp2 (t)
0
0.1
−0.5
0
−0.1 −1
0 0.5 1 1.5 2 2.5 3 20 21 22 23 24 25
Time (s) Time (s)
(b)
Figure 11: PID controller for piecewise linear system at the transient phase.
12 Mathematical Problems in Engineering
0.25 0.4
0.2 0.2
0.15 0
0.1 −0.2
0.05 −0.4
0 −0.6
0 0.5 1 1.5 2 2.5 3 20 21 22 23 24 25
Time (s) Time (s)
(a)
0.3 1
0.2 0
0.1 −1
0 −2
−0.1 −3
0 0.5 1 1.5 2 2.5 3 20 21 22 23 24 25
Time (s) Time (s)
(b)
Figure 12: MRAC controller for piecewise linear system at the transient phase.
Table 3: Mean square error contrast. about 0.5 seconds for 𝑥𝑝1 to track 𝑥𝑚1 and about 0.6 seconds
for 𝑥𝑝2 to track 𝑥𝑚2 . We can get that the switch controller has
PID MRAC Switch
better performance for the piecewise linear system than using
𝑓MSE (𝑥𝑚1 − 𝑥𝑝1 ) 0.616 0.0493 0.0015 MRAC controller. And the mean square errors are smaller,
𝑓MSE (𝑥𝑚2 − 𝑥𝑝2 ) 0.0655 0.1721 0.0266 controlled by switch controller, shown in Table 5.
5. Conclusion
𝑓MSE (𝑥𝑚1 (𝑡) − 𝑥𝑝1 (𝑡)) is the mean square error of 𝑥𝑝1 to track In this paper, the adaptive switch controller consisting of
𝑥𝑚1 , 𝑓MSE (𝑥𝑚2 (𝑡) − 𝑥𝑝2 (𝑡)) is the mean square error of 𝑥𝑝2 to PID and MRAC controllers for piecewise linear systems is
track 𝑥𝑚2 . From Table 3, we can get that the switch controller studied. The controller adaptively chooses the proportion
has better performance than PID controller and MRAC of these two components and makes the designed system
controller. Table 4 shows the system performance at transient have faster response time and better tracking performance,
phase controlled by PID, MRAC, and the switch controller. simultaneously. The integrated controller and a state feedback
With Table 4, PID technique has smaller mean square errors control law 𝑢(𝑡) have been designed so that the closed-
than MRAC and switch techniques at the transient phase; loop system is stable and 𝑥(𝑡) asymptotically tracks 𝑥𝑚 (𝑡).
at the same time, switch controller has smaller mean square Hydraulic actuator model was built and its piecewise linear
errors when the input changes at 𝑡 = 20 s. characteristics were used to demonstrate the asymptotic
tracking of 𝑥𝑚 (𝑡) by piston displacement. Unlike conven-
Case 2. Simulations are performed for the reference input tional MRAC and PID controllers, the adaptive switch con-
signal 𝑟(𝑡) = 10. From Figures 14 and 15, we can get that 𝑥 troller combines advantages of PID and MRAC; it makes the
needs less time to track 𝑥𝑚 . For the piecewise linear model, piecewise linear control have faster tracking response time
it takes about 1 second for 𝑥𝑝1 to track 𝑥𝑚1 and about 1.25 and better performance, especially in oil production; it needs
seconds for 𝑥𝑝2 to track 𝑥𝑚2 by using MRAC controller and less time to control the actuator displacement and has less
more than that by using the switch controller, which takes hydraulic shock.
Mathematical Problems in Engineering 13
0.25 0.4
0.2
0.2
xm1 (t), xp1 (t)
0 −0.4
0 0.5 1 1.5 2 2.5 3 20 21 22 23 24 25
Time (s) Time (s)
(a)
0.3 0
−0.1
0.2
−0.2
0.1
−0.3
0
−0.4
−0.1 −0.5
0 0.5 1 1.5 2 2.5 3 20 21 22 23 24 25
Time (s) Time (s)
(b)
Figure 13: Switch controller for piecewise linear system at the transient phase.
xm1 (t), xp1 (t)
2.5
2
1.5
1
0.5
0
0 0.5 1 1.5 2 2.5 3
Time (s)
xm1 (t)
xp1 (t)
6
xm2 (t), xp2 (t)
4
2
0
−2
0 0.5 1 1.5 2 2.5 3
Time (s)
xm2 (t)
xp2 (t)
xm1 (t)
xp1 (t)
3
xm2 (t), xp2 (t) 2
1
0
−1
0 0.5 1 1.5 2 2.5 3
Time (s)
xm2 (t)
xp2 (t)
Table 5: Mean square error contrast. Electronics Engineers. Transactions on Automatic Control, vol. 57,
no. 2, pp. 522–528, 2012.
MRAC SWITCH
[8] G. Feng, “Stability analysis of piecewise discrete-time linear
𝑓MSE (𝑥𝑚1 − 𝑥𝑝1 ) 0.0499 0.0387 systems,” Institute of Electrical and Electronics Engineers. Trans-
𝑓MSE (𝑥𝑚2 − 𝑥𝑝2 ) 0.7186 0.2753 actions on Automatic Control, vol. 47, no. 7, pp. 1108–1112, 2002.
[9] E. D. Sontag, “Nonlinear regulation: the piecewise linear
approach,” Institute of Electrical and Electronics Engineers.
Conflicts of Interest Transactions on Automatic Control, vol. 26, no. 2, pp. 346–358,
1981.
The authors declare that they have no conflicts of interest.
[10] C. Briat, “Stability analysis and control of a class of LPV sys-
tems with piecewise constant parameters,” Systems and Control
References Letters, vol. 82, article no. 3921, pp. 10–17, 2015.
[11] Y. Han, Q. Cao, Y. Chen, and M. Wiercigroch, “Chaotic
[1] A. Mohanty, S. Gayaka, and B. Yao, “An adaptive robust thresholds for the piecewise linear discontinuous system with
observer for velocity estimation in an electro-hydraulic system,” multiple well potentials,” International Journal of Non-Linear
International Journal of Adaptive Control and Signal Processing, Mechanics, vol. 70, pp. 145–152, 2015.
vol. 26, no. 12, pp. 1076–1089, 2012.
[12] Y. Lin, Y. Shi, and R. Burton, “Modeling and robust discrete-
[2] D. X. Ba, K. K. Ahn, D. Q. Truong, and H. G. Park, “Integrated
time sliding-mode control design for a fluid power electrohy-
model-based backstepping control for an electro-hydraulic
draulic actuator (EHA) system,” IEEE/ASME Transactions on
system,” International Journal of Precision Engineering and
Mechatronics, vol. 18, no. 1, pp. 1–10, 2013.
Manufacturing, vol. 17, no. 5, pp. 565–577, 2016.
[3] T. Samakwong and W. Assawinchaichote, “PID Controller [13] M. Chen and G. Feng, “Delay-Dependent H∞ filtering of
Design for Electro-hydraulic Servo Valve System with Genetic piecewise-linear systems with time-varying delays,” IEEE Trans-
Algorithm,” in Proceedings of the International Electrical Engi- actions on Circuits and Systems. I. Regular Papers, vol. 55, no. 7,
neering Congress, iEECON 2016, pp. 91–94, tha, March 2016. pp. 2087–2096, 2008.
[4] Q. Zhou, L. Hongyi, C. Wu et al., “Adaptive fuzzy control [14] P. Doležel and I. Taufer, “Piecewise-linear artificial neural
of nonlinear systems with unmodeled dynamics and input networks for PID controller tuning,” Acta Montanistica Slovaca,
saturation using small-gain approach,” Ieee Transactions on Man vol. 17, no. 3, pp. 224–233, 2012.
and Cybernetics Systems, vol. 99, pp. 1–11, 2016. [15] P. Dolezel, P. Rozsival, M. Mariska, and L. Havlicek, “PID
[5] B. Niu, C. K. Ahn, H. Li, and M. Liu, “Adaptive Control for controller design for nonlinear oscillative plants using piecewise
Stochastic Switched Nonlower Triangular Nonlinear Systems linear neural network,” in Proceedings of the 2013 International
and Its Application to a One-Link Manipulator,” IEEE Trans- Conference on Process Control, PC 2013, pp. 19–24, svk, June
actions on Systems, Man, and Cybernetics: Systems, pp. 1–14. 2013.
[6] G. K. Lowe and M. A. Zohdy, “Modeling nonlinear sys- [16] R. H. B. Fey, R. M. T. Wouters, and H. Nijmeijer, “Proportional
tems using multiple piecewise linear equations,” Lithuanian and derivative control for steady-state vibration mitigation in a
Association of Nonlinear Analysts (LANA). Nonlinear Analysis. piecewise linear beam system,” Nonlinear Dynamics, vol. 60, no.
Modelling and Control, vol. 15, no. 4, pp. 451–458, 2010. 4, pp. 535–549, 2010.
[7] Q. Sang and G. Tao, “Adaptive control of piecewise linear [17] W. O. Dennis, Adaptive control of plants with jumping parame-
systems: the state tracking case,” Institute of Electrical and ters [Master, thesis], Elective, University of Virginia, USA, 1995.
Mathematical Problems in Engineering 15
International
Journal of Journal of
Mathematics and
Mathematical
#HRBQDSDĮ,@SGDL@SHBR
Sciences