Open AccessArticle

Neural Adaptive Funnel Dynamic Surface Control with Disturbance-Observer for the PMSM with Time Delays

Menghan Li

Shaobo Li

^2,*

Junxing Zhang

²,

Fengbin Wu

and

Tao Zhang

School of Mechanical Engineering, Guizhou University, Guiyang 550025, China

State Key Laboratory of Public Big Data, Guizhou University, Guiyang 550025, China

School of Computer Science and Technology, Guizhou University, Guiyang 550025, China

Author to whom correspondence should be addressed.

Entropy 2022, 24(8), 1028; https://doi.org/10.3390/e24081028

Submission received: 22 June 2022 / Revised: 18 July 2022 / Accepted: 21 July 2022 / Published: 26 July 2022

Download

Browse Figures

Review Reports Versions Notes

Abstract

This paper suggests an adaptive funnel dynamic surface control method with a disturbance observer for the permanent magnet synchronous motor with time delays. An improved prescribed performance function is integrated with a modified funnel variable at the beginning of the controller design to coordinate the permanent magnet synchronous motor with the output constrained into an unconstrained one, which has a faster convergence rate than ordinary barrier Lyapunov functions. Then, the specific controller is devised by the dynamic surface control technique with first-order filters to the unconstrained system. Therein, a disturbance-observer and the radial basis function neural networks are introduced to estimate unmatched disturbances and multiple unknown nonlinearities, respectively. Several Lyapunov-Krasovskii functionals are constructed to make up for time delays, enhancing control performance. The first-order filters are implemented to overcome the “complexity explosion” caused by general backstepping methods. Additionally, the boundedness and binding ranges of all the signals are ensured through the detailed stability analysis. Ultimately, simulation results and comparison experiments confirm the superiority of the controller designed in this paper.

Keywords:

disturbance observer; dynamic surface control; permanent magnetic synchronous motor; funnel control; radial basis function neural networks

1. Introduction

The permanent magnet synchronous motor (PMSM) is a separately excited generator composed of a stator and a rotator. In recent decades, PMSMs are widely used in aerospace, defense, and other major fields due to the advantages of simple structure, small size, and low noise [1,2]. With the purpose of environmental protection, PMSMs are currently also used in the fields of new energy vehicles, and other major fields [3,4]. However, PMSMs are systems with nonlinearities, strong coupling, and time-varying. Hence, there has been considerable interest in the high-precision control of PMSMs, which is of great help to enhance the aviation industry and ecological conservation for a nation.

The most classical method of controlling PMSMs is backstepping control. The backstepping method splits the n-order complicated system into n subsystems and realizes the virtual control of every subsystem. In the former

n - 1

steps, it will derive a virtual controller in every step, significantly simplifying the computational process of the controller. Nevertheless, the backstepping method also suffers from the “explosion of complexity” caused by the repeated derivatives of the virtual controller, owing to its inherent properties of it. To circumvent this obstacle, a first-order filter combined with the backstepping control approach [5] is created by Swaroop, called dynamic surface control [6]. Despite dynamic surface control methods [7,8,9], that can reduce computing efforts, various nonlinear factors such as time delays, external disturbances, and physical constraints are ubiquitous in real industrial scenarios [10,11], which may diminish the controlling precision of the PMSM systems. Researchers have proposed proportional integral derivative (PID) control [12], neural network (NN) [5], time delay control [13,14], disturbance observer (DO) [15,16], and constraint control [17,18] methods for different nonlinearities to reach satisfying control results. Hence, the key point is how to design an effective controller to address the various nonlinear uncertainties such as unknown functions, mismatched disturbance, state constraints, and time delays.

For the unknown characteristics and uncertain disturbances, the NN-based adaptive control that has superb evaluating abilities are employed to obtain the high-performance control of nonlinear systems [19,20]. For instance, the authors in [19] design a radial basis function NN (RBFNN) to estimate the error of the high-gain observer and the lumped interference. The authors in [20] utilize a feedforward artificial NN to renew the training parameters of the proportion integration differential controller, offering high dynamic performance. Due to their advantages, RBFNNs are introduced in this paper to estimate the unknown uncertainties. To achieve advanced transient and steady-state performance, NNs still need to be combined with other control strategies.

For the uncertain load disturbances, considerable pioneering investigations based on the DO have been frequently suggested to achieve the steady operation of the PMSM [21]. Based on the sliding mode control technology, a DO is proposed to attain desired anti-load-disturbance capability [11]. A DO is fused into the super-twisting sliding mode method to compensate for the lumped disturbance [22]. A second-order DO is utilized to estimate the parameters perturbations, ensuring the accuracy of the PMSM system [16]. Though the DO-based schemes can effectively suppress the uncertain perturbations, the transient performance that is significant for the robustness of the system is not negligible. To reduce the calculation burdens and further enhance the transient performance, this paper introduces a finite-time second-order command filter to approximate the matched and mismatched disturbance in a finite time with the aid of these investigations.

The physical constraints complicate controller design process, proper strategies which are proposed by researchers to improve the transient performance and stability of nonlinear systems. They have introduced barrier Lyapunov functions [23], prescribed performance functions (PPF) [24,25], and funnel control strategies. As the novel control methods developed, the barrier Lyapunov functions need to be modified to suit specific changing situations, resulting in its non-universality. Likewise, the demand for precise initial values limits the applications of PPF strategies. Based on these, the improved PPFs are applied to further attain superior transient performance [26,27]. Compared with PPFs and barrier Lyapunov functions, the funnel control method is considered a promising way to deal with the constraints owing to its effectiveness for the output overshoot [28,29], and does not need precise initial conditions. The funnel control method has been applied in vehicles [30], and robots [31] because it can avoid the modification of the controllers. According to the above investigations, this paper suggests a funnel controller for the output-constrained PMSM system to achieve excellent control performance.

It is known that time delays harm the dynamic performance and the stability of the PMSMs. The time delay control strategies that can strengthen both transient and steady-state responses, recently, have been involved in many pioneering backstepping investigations. The studies can be roughly divided into two groups namely Smith predictor control tools [32,33] and Lyapunov-Krasovskii functional methods [34,35]. For example, a smith predictor is combined with the speed controller to eliminate the time delays [32]. An appropriate Lyapunov-Krasovskii functional is employed to ensure the asymptotic stability of the delay-dependent PMSM system [34]. A hyperbolic function is utilized to deal with the delay caused by the low-pass filter [36]. From these, a proper Lyapunov-Krasovskii functional is designed in this paper to address the time delay issues, which enhances both the transient and steady-state performances of PMSM.

Motivated by these discussions, a neural adaptive funnel dynamic surface control (FDSC) scheme combined with the DO is designed in this article to address the output-restrained system with time delays and load interference. The primary contributions of this paper are outlined as:

(a): The prescribed performance control method can ensure the tracking error converges to a predefined arbitrary small residual set [24,25]. However, the transient performance still needs to be improved. Upon this, a neural adaptive funnel control strategy with advanced transient performance is proposed to restrict the output response of PMSM into a certain funnel region. The simulation section compares the funnel control method with neural dynamic surface control (NDSC) and PID methods, and the funnel controller has smaller tracking error and better transient performance. The advanced transient performance makes the devised controller more applicable.
(b): The constraint-considered controller can reasonably simulate the physical constraints in the actual operating environment of the PMSM [23,25]. Nevertheless, there are many other nonlinear factors in the actual industries that need to be taken into account. The FDSC method considers output constraints, time delays, and mismatched external load interference, which allows FDSC to simulate the actual situation more realistically. The Lyapunov-Krasovskii functionals and DO are devised to suppress the time delays and approximate the unmatched external load interference. The simulation section shows that the FDSC method has smaller steady-state and transient errors, and the robustness and dynamic performances are strengthened compared with the NDSC and the PID schemes.

Notation: For any variable

c

\hat{c}

denotes the estimated value of

c

\tilde{c} = c - \hat{c}

represents the state error.

‖ ● ‖

indicates the 2-norm of

●

2. System Formulation and Preliminaries

2.1. System Statement

Under the

(d - q)

coordinate frame, formulate the PMSM mathematical model as [7]:

{\begin{cases} \dot{θ} = ω \\ \dot{ω} = 3 n_{p} / (2 J) [(L_{d} - L_{q}) i_{d} i_{q} + φ i_{q}] - B ω / J - T_{L} / J \\ {\dot{i}}_{q} = - R_{s} i_{q} / L_{q} - n_{p} L_{d} i_{d} ω / L_{q} - n_{p} φ ω / L_{q} + u_{q} / L_{q} \\ {\dot{i}}_{d} = - R_{s} i_{d} / L_{d} + n_{p} L_{q} i_{q} ω / L_{d} + u_{d} / L_{d}, \end{cases}

(1)

where

u_{q}

and

u_{d}

are control variables. The descriptions of PMSM parameters are exhibited in Table 1.

To simplify (1), define

a_{1} = 3 n_{p} φ / 2, a_{2} = 3 n_{p} (L_{d} - L_{q}) / 2, b_{1} = - R_{s} / L_{q}, b_{2} = - n_{p} L_{d} / L_{q}, b_{3} = - n_{p} φ / L_{q}, b_{4} = 1 / L_{q}, c_{1} = - R_{s} / L_{d}, c_{2} = n_{p} L_{q} / L_{d}, c_{3} = 1 / L_{d} .

Moreover, describe the states variables as:

x_{1} = θ, x_{2} = ω, x_{3} = i_{q}, x_{4} = i_{d} .

(2)

Considering time delays and asymmetric output constraints, the model (1) can be reconstructed as:

{\begin{array}{l} {\dot{x}}_{1} = x_{2} + Δ l_{1} [x (t - τ_{1})] \\ {\dot{x}}_{2} = \frac{a_{1}}{J} x_{3} + \frac{a_{2}}{J} x_{3} x_{4} - \frac{B}{J} x_{2} - \frac{T_{L}}{J} + Δ l_{2} [x (t - τ_{2})] + Δ E \\ {\dot{x}}_{3} = b_{1} x_{3} + b_{2} x_{2} x_{4} + b_{3} x_{2} + b_{4} u_{q} + Δ l_{3} [x (t - τ_{3})] \\ {\dot{x}}_{4} = c_{1} x_{4} + c_{2} x_{2} x_{3} + c_{3} u_{d} + Δ l_{4} [x (t - τ_{4})], \end{array}

(3)

where

x = {(x_{1}, x_{2}, x_{3}, x_{4})}^{T} \in R^{4}

, and the state variable

x_{1}

subjects to:

x_{1} \in Π_{x_{1}} : = {x_{1} \in R : x_{d} (t) - f_{1} (t) < x_{1} (t) < x_{d} (t) + f_{1} (t)} .

(4)

Remark 1.

In the actual PMSM operating environment, there are plenty of unknown uncertainties, such as time delays, physical constraints, and matched and mismatched disturbances. It is essential to consider these limitations in the PMSM system to guarantee that the controller designed in this paper approach reality. It is worth mentioning that for the PMSM systems, the unmatched interference

Δ E

and the time delays

Δ l_{i} (x (t - τ_{i})), i = 1, \dots, 4

in (3) as well as the time-varying symmetric output constraints in (4) are firstly considered simultaneously. Thereby, the controller devised here based on (3) is more suitable for real industrial fields.

Based on the above observations, the control objective of this paper is to devise a neural adaptive funnel dynamic surface controller based on DO for (3) to realize:

(a): The system output $x_{1}$ tracks the desired signal $x_{d}$ , where the transient control behavior of (3) can be retained by (4).
(b): The other signals in the resulting system are bounded.

To ensure these objectives, the following assumption and lemmas are provided.

Assumption 1

([37]). The continuous desired signal

x_{d} (t)

and its ith-order derivatives

x_{d}^{(i)} (t), (i = 0, \dots, 4)

are bounded. The continuous state-constrained functions

f_{1} (t)

as well as its jth-order derivatives

f_{1}^{(j)} (t), (j = 0, \dots, 4)

are bounded.

Lemma 1

([38]). For any

f (η_{1}, \dots, η_{n}) : R^{m_{r}} \times \dots \times R^{m_{n}} \to R,

there are smooth functions

ω_{i} (η_{i}) > 0 :

R^{m_{i}} \to R

satisfying

| f (η_{1}, \dots, η_{n}) | \leq \sum_{i = 1}^{n} ω_{i} (η_{i}) .

For the initial value

f (0, \dots, 0) = 0,

ω_{i} (0) = 0

fulfills

| f (0, \dots, 0) | \leq \sum_{i = 1}^{n} ω_{i} (0)

as well.

Remark 2.

According to Lemma 1, there are continuous positive functions

ξ_{i k}, i = 1, \dots, 4

restricting the time-delay terms

Δ l_{i} (x (t - τ_{i})), i = 1, \dots, 4

in the system (3) as

Δ l_{i} (x (t - τ_{i}))

\leq \sum_{k = 1}^{4} ξ_{i k} (x_{k} (t - τ_{i}))

. With the aid of Young’s inequality, it leads to:

e_{i} Δ l_{i} [x (t - τ_{i})] \leq 1 / 2 n e_{i}^{2} + 1 / 2 \sum_{j = 1}^{n} ξ_{i j}^{2} (x_{j} (t - τ_{i})), i = 1, \dots, 4 .

(5)

Lemma 2

([39,40]). For any real variables

p, q

, and the positive constants

m_{i}, i = 1, 2, 3

, the following inequality holds:

{| a |}^{m_{1}} {| b |}^{m_{2}} ⩽ \frac{m_{1}}{m_{1} + m_{2}} m_{3} {| a |}^{m_{1} + m_{2}} + \frac{m_{2}}{m_{1} + m_{2}} m_{3}^{- m_{1} / m_{2}} {| b |}^{m_{1} + m_{2}} .

(6)

Lemma 3

([41]). For

σ > 0,

there exists the set

Ω_{e} : = {e \in R : | e | \leq 0.2554 σ} .

For

e \notin Ω_{e}

, the inequality

1 - 16 {t a n h}^{2} (e / σ) < 0

holds.

Thereafter, function arguments are sometimes dropped without confusion.

2.2. Neural Network Systems and Function Approximation

Since the RBFNNs can approximate the unknown hard-to-calculated functions that are in a closed set at any precision [42,43], this paper introduces an RBFNN function to estimate unknown functions

W (X)

W (X) = ℘^{* T} P (X) + ψ (X), \forall X \in Ω_{X},

(7)

where

X = {[x_{1}, x_{2}, \dots, x_{n}]}^{T}

indicates the input vector,

ψ (X)

fulfills

| ψ (X) | < ψ_{M}

with a bounded constant

ψ_{M}

P (X) =

{[p_{1} (x), p_{2} (x), \dots, p_{l} (x)]}^{T}

is a vector of basis function, and select

p_{i} (x)

as the versatile Gaussian functions:

p_{i} (x) = \exp [- {(x - ν_{i})}^{T} (x - ν_{i}) / χ_{i}^{2}], i = 1, 2, \dots, m .

(8)

Construct the expected weight vector

℘^{*}

as:

℘^{*} = \arg \min_{℘ \in R^{n}} {\sup_{X \in D_{X}} ‖ W (X) - {\hat{℘}}^{T} P (X) ‖} .

(9)

The 2-norms of the variables are utilized to assess the weights allowing for a reduction in the computational burden of RBFNN [42,43]. Consequently, it leads to:

β_{i} = {‖ ℘_{i} ‖}^{2} = ℘_{i}^{T} ℘_{i}, i = 1, \dots, 4 .

(10)

3. Design of Neural Adaptive Funnel Control

This section introduces the funnel controller designed in this paper. In Section 3.1, a funnel-type variable with improved PPF is introduced to remove the output constraint of the variable

x_{1}

. In Section 3.2, a DO is proposed to approximate the matched and mismatched disturbance that is utilized in step 2. In Section 3.3, four steps are listed to show the whole design procedure of the funnel controller for the PMSM system.

3.1. Funnel Control with Improved Prescribed Performance Function

The core point of the funnel control is devising the given functions of the envelope as the boundaries that restrict the tracking error

s_{1} (t)

, where

s_{1} (t) = x_{1} - x_{d}

. Based on this idea, a funnel-type function is selected at the very beginning of the controller design as:

g (t) = F_{k} (f_{1} (t), G (t), ‖ s_{1} (t) ‖) .

(11)

A PPF

f^{*} (t)

[44] is constructed to obtain both steady-state and transient performances of

s_{1} (t)

f^{*} (t) = (f_{0} - f_{\infty}) \exp (- π t) + f_{\infty},

(12)

where the design parameters 3

f_{0} > f_{\infty} > 0

and the minimum convergence rate value

π > 0

From [44], we can attain that the control objectives can be achieved when the following condition satisfied:

| s_{1} (t) | < ω f^{*} (t),

(13)

where the design constants

ω > 0

To obtain better steady performance, a funnel control with envelope boundaries is introduced. An improved PPF based on (12) is proposed:

f_{1} (t) = f_{0} \exp (- π t) + \frac{t}{π (t + 1)} f_{\infty},

(14)

where

f_{\infty} / π

represents the steady-state error.

Remark 3.

The improved PPF

f_{1} (t)

indicated by (14) has the same properties as (12). Moreover, the improved PPF has a faster convergence speed and more outstanding transient performance than the conventional PPF while choosing the same parameters values of

f_{0}, f_{\infty}, π

. As an example, Figure 1 gives the profiles of (14) and (12) with

f_{0} = 1, f_{\infty} = 0.05, π = 3, 5, 10

The gain function

F_{k} (\cdot)

[31] can be regulated by:

F_{k} (f_{1} (t), G (t), ‖ s_{1} (t) ‖) = G (t) / (f_{1} (t) - ‖ s_{1} (t) ‖) .

(15)

It can be concluded from (15) that the value of

F_{k} (\cdot)

decreases as

f_{1} (t)

moves away from

‖ s_{1} (t) ‖

when

G (t)

is supposed to be fixed. Regrettably, the gain

F_{k} (\cdot)

formulated by (15) is not differentiable at the point of

s_{1} (t) = 0

, which does not meet the using conditions of backstepping.

To circumvent the obstacle, a modified funnel variable

η_{1}

by virtue of (14) and [31] is first created as follows:

η_{1} = s_{1}^{2} / (f_{1}^{2} - s_{1}^{2}) .

(16)

Remark 4.

f_{1} \to s_{1}

, the value of

η_{1}

will be infinite. For this reason, assuming that the starting value of the tracking error

s_{1} (0)

is limited by the funnel boundaries. Choosing a proper design parameter

f_{0}

can effectively eliminate the infinite starting value.

Remark 5.

Compared with prescribed performance control [45], the improved PPF-based funnel control not only provides better transient performance but is also simpler and more efficient without calculating inverse conversion errors, which is conducive to the stable operation of PMSM.

The time derivatives of

η_{i}

in (16) are obtained as:

{\dot{η}}_{1} = \frac{2 s_{1} f_{1}^{2}}{{(f_{1}^{2} - s_{1}^{2})}^{2}} ({\dot{s}}_{1} - s_{1} \frac{{\dot{f}}_{1}}{f_{1}}) = Γ_{1} ({\dot{s}}_{1} - s_{1} \frac{{\dot{f}}_{1}}{f_{1}}),

(17)

where the variable

Γ_{1} = 2 s_{1} f_{1}^{2} / {(f_{1}^{2} - s_{1}^{2})}^{2}

3.2. Disturbance Observer

For the difficult-to-compute external disturbance, a DO is an effective utensil to observe the matched and mismatched interference. Hence, introduce the intermediate variables

d, D

, and conceive the DO as [46]:

{\begin{cases} {\dot{\hat{x}}}_{2} = a_{1} / J x_{3} + a_{2} / J x_{3} x_{4} - B / J x_{2} - T_{L} / J + Δ l_{2} (x (t - τ_{2})) + Δ E_{1} \\ Δ E_{1} = - κ_{1} ι_{1}^{1 / 3} {| {\hat{x}}_{2} - x_{2} |}^{2 / 3} sgn ({\hat{x}}_{2} - x_{2}) + Δ {\hat{E}}_{2} \\ Δ {\dot{\hat{E}}}_{2} = - κ_{1} ι_{1}^{1 / 2} {| Δ {\hat{E}}_{2} - Δ E_{2} |}^{1 / 2} {sgn}^{1 / 2} (Δ {\hat{E}}_{2} - Δ E_{2}) + Δ \hat{E} \\ Δ \dot{\hat{E}} = - κ_{2} ι_{1} sgn (Δ \hat{E} - Δ {\dot{\hat{E}}}_{2}), \end{cases}

(18)

where coefficients

κ_{1} > 0, κ_{2} > 0, ι_{1} > 0

Describing variables as

{\tilde{x}}_{2} = {\hat{x}}_{2} - x_{2}, Δ \tilde{E} = Δ \hat{E} - Δ E, Δ {\tilde{E}}_{2} = Δ {\hat{E}}_{2} - Δ E_{2}

, it leads to:

{\begin{cases} {\dot{\tilde{x}}}_{2} = - κ_{1} ι_{1}^{1 / 3} {| {\hat{x}}_{2} - x_{2} |}^{2 / 3} sgn ({\hat{x}}_{2} - x_{2}) + Δ \tilde{E} \\ Δ \dot{\tilde{E}} = - κ_{1} ι_{1}^{1 / 2} {| Δ {\hat{E}}_{2} - Δ E_{2} |}^{1 / 2} sgn (Δ {\hat{E}}_{2} - Δ E_{2}) + (Δ \hat{E} - Δ E) \\ Δ \tilde{E} \in - κ_{2} ι_{1} sgn (Δ \hat{E} - Δ {\dot{\hat{E}}}_{2}) + [- ι_{1}, ι_{1}] . \end{cases}

(19)

Supposing a negligible rate of change for load disturbance

Δ E_{2}

, then it is bounded. According to

Δ {\tilde{E}}_{2} = Δ {\hat{E}}_{2} - Δ E_{2}

, it can be concluded that

Δ \tilde{E}

is bounded. Similarly, we can attain that

Δ \hat{E}

and

{\hat{x}}_{2}

are limited to the small adjacency of

Δ E

and

x_{2}

, respectively.

Remark 6.

The interference error can be well estimated by the excellent DO (18). In other words, we can select suitable parameters

κ_{1}, κ_{2}

and

ι_{1}

to tackle the matched and unmatched load disturbance. Moreover, the excellent DO can prevent the system tremors due to the symbolic function in the classical sliding mode control method simultaneously [21].

3.3. Neural Adaptive Funnel Controller Design

This subsection will provide a dynamic surface controller by integrating the funnel control technology. First of all, let us introduce the following coordinate conversion:

e_{1} = η_{1}, e_{2} = x_{2} - u_{2 c}, e_{3} = x_{3} - u_{3 c}, e_{4} = x_{4} .

(20)

Define the first-order filters as:

λ_{i} {\dot{u}}_{i c} + u_{i c} = u_{i}, u_{i c} (0) = u_{i} (0), i = 2, 3,

(21)

where

λ_{i}

are constants.

Remark 7.

This paper introduces first-order filters to bypass the repeated derivatives of

u_{i}

so that the “explosion of complexity” can be eliminated by designing proper filters. However, there exist filter errors that influence the control precision by introducing the first-order filters into the controller design procedure. When choosing proper Lyapunov functions to devise the virtual and actual controller, the filter errors need to be taken into account to improve the control accuracy.

Akin to (20), describe the filter errors

ϕ_{i}

as:

ϕ_{i} = u_{i c} - u_{i}, i = 2, 3 .

(22)

Fusing (3) and (16) into (20), and take the derivatives of

e_{i}, i = 1, \dots, 4

in (20):

{\begin{cases} {\dot{e}}_{1} = Γ_{1} {e_{2} + ϕ_{2} + u_{2} - {\dot{x}}_{d} + Δ l_{1} [x (t - τ_{1})] - s_{1} {\dot{f}}_{1} / f_{1}} \\ {\dot{e}}_{2} = e_{3} + ϕ_{3} + u_{3} + (\frac{a_{1}}{J} - 1) x_{3} + \frac{a_{2}}{J} x_{3} x_{4} - \frac{B}{J} x_{2} - \frac{T_{L}}{J} - {\dot{u}}_{2 c} + Δ l_{2} [x (t - τ_{2})] + Δ E \\ {\dot{e}}_{3} = b_{1} x_{3} + b_{2} x_{2} x_{4} + b_{3} x_{2} + b_{4} u_{q} + Δ l_{3} [x (t - τ_{3})] - {\dot{u}}_{3 c} \\ {\dot{e}}_{4} = c_{1} x_{4} + c_{2} x_{2} x_{3} + c_{3} u_{d} + Δ l_{4} [x (t - τ_{4})] . \end{cases}

(23)

This paper utilized RBFNNs to estimate the unknown functions, so there exist estimation errors that need to be computed. Describe the approximated errors

{\tilde{β}}_{i}

as:

{\tilde{β}}_{i} = β_{i} - {\hat{β}}_{i}, i = 1, \dots, 4 .

(24)

The next steps are further design procedures of the neural adaptive FDSC. The brief structure of the control procedure is shown in Figure 2.

Step 1. The design of the stunning control law $u_{2}$ and the adaptive law ${\dot{\hat{β}}}_{1}$ .

Taking the tracking error

e_{1}

, filter error

ϕ_{2}

, approximation error

{\tilde{β}}_{1}

, and the Lyapunov–Krasovskii functional

V_{T}

into account, choose the Lyapunov function

V_{1}

as:

V_{1} = 1 / 2 e_{1}^{2} + 1 / 2 ϕ_{2}^{2} + 1 / (2 d_{1}) {\tilde{β}}_{1}^{2} + V_{T},

(25)

subject to the Lyapunov–Krasovskii functional

V_{T}

as:

V_{T} = \frac{1}{2} \sum_{i = 1}^{4} \sum_{j = 1}^{4} \exp [- ƛ (t - τ_{i})] \int_{t - τ_{i}}^{t} \exp (ƛ v) ξ_{i j}^{2} [x_{j} (v)] d v,

(26)

where the design constants

d_{1} > 0, ƛ > 0

ξ_{i j}

are utilized to address the time delays below.

Remark 8.

To further improve the precision of the controller designed later, theproper Lyapunov-Krasovskii functional

V_{T}

is utilized to compensate for the time-varying delays.

Taking the derivative of

V_{T}

in (26) obtains:

{\dot{V}}_{T} = \frac{1}{2} \sum_{i = 1}^{4} \sum_{j = 1}^{4} \exp [- ƛ (t - τ_{i})] ξ_{i j}^{2} [x_{j} (t)] - \frac{1}{2} \sum_{i = 1}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - ƛ V_{T} .

(27)

With (24) and (27), deriving

V_{1}

in (25) yields:

{\dot{V}}_{1} = e_{1} {\dot{e}}_{1} - \frac{1}{d_{1}} {\tilde{β}}_{1} {\dot{\hat{β}}}_{1} + \frac{1}{2} \sum_{i = 1}^{4} \sum_{j = 1}^{4} \exp [- ƛ (t - τ_{i})] ξ_{i j}^{2} [x_{j} (t)] - \frac{1}{2} \sum_{i = 1}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] + ϕ_{2} {\dot{ϕ}}_{2} - ƛ V_{T} .

(28)

Fusing (23) into (28) generates:

\begin{matrix} {\dot{V}}_{1} = e_{1} Γ_{1} {e_{2} + ϕ_{2} + u_{2} - {\dot{x}}_{d} + Δ l_{1} [x (t - τ_{1})] - s_{1} \frac{{\dot{f}}_{1}}{f_{1}}} - \frac{1}{2} \sum_{i = 1}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] \\ + \frac{1}{2} \sum_{i = 1}^{4} \sum_{j = 1}^{4} \exp [- ƛ (t - τ_{i})] ξ_{i j}^{2} [x_{j} (t)] - \frac{1}{d_{1}} {\tilde{β}}_{1} {\dot{\hat{β}}}_{1} - ƛ V_{T} + ϕ_{2} {\dot{ϕ}}_{2} . \end{matrix}

(29)

Based on Remark 2 with

n = 4

, the following inequality can be obtained:

e_{1} Γ_{1} Δ l_{1} [x (t - τ_{1})] \leq 2 e_{1}^{2} Γ_{1}^{2} + \frac{1}{2} \sum_{j = 1}^{4} ξ_{1 j}^{2} [x_{j} (t - τ_{1})] .

(30)

Taking (30) into (29) derives:

{\dot{V}}_{1} \leq e_{1} Γ_{1} (e_{2} + ϕ_{2} + u_{2} - {\dot{x}}_{d} - s_{1} \frac{{\dot{f}}_{1}}{f_{1}}) - \frac{1}{2} \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \frac{1}{d_{1}} {\tilde{β}}_{1} {\dot{\hat{β}}}_{1} + 2 e_{1}^{2} Γ_{1}^{2} + T (x) + ϕ_{2} {\dot{ϕ}}_{2} - ƛ V_{T},

(31)

where the time delay function

T (x) = 1 / 2 \sum_{i = 1}^{4} \sum_{j = 1}^{4} (- \exp (ƛ τ_{i}) ξ_{i j}^{2} (x_{j} (t)))

. We can infer that

\lim_{e_{1} \to 0} (T (x) / e_{1}) \to \infty

. Consequently,

(16 / e_{1}) \tanh^{2} (e_{1} / σ_{1}) T

is brought in (32) to further eliminate the time delays. Then, rewrite (31) as:

\begin{matrix} {\dot{V}}_{1} \leq e_{1} [Γ_{1} e_{2} + Γ_{1} ϕ_{2} + Γ_{1} u_{2} + \frac{16}{e_{1}} T \tanh^{2} (\frac{e_{1}}{σ_{1}}) + 2 e_{1} Γ_{1}^{2} - Γ_{1} {\dot{x}}_{d} - Γ_{1} s_{1} \frac{{\dot{f}}_{1}}{f_{1}}] \\ + ϕ_{2} {\dot{ϕ}}_{2} - ƛ V_{T} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T - \frac{1}{2} \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \frac{1}{d_{1}} {\tilde{β}}_{1} {\dot{\hat{β}}}_{1} . \end{matrix}

(32)

Then this paper utilizes an unknown function

W_{1} (X_{1})

to generalize the unknown functions:

W_{1} (X_{1}) = 3 e_{1} Γ_{1}^{2} + e_{1} + 16 / e_{1} T \tanh^{2} (e_{1} / σ_{1}) - Γ_{1} {\dot{x}}_{d},

(33)

where

W_{1} (X_{1})

is an unknown function. According to (11), (16) and (20), it can be learned that

e_{1}

is composed of a known desired signal

x_{d}

and an unknown state

x_{1}

Γ_{1}

consists of a known desired signal

x_{d}

, an assured function

f_{1}

, and an unknown state

x_{1}

. The RBFNN is developed to estimate the unknown function

W_{1} (X_{1})

, then

X_{1} = {[x_{1}, \dots, x_{4}, x_{d}, {\dot{x}}_{d}]}^{T}

For the unknown and uncertain functions, RBFNN is considered as a promising tool to approximate them at any precision. An RBFNN is designed to approximate

W_{1} (X_{1})

W_{1} (X_{1}) = ℘_{1}^{T} P_{1} (X_{1}) + ψ_{1} (X_{1}), | ψ_{1} (X_{1}) | ⩽ ψ_{M},

(34)

where

ψ_{M}

stands for the positive bounded constant.

As it can be seen from (33), (32) can be simplified by an RBFNN:

\begin{matrix} {\dot{V}}_{1} \leq e_{1} (Γ_{1} e_{2} + Γ_{1} ϕ_{2} + Γ_{1} u_{2} + ℘_{1}^{T} P_{1} (X_{1}) + ψ_{1} (X_{1}) - Γ_{1} s_{1} \frac{{\dot{f}}_{1}}{f_{1}} - e_{1} Γ_{1}^{2} - e_{1}) - \frac{1}{d_{1}} {\tilde{β}}_{1} {\dot{\hat{β}}}_{1} \\ + [1 - 16 / e_{1} T \tanh^{2} (e_{1} / σ_{1})] T + ϕ_{2} {\dot{ϕ}}_{2} - ƛ V_{T} - \frac{1}{2} \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] . \end{matrix}

(35)

According to (34) and Lemma 2 with

m_{1} = m_{2} = m_{3} = 1

, functions that are difficult to compute can be deflated to:

{\begin{cases} e_{1} Γ_{1} ϕ_{2} \leq e_{1}^{2} Γ_{1}^{2} + \frac{1}{4} ϕ_{2}^{2} \\ e_{1} W_{1} (X_{1}) = e_{1} [℘_{1}^{T} P_{1} (X_{1}) + ψ_{1} (X_{1})] \leq \frac{1}{4 μ_{1}^{2}} β_{1} e_{1}^{2} P_{1}^{T} P_{1} + μ_{1}^{2} + \frac{ψ_{M}^{2}}{4} + e_{1}^{2}, \end{cases}

(36)

where

μ_{1}

stands for the positive design constant.

Integrating (36) into (35) generates:

\begin{array}{l} {\dot{V}}_{1} \leq & e_{1} (Γ_{1} e_{2} + Γ_{1} u_{2} - Γ_{1} s_{1} \frac{{\dot{f}}_{1}}{f_{1}}) + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T - \frac{1}{2} \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \frac{1}{d_{1}} {\tilde{β}}_{1} {\dot{\hat{β}}}_{1} \\ - ƛ V_{T} + ϕ_{2} {\dot{ϕ}}_{2} + \frac{1}{4 μ_{1}^{2}} e_{1}^{2} β_{1} P_{1}^{T} P_{1} + \frac{1}{4} ϕ_{2}^{2} + \frac{ψ_{M}^{2}}{4} + μ_{1}^{2} . \end{array}

(37)

Devise the stunning control law

u_{2}

and the adaptive law

{\dot{\hat{β}}}_{1}

as:

\begin{array}{l} u_{2} = - \frac{(f_{1}^{2} - s_{1}^{2}) s_{1}}{2 f_{1}^{2}} (k_{1} + \frac{1}{4 μ_{1}^{2}} {\hat{β}}_{1} P_{1}^{T} P_{1}) + s_{1} \frac{{\dot{f}}_{1}}{f_{1}} \\ {\dot{\hat{β}}}_{1} = \frac{d_{1}}{4 μ_{1}^{2}} e_{1}^{2} P_{1}^{T} P_{1} - γ_{1} {\hat{β}}_{1}, \end{array}

(38)

where the design constants

k_{1} > 0, γ_{1} > 0

Inserting the control law

u_{2}

and adaptive law

{\dot{\hat{β}}}_{1}

in (38) into (37) derives:

\begin{array}{l} {\dot{V}}_{1} \leq & - k_{1} e_{1}^{2} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T - \frac{1}{2} \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] + \frac{γ_{1}}{d_{1}} {\tilde{β}}_{1} {\hat{β}}_{1} \\ - ƛ V_{T} + Γ_{1} e_{1} e_{2} + ϕ_{2} {\dot{ϕ}}_{2} + μ_{1}^{2} + \frac{ψ_{M}^{2}}{4} + \frac{ϕ_{2}^{2}}{4} . \end{array}

(39)

According to (21)–(24), and (38), differentiate the filter error

ϕ_{2}

in (22) as:

{\dot{ϕ}}_{2} = {\dot{u}}_{2 c} - {\dot{u}}_{2} = - \frac{ϕ_{2}}{λ_{2}} + M_{2} (e_{1}, e_{2}, ϕ_{2}, {\hat{β}}_{1}, x_{d}, {\dot{x}}_{d}, {\ddot{x}}_{d}),

(40)

where

M_{2} (e_{1}, e_{2}, ϕ_{2}, {\hat{β}}_{1}, x_{d}, {\dot{x}}_{d}, {\ddot{x}}_{d})

is a smooth function.

In virtual of [47], we can infer that there exists an upper value

{\bar{M}}_{2} ({\bar{M}}_{2} \geq 0)

for the original conditions to restrict

M_{2} (e_{1}, e_{2}, ϕ_{2}, {\hat{β}}_{1}, x_{d}, {\dot{x}}_{d}, {\ddot{x}}_{d})

within the prescribed set, it produces:

{\dot{ϕ}}_{2} \leq - \frac{ϕ_{2}}{λ_{2}} + {\bar{M}}_{2},

(41)

where

{\bar{M}}_{2} > | M_{2} (e_{1}, e_{2}, ϕ_{2}, {\hat{β}}_{1}, x_{d}, {\dot{x}}_{d}, {\ddot{x}}_{d}) | .

Based on Lemma 2 with

m_{1} = m_{2} = m_{3} = 1

, consider

a = ϕ_{2}, b = {\bar{M}}_{2}

. Then, the hard-to-compute function is converted into:

ϕ_{2} {\dot{ϕ}}_{2} = - ϕ_{2}^{2} / λ_{2} + ϕ_{2} {\bar{M}}_{2} \leq - (1 / λ_{2} - 1 / 4) ϕ_{2}^{2} + {\bar{M}}_{2}^{2} .

(42)

Substituting (42) into (39) generates:

\begin{array}{l} {\dot{V}}_{1} \leq & - k_{1} e_{1}^{2} - 1 / 2 \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - (\frac{1}{λ_{2}} - \frac{1}{4}) ϕ_{2}^{2} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{γ_{1}}{d_{1}} {\tilde{β}}_{1} {\hat{β}}_{1} \\ - ƛ V_{T} + Γ_{1} e_{1} e_{2} + {\bar{M}}_{2}^{2} + ψ_{M}^{2} / 4 + μ_{1}^{2} . \end{array}

(43)

Step 2. The design of the stunning control law $u_{3}$ and the adaptive law ${\dot{\hat{β}}}_{2}$ .

Considering the state error

e_{2}

, filter error

ϕ_{3}

and the approximation error

{\tilde{β}}_{2}

, the second Lyapunov function

V_{2}

is chosen as:

V_{2} = V_{1} + 1 / 2 e_{2}^{2} + 1 / 2 ϕ_{3}^{2} + 1 / (2 d_{2}) {\tilde{β}}_{2}^{2},

(44)

where

d_{2}

denotes the positive constants.

With (24), taking the derivative of

V_{2}

in (44) leads to:

{\dot{V}}_{2} = {\dot{V}}_{1} + e_{2} {\dot{e}}_{2} + ϕ_{3} {\dot{ϕ}}_{3} - 1 / d_{2} {\tilde{β}}_{2} {\dot{\hat{β}}}_{2} .

(45)

Integrating (23) and (43) into (45) produces:

\begin{array}{l} {\dot{V}}_{2} \leq & - k_{1} e_{1}^{2} - 1 / 2 \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - (\frac{1}{λ_{2}} - \frac{1}{4}) ϕ_{2}^{2} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{γ_{1}}{d_{1}} {\tilde{β}}_{1} {\hat{β}}_{1} \\ + e_{2} {e_{3} + ϕ_{3} + u_{3} + (a_{1} / J - 1) x_{3} + a_{2} / J x_{3} x_{4} - B / J x_{2} - T_{L} / J + Δ l_{2} [x (t - τ_{2})] \\ - {\dot{u}}_{2 c} + Δ E} - ƛ V_{T} + ϕ_{3} {\dot{ϕ}}_{3} + Γ_{1} e_{1} e_{2} + {\bar{M}}_{2}^{2} + μ_{1}^{2} + ψ_{M}^{2} / 4 - 1 / d_{2} {\tilde{β}}_{2} {\dot{\hat{β}}}_{2} . \end{array}

(46)

Resembling (30), define

n = 4

, then it derives:

e_{2} Δ l_{2} [x (t - τ_{2})] \leq 2 e_{2}^{2} + \frac{1}{2} \sum_{j = 1}^{4} ξ_{2 j}^{2} [x_{j} (t - τ_{2})] .

(47)

Assume that

| Δ \hat{E} - Δ E = Δ \tilde{E} | \leq q_{1}

, and the constant

q_{1} > 0

. It is hard to determine the positive and negative of

e_{2}

, an inequality

(Δ \hat{E} - Δ E) e_{2} \leq | Δ \hat{E} - Δ E | | e_{2} | \leq q_{1} | e_{2} |

is brought to ensure that the following inequality holds. Pouring (47) into (46) gets:

\begin{array}{l} {\dot{V}}_{2} \leq & - k_{1} e_{1}^{2} - \frac{1}{2} \sum_{i = 3}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - (\frac{1}{λ_{2}} - \frac{1}{4}) ϕ_{2}^{2} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{γ_{1}}{d_{1}} {\tilde{β}}_{1} {\hat{β}}_{1} \\ + e_{2} [e_{3} + ϕ_{3} + Δ \hat{E} + (a_{1} / J - 1) x_{3} + u_{3} - B / J x_{2} + a_{2} / J x_{3} x_{4} - T_{L} / J + 2 e_{2} - {\dot{u}}_{2 c} \\ + Γ_{1} e_{1}] - ƛ V_{T} + ϕ_{3} {\dot{ϕ}}_{3} + {\bar{M}}_{2}^{2} + q_{1} | e_{2} | + μ_{1}^{2} + ψ_{M}^{2} / 4 - 1 / d_{2} {\tilde{β}}_{2} {\dot{\hat{β}}}_{2} . \end{array}

(48)

Design

F_{2} (X_{2})

F_{2} (X_{2}) = (\frac{a_{1}}{J} - 1) x_{3} + \frac{a_{2}}{J} x_{3} x_{4} - \frac{B}{J} x_{2} - \frac{T_{L}}{J} + 4 e_{2} + Γ_{1} e_{1},

(49)

where

e_{1} = x_{1} - x_{d}, e_{2} = x_{2} - α_{2 c}

Γ_{1}

is composed of a known desired signal

x_{d}

, a given function

f_{1}

, and an unknown state

x_{1}

. Based on the structure of the uncertain function

F_{2} (X_{2})

X_{2}

is composed of

x_{1}, x_{2}, x_{3}, x_{4}, x_{d}

and

u_{2 c}

. Then it leads to

X_{2} = {[x_{1}, \dots, x_{4}, x_{d}, u_{2 c}]}^{T}

A piecewise function

W_{2} (X_{2})

is devised as:

W_{2} (X_{2}) = {\begin{cases} F_{2} (X_{2}) - q_{1}, e_{2} \leq 0 \\ F_{2} (X_{2}) + q_{1}, e_{2} > 0, \end{cases}

(50)

where

X_{2} = {[x_{1}, \dots, x_{4}, x_{d}, u_{2 c}]}^{T}

Integrating (50) into (48) generates:

\begin{array}{l} {\dot{V}}_{2} \leq & - k_{1} e_{1}^{2} - \frac{1}{2} \sum_{i = 2}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - (\frac{1}{λ_{2}} - \frac{1}{2}) ϕ_{2}^{2} - \frac{1}{d_{2}} {\tilde{β}}_{2} {\dot{\hat{β}}}_{2} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{γ_{1}}{d_{1}} {\tilde{β}}_{1} {\hat{β}}_{1} \\ + e_{2} [e_{3} + ϕ_{3} + u_{3} + Δ \hat{E} + W_{2} (X_{2}) - 2 e_{2} - {\dot{u}}_{2 c}] - ƛ V_{T} + ϕ_{3} {\dot{ϕ}}_{3} + {\bar{M}}_{2}^{2} + μ_{1}^{2} + \frac{ψ_{M}^{2}}{4} . \end{array}

(51)

Akin to (34), the unknown function

W_{2} (X_{2})

can be approximated by an RBFNN:

W_{2} (X_{2}) = ℘_{2}^{T} P_{2} (X_{2}) + ψ_{2} (X_{2}), | ψ_{2} (X_{2}) | < ψ_{M} .

(52)

Re-express (51) as:

\begin{array}{l} {\dot{V}}_{2} \leq & - k_{1} e_{1}^{2} - \frac{1}{2} \sum_{i = 3}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - (\frac{1}{λ_{2}} - \frac{1}{2}) ϕ_{2}^{2} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{γ_{1}}{d_{1}} {\tilde{β}}_{1} {\hat{β}}_{1} \\ + e_{2} [e_{3} + ϕ_{3} + u_{3} + Δ \hat{E} + ℘_{2}^{T} P_{2} (X_{2}) + ψ_{2} (X_{2}) - {\dot{u}}_{2 c} - 2 e_{2}] - ƛ V_{T} + ϕ_{3} {\dot{ϕ}}_{3} + {\bar{M}}_{2}^{2} \\ + μ_{1}^{2} + ψ_{M}^{2} / 4 - 1 / d_{2} {\tilde{β}}_{2} {\dot{\hat{β}}}_{2} . \end{array}

(53)

Similar to (36), the inequalities below can be obtained with

m_{1} = m_{2} = m_{3} = 1

{\begin{matrix} e_{2} ϕ_{3} \leq e_{2}^{2} + \frac{1}{4} ϕ_{3}^{2} \\ e_{2} W_{2} (X_{2}) = e_{2} [℘_{2}^{T} P_{2} (X_{2}) + ψ_{2} (X_{2})] \leq 1 / (4 μ_{2}^{2}) β_{2} e_{2}^{2} P_{2}^{T} P_{2} + μ_{2}^{2} + ψ_{M}^{2} / 4 + e_{2}^{2}, \end{matrix}

(54)

where

μ_{2}

represents the positive design parameter.

Integrating (54) into (53) produces:

\begin{array}{l} {\dot{V}}_{2} \leq & - k_{1} e_{1}^{2} - \frac{1}{2} \sum_{i = 3}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - (\frac{1}{λ_{2}} - \frac{1}{2}) ϕ_{2}^{2} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{γ_{1}}{d_{1}} {\tilde{β}}_{1} {\hat{β}}_{1} \\ + e_{2} (e_{3} + u_{3} + Δ \hat{E} - {\dot{u}}_{2 c}) - ƛ V_{T} + ϕ_{3} {\dot{ϕ}}_{3} + {\bar{M}}_{2}^{2} + \sum_{i = 1}^{2} μ_{i}^{2} + + \frac{ψ_{M}^{2}}{2} + \frac{ϕ_{3}^{2}}{4} \\ + \frac{1}{4 μ_{2}^{2}} e_{2}^{2} β_{2} P_{2}^{T} P_{2} - \frac{1}{d_{2}} {\tilde{β}}_{2} {\dot{\hat{β}}}_{2} . \end{array}

(55)

Analogous to (38), choose the stunning control law

u_{3}

and the adaptive law

{\dot{\hat{β}}}_{2}

as:

\begin{array}{l} u_{3} = - (k_{2} e_{2} + \frac{1}{4 μ_{2}^{2}} {\hat{β}}_{2} e_{2} P_{2}^{T} P_{2} + Δ \hat{E}) + {\dot{u}}_{2 c} \\ {\dot{\hat{β}}}_{2} = \frac{d_{2}}{4 μ_{2}^{2}} e_{2}^{2} P_{2}^{T} P_{2} - γ_{2} {\hat{β}}_{2}, \end{array}

(56)

where the design parameters

k_{2} > 0, γ_{2} > 0

Remark 9.

Up to this step, the external interference

Δ E

can be thoroughly approximated by the DO (18) with the stunning controller (56), thus the stable operation of the controlled system is further enhanced.

Pouring stunning control law

u_{3}

and the adaptive law

{\dot{\hat{β}}}_{2}

in (56) into (55) leads to:

\begin{array}{l} {\dot{V}}_{2} \leq & - \sum_{i = 1}^{2} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{i = 3}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - (\frac{1}{λ_{2}} - \frac{1}{2}) ϕ_{2}^{2} + \sum_{i = 1}^{2} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T \\ - ƛ V_{T} + e_{2} e_{3} + ϕ_{3} {\dot{ϕ}}_{3} + {\bar{M}}_{2}^{2} + \sum_{i = 1}^{2} μ_{i}^{2} + ψ_{M}^{2} / 2 + ϕ_{3}^{2} / 4 . \end{array}

(57)

Akin to (42), we can obtain:

ϕ_{3} {\dot{ϕ}}_{3} \leq - (\frac{1}{λ_{3}} - \frac{1}{4}) ϕ_{3}^{2} + {\bar{M}}_{3}^{2},

(58)

where the function

{\bar{M}}_{3} > 0

Substituting (58) into (57) produces:

\begin{array}{l} {\dot{V}}_{2} \leq & - \sum_{i = 1}^{2} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{i = 3}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{2} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T \\ - ƛ V_{T} + e_{2} e_{3} + \sum_{i = 1}^{2} μ_{i}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + ψ_{M}^{2} / 2 . \end{array}

(59)

Step 3. The design of the actual controller $u_{q}$ and the adaptive law ${\dot{\hat{β}}}_{3}$ .

With the state error

e_{3}

and estimation error

{\tilde{β}}_{3}

considered, the third Lyapunov function

V_{3}

is chosen as:

V_{3} = V_{2} + \frac{1}{2} e_{3}^{2} + \frac{1}{2 d_{3}} {\tilde{β}}_{3}^{2},

(60)

where

d_{3}

stands for the known positive constant.

Based on (24), derive

V_{3}

in (60) as:

{\dot{V}}_{3} = {\dot{V}}_{2} + e_{3} {\dot{e}}_{3} - \frac{1}{d_{3}} {\tilde{β}}_{3} {\dot{\hat{β}}}_{3} .

(61)

Combining (23) and (59), (61)becomes:

\begin{array}{l} {\dot{V}}_{3} \leq & - \sum_{i = 1}^{2} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{i = 3}^{4} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{2} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{ψ_{M}^{2}}{2} \\ - ƛ V_{T} + e_{3} {b_{1} x_{3} + b_{3} x_{2} + b_{2} x_{2} x_{4} + b_{4} u_{q} + Δ l_{3} [x (t - τ_{3})] + e_{2} - {\dot{u}}_{3 c}} + \sum_{i = 1}^{2} μ_{i}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} - \frac{1}{d_{3}} {\tilde{β}}_{3} {\dot{\hat{β}}}_{3} . \end{array}

(62)

Analogous to (30), the following inequality holds with

n = 4

e_{3} Δ l_{3} [x (t - τ_{3})] \leq 2 e_{3}^{2} + \frac{1}{2} \sum_{j = 1}^{4} ξ_{3 j}^{2} [x_{j} (t - τ_{3})] .

(63)

Pouring (63) into (62) leads to:

\begin{array}{l} {\dot{V}}_{3} \leq & - \sum_{i = 1}^{2} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{2} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{ψ_{M}^{2}}{2} \\ - ƛ V_{T} + e_{3} (b_{1} x_{3} + b_{3} x_{2} + b_{2} x_{2} x_{4} + b_{4} u_{q} + e_{2} + 2 e_{3} - {\dot{u}}_{3 c}) + \sum_{i = 1}^{2} μ_{i}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} - \frac{1}{d_{3}} {\tilde{β}}_{3} {\dot{\hat{β}}}_{3} . \end{array}

(64)

Design the unknown function

W_{3} (X_{3})

as:

W_{3} (X_{3}) = b_{1} x_{3} + b_{2} x_{2} x_{4} + b_{3} x_{2} + e_{2} + 3 e_{3},

(65)

where

e_{2} = x_{2} - α_{2 c}, e_{3} = x_{3} - α_{3 c}

. According to the composition of the unknown function

W_{3} (X_{3})

, it can be derived that

X_{3}

consists of the unknown and uncertain states

x_{2}, x_{3}, x_{4}, α_{2 c}

and

α_{3 c}

. Then

X_{3} = {[x_{2}, x_{3}, x_{4}, u_{2 c}, u_{3 c}]}^{T}

Consequently, rewrite (64) as:

\begin{array}{l} {\dot{V}}_{3} \leq & - \sum_{i = 1}^{2} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{2} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T \\ - ƛ V_{T} + e_{3} [W_{3} (X_{3}) + b_{4} u_{q} - e_{3} - {\dot{u}}_{3 c}] + \sum_{i = 1}^{2} μ_{i}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + ψ_{M}^{2} / 2 - 1 / d_{3} {\tilde{β}}_{3} {\dot{\hat{β}}}_{3} . \end{array}

(66)

Utilize an RBFNN to approximate

W_{3} (X_{3})

W_{3} (X_{3}) = ℘_{3}^{T} P_{3} (X_{3}) + ψ_{3} (X_{3}), | ψ_{3} (X_{3}) | ⩽ ψ_{M} .

(67)

Resembling (36), it generates:

e_{3} W_{3} (X_{3}) = e_{3} [℘_{3}^{T} P_{3} (X_{3}) + ψ_{3} (X_{3})] \leq \frac{1}{4 μ_{3}^{2}} β_{3} e_{3}^{2} P_{3}^{T} P_{3} + μ_{3}^{2} + \frac{ψ_{M}^{2}}{4} + e_{3}^{2},

(68)

where

μ_{3}

represents the positive design parameter.

Then, reperform (66) as:

\begin{array}{l} {\dot{V}}_{3} \leq & - \sum_{i = 1}^{2} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{2} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \frac{3}{4} ψ_{M}^{2} \\ - ƛ V_{T} + e_{3} (b_{4} u_{q} - {\dot{u}}_{3 c}) + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 1}^{3} μ_{i}^{2} + \frac{1}{4 μ_{3}^{2}} e_{3}^{2} β_{3} P_{3}^{T} P_{3} - \frac{1}{d_{3}} {\tilde{β}}_{3} {\dot{\hat{β}}}_{3} . \end{array}

(69)

Design the real controller

u_{q}

and the adaptive law

{\dot{\hat{β}}}_{3}

as:

\begin{array}{l} u_{q} = - \frac{1}{b_{4}} (k_{3} e_{3} + \frac{1}{4 μ_{3}^{2}} {\hat{β}}_{3} e_{3} P_{3}^{T} P_{3} - {\dot{u}}_{3 c}) \\ {\dot{\hat{β}}}_{3} = \frac{d_{3}}{4 μ_{3}^{2}} e_{3}^{2} P_{3}^{T} P_{3} - γ_{3} {\hat{β}}_{3}, \end{array}

(70)

where the design parameters

k_{3} > 0, γ_{3} > 0

At this point, the actual controller for the

q

-axis has been designed.

Inserting the real controller

u_{q}

and the adaptive law

{\dot{\hat{β}}}_{3}

in (70) into (69) derives:

\begin{array}{l} {\dot{V}}_{3} \leq & - \sum_{i = 1}^{3} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{3} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T \\ - ƛ V_{T} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 1}^{3} μ_{i}^{2} + \frac{3}{4} ψ_{M}^{2} . \end{array}

(71)

Step 4. The design of the actual controller $u_{d}$ and the adaptive law ${\dot{\hat{β}}}_{4}$ .

Consider the state error

e_{4}

and the estimation error

{\tilde{β}}_{4}

, then the Lyapunov function

V_{4}

is chosen as:

V_{4} = V_{3} + 1 / 2 e_{4}^{2} + 1 / (2 d_{4}) {\tilde{β}}_{4}^{2},

(72)

where

d_{4}

represents the positive design parameter.

Combining (24), differentiate

V_{4}

in (72) produces:

{\dot{V}}_{4} = {\dot{V}}_{3} + e_{4} {\dot{e}}_{4} - 1 / d_{4} {\tilde{β}}_{4} {\dot{\hat{β}}}_{4} .

(73)

Integrating (23) and (71) into (73) yields:

\begin{array}{l} {\dot{V}}_{4} \leq & - \sum_{i = 1}^{3} k_{i} e_{i}^{2} - \frac{1}{2} \sum_{j = 1}^{4} ξ_{i j}^{2} [x_{j} (t - τ_{i})] - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{3} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T \\ - ƛ V_{T} + e_{4} {c_{1} x_{4} + c_{2} x_{2} x_{3} + Δ l_{4} [x (t - τ_{4})] + c_{3} u_{d}} + 3 / 4 ψ_{M}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 1}^{3} μ_{i}^{2} \\ - \frac{1}{d_{4}} {\tilde{β}}_{4} {\dot{\hat{β}}}_{4} . \end{array}

(74)

Analogous to (30), the following inequality holds with

n = 4

e_{4} Δ l_{4} [x (t - τ_{4})] \leq 2 e_{4}^{2} + \frac{1}{2} \sum_{j = 1}^{4} ξ_{4 j}^{2} [x_{j} (t - τ_{4})] .

(75)

Design the unknown function

W_{4} (X_{4})

as:

W_{4} (X_{4}) = c_{1} x_{4} + c_{2} x_{2} x_{3} + 3 e_{4},

(76)

where

e_{4} = x_{4}

, the uncertain function

X_{4}

consists of

x_{2}, x_{3}

and

x_{4}

. Then

X_{4} = {[x_{2}, x_{3}, x_{4}]}^{T}

Then, (74) can be redrafted as:

\begin{array}{l} {\dot{V}}_{4} \leq & - \sum_{i = 1}^{3} k_{i} e_{i}^{2} - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} + \sum_{i = 1}^{3} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 3}^{3} μ_{i}^{2} \\ - ƛ V_{T} + e_{4} [W_{4} (X_{4}) + c_{3} u_{d} - e_{4}] + 3 / 4 ψ_{M}^{2} - 1 / d_{4} {\tilde{β}}_{4} {\dot{\hat{β}}}_{4} . \end{array}

(77)

Similarly, the unknown function

W_{4} (X_{4})

can be estimated by an RBFNN:

W_{4} (X_{4}) = ℘_{4}^{T} P_{4} (X_{4}) + ψ_{4} (X_{4}), | ψ_{4} (X_{4}) | ⩽ ψ_{M} .

(78)

Pouring (78) into (77) leads to:

\begin{array}{l} {\dot{V}}_{4} \leq & - \sum_{i = 1}^{3} k_{i} e_{i}^{2} - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} - \frac{1}{d_{4}} {\tilde{β}}_{4} {\dot{\hat{β}}}_{4} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 1}^{3} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T \\ - ƛ V_{T} + e_{4} [- e_{4} + c_{3} u_{d} + ℘_{4}^{T} P_{4} (X_{4}) + ψ_{4} (X_{4})] + \sum_{i = 1}^{3} μ_{i}^{2} + 3 / 4 ψ_{M}^{2} . \end{array}

(79)

Analogous to (36), it derives:

e_{4} W_{4} (X_{4}) = e_{4} [℘_{4}^{T} P_{4} (X_{4}) + ψ_{4} (X_{4})] \leq \frac{1}{4 μ_{4}^{2}} e_{4}^{2} β_{4} P_{4}^{T} P_{4} + μ_{4}^{2} + \frac{ψ_{M}^{2}}{4} + e_{4}^{2},

(80)

where

μ_{4}

stands for the positive design parameter.

Combining (80) with (79) generates:

\begin{array}{l} {\dot{V}}_{4} \leq & - \sum_{i = 1}^{3} k_{i} e_{i}^{2} - \sum_{i = 2}^{3} (\frac{1}{λ_{i}} - \frac{1}{2}) ϕ_{i}^{2} - \frac{1}{d_{4}} {\tilde{β}}_{4} {\dot{\hat{β}}}_{4} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 1}^{3} \frac{γ_{i}}{d_{i}} {\tilde{β}}_{i} {\hat{β}}_{i} + [1 - 16 \tanh^{2} (\frac{e_{1}}{σ_{1}})] T \\ - ƛ V_{T} + c_{3} e_{4} u_{d} + \sum_{i = 1}^{4} μ_{i}^{2} + ψ_{M}^{2} + \frac{1}{4 μ_{4}^{2}} e_{4}^{2} β_{4} P_{4}^{T} P_{4} . \end{array}

(81)

Devise the real controller

u_{d}

and the adaptive law

{\dot{\hat{β}}}_{4}

as:

\begin{array}{l} u_{d} = - \frac{1}{c_{3}} [k_{4} e_{4} + 1 / (4 μ_{4}^{2}) {\hat{β}}_{4} e_{4} P_{4}^{T} P_{4}] \\ {\dot{\hat{β}}}_{4} = \frac{d_{4}}{4 μ_{4}^{2}} e_{4}^{2} P_{4}^{T} P_{4} - γ_{4} {\hat{β}}_{4}, \end{array}

(82)

where the design constants

k_{4} > 0, γ_{4} > 0

Fusing the real controller

u_{d}

and the adaptive law

{\dot{\hat{β}}}_{4}

in (82) into (81) produces:

\begin{array}{l} \dot{V} \leq & - \sum_{i = 1}^{4} k_{i} e_{i}^{2} - \sum_{i = 2}^{3} (1 / λ_{i} - 1 / 2) ϕ_{i}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 1}^{4} γ_{i} / d_{i} {\tilde{β}}_{i} {\hat{β}}_{i} \\ - ƛ V_{T} + \sum_{i = 1}^{4} μ_{i}^{2} + ψ_{M}^{2} + [1 - 16 \tanh^{2} (e_{1} / σ_{1})] T . \end{array}

(83)

At the very beginning of this subsection, a funnel variable (16) based on modified PPF is designed in this paper to guarantee that the tracking error narrows down to a prescribed funnel type scope. The modified PPF avoids the demand for the precise initial value of the output variable. Among these steps, RBFNNs are utilized to approximate the unknown hard-to-compute functions (33), (50), (65) and (76) based on their infinite approximation capability. The Lyapunov Krasovskii functional eliminates the time delays

Δ l_{i} [x (t - τ_{i})]

in (3) with the aid of Lemma 1 and Remark 2. A DO (18) is introduced in step 2 to observe the uncertain matched and mismatched disturbance.

Up to this step, the design process of the FDSC with time delays and output constraints is finished. The effectiveness of the controller and the boundedness of all the variables need to be discussed in Section 4 and proved in Section 5.

4. Stability Analysis

For arbitrary predefined

p > 0

, the compact sets are described as:

{\begin{cases} Ω_{1} = {(e_{1}, ϕ_{2}, {\hat{β}}_{1}, ξ_{11}, ξ_{12}, \dots, ξ_{44}) : 2 e_{1}^{2} + 2 ϕ_{2}^{2} + \frac{2}{d_{1}} {\tilde{β}}_{1}^{2} + 4 V_{T} \leq 4 p} \\ Ω_{2} = {(\begin{array}{l} e_{1}, e_{2}, ϕ_{2}, ϕ_{3}, {\hat{β}}_{1}, {\hat{β}}_{2}, \\ ξ_{11}, ξ_{12}, \dots, ξ_{44} \end{array}) : 2 \sum_{i = 1}^{2} e_{i}^{2} + 2 \sum_{i = 2}^{3} ϕ_{i}^{2} + \sum_{i = 1}^{2} \frac{2}{d_{i}} {\tilde{β}}_{i}^{2} + 4 V_{T} \leq 4 p} \\ Ω_{3} = {(\begin{array}{l} e_{1}, e_{2}, e_{3}, ϕ_{2}, ϕ_{3}, {\hat{β}}_{1}, {\hat{β}}_{2}, \\ {\hat{β}}_{3}, ξ_{11}, ξ_{12}, \dots, ξ_{44} \end{array}) : 2 \sum_{i = 1}^{3} e_{i}^{2} + 2 \sum_{i = 2}^{3} ϕ_{i}^{2} + \sum_{i = 1}^{3} \frac{2}{d_{i}} {\tilde{β}}_{i}^{2} + 4 V_{T} \leq 4 p} \\ Ω_{4} = {(\begin{array}{l} e_{1}, \dots, e_{4}, ϕ_{2}, ϕ_{3}, {\hat{β}}_{1}, \dots, \\ {\hat{β}}_{4}, ξ_{11}, ξ_{12}, \dots, ξ_{44} \end{array}) : 2 \sum_{i = 1}^{4} e_{i}^{2} + 2 \sum_{i = 2}^{3} ϕ_{i}^{2} + \sum_{i = 1}^{4} \frac{2}{d_{i}} {\tilde{β}}_{i}^{2} + 4 V_{T} \leq 4 p} . \end{cases}

(84)

Theorem 1.

According to Assumption 1, the neural adaptive FDSC approach designed for the PMSM system (3) in this paper includes four controllers

u_{2}, u_{3}, u_{q}, u_{d}

and four adaptive laws

{\dot{\hat{β}}}_{i}, i = 1, \dots, 4

. For the initial conditions, if

Ω_{i}, i = 1, \dots, 4

- f_{1} (0) < s_{1} (0) < f_{1} (0)

and

x_{d} \in (- d, d)

are fulfilled, the core purposes of this paper will be realized.

Proof of Theorem 1.

This part proves the boundedness and the binding ranges of the variables, demonstrating the effectiveness of the designed controller in this paper for the PMSM system. □

4.1. Verification of the Boundedness for All Variables

Considering the state errors

e_{i}, i = 1, \dots, 4

, the filter errors

ϕ_{i}, i = 1, 2

, and the estimation errors

{\tilde{β}}_{i}, i = 1, \dots, 4

, design the whole Lyapunov function

V

as:

V = V_{4} = 1 / 2 \sum_{i = 1}^{4} e_{i}^{2} + 1 / 2 \sum_{i = 2}^{3} ϕ_{i}^{2} + \sum_{i = 1}^{4} 1 / (2 d_{i}) {\tilde{β}}_{i}^{2} .

(85)

Then, it can be taken from (83) the derivative of

V

in (85) as:

\begin{array}{l} \dot{V} \leq & - \sum_{i = 1}^{4} k_{i} e_{i}^{2} - \sum_{i = 2}^{3} (1 / λ_{i} - 1 / 2) ϕ_{i}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2} + \sum_{i = 1}^{4} γ_{i} / d_{i} {\tilde{β}}_{i} {\hat{β}}_{i} \\ - ƛ V_{T} + \sum_{i = 1}^{4} μ_{i}^{2} + ψ_{M}^{2} + [1 - 16 \tanh^{2} (e_{1} / σ_{1})] T . \end{array}

(86)

Reorganize (86) as:

\dot{V} \leq - b_{0} V + φ_{0} + a_{1},

(87)

where

b_{0} = \min {2 k_{1}, 2 k_{2}, 2 k_{3}, 2 k_{4}, 2 γ_{1}, 2 γ_{2}, 2 γ_{3}, 2 γ_{4}, ƛ, (2 / λ_{2} - 1), (2 / λ_{3} - 1)}

φ_{0} = \sum_{i = 1}^{4} μ_{i}^{2} + \sum_{i = 2}^{3} {\bar{M}}_{i}^{2}

+ ψ_{M}^{2}

, a_{1} = [1 - 16 \tanh^{2} (e_{1} / σ_{1})] T

To ensure every term in (87) satisfying the stability conditions, it needs to choose proper design parameters that make up

b_{0}

to guarantee

b_{0} > 0

. With the aid of Lemma 3, the value of

a_{1}

in (87) depends on two cases: (I) For

e_{1} \notin Ω_{e_{1}}

, we can deduce that

a_{1} \leq 0

due to

T (x) \geq 0

; (II) For

e_{1} \in Λ

, it can be inferred that

| e_{1} | \leq 0.2554 σ_{1}

and

σ_{1} > 0

. Based on the above discussion, we can conclude that the boundedness of

e_{1}

and

a_{1}

are both assured. Consequently, a constant

φ_{1} > 0

can be chosen to fulfill

| φ_{0} + a_{1} | < φ_{1}

. Then redraft (87) as:

\dot{V} \leq - b_{0} V + φ_{1} .

(88)

Multiplying both sides by

\exp (b_{0} t)

, then (88) becomes

d (V (t) \exp (b_{0} t)) / d t \leq φ_{1} \exp (b_{0} t)

. Integrating the inequality over

[0, t]

, it leads to:

V (t) \leq [V (0) - \frac{φ_{1}}{b_{0}}] \exp (b_{0} t) + \frac{φ_{1}}{b_{0}} \leq V (0) + \frac{φ_{1}}{b_{0}} .

(89)

Especially, the following results can be derived from (89):

\lim_{t \to \infty} | e_{i} | \leq \sqrt{\frac{2 φ_{1}}{b_{0}}}, i = 1, \dots, 4 .

(90)

It can be derived that

e_{i}, i = 1, \dots, 4

are bounded. Analogously, the boundedness of

ϕ_{2}, ϕ_{3}

, and

{\tilde{β}}_{i}, i = 1, \dots, 4

are ensured with (85). From (24), it can be concluded that

{\hat{β}}_{i}, i = 1, \dots, 4

are bounded. We can know that

s_{1}

is bounded from (16). Furthermore, it can be obtained that

x_{1}

is bounded according to

s_{1} = x_{1} - x_{d}

and the boundedness of

x_{d}

. Then, the virtual controller

u_{2}

is bounded based on (38). Consequently, the boundedness of

x_{2}

can be ensured from (20) and (21). Similarly, it can be derived that

u_{2}, u_{d}, u_{q}, x_{i}, i = 2, 3, 4

are also bounded. As a result, the designed controller demonstrates that all the signals of the closed-loop are uniformly bounded.

4.2. Deduction of the Binding Ranges for Variables

According to (85), one derives:

e_{i}^{2} \leq 2 [V (0) - \frac{φ_{1}}{b_{0}}] e^{- b_{0} t} + \frac{2 φ_{1}}{b_{0}}, i = 2, 3, 4 .

(91)

Then we can get that

| e_{i} | \leq \sqrt{2 [V (0) - φ_{1} / b_{0}] \exp (- b_{0} t) + 2 φ_{1} / b_{0}}

| e_{i} | \leq \sqrt{2 φ_{1} / b_{0}}

when

V (0) = φ_{1} / b_{0}

. If

V (0) \neq φ_{1} / b_{0}

, for arbitrary known

\sqrt{2 [V (0) - φ_{1} / b_{0}] \exp (- b_{0} t) + 2 φ_{1} / b_{0}} > 2 \sqrt{φ_{1} / b_{0}}

, there exists

T

for each

t > T

, it produces

\sqrt{2 [V (0) - φ_{1} / b_{0}] \exp (- b_{0} t) + 2 φ_{1} / b_{0}}

. When

t \to \infty

, the transformation error

| e_{i} | \leq \sqrt{2 φ_{1} / b_{0}}

According to (16),

η_{1} \to \pm \infty

when and only when

s_{1} \to \pm f_{1}

, it means that if

e_{1} \to \pm f_{1}

Δ f_{i}

will approach infinite. For the arbitrary starting value

Δ f_{i}

, it satisfies

| s_{1} (0) | < f_{1} (0)

. For arbitrary

t > 0

, it derives

| s_{1} (t) | < f_{1} (t)

. According to

s_{1} = x_{1} - x_{d}

, we can obtain the arbitrary initial condition

x_{d} (0) - f_{1} (0) < x_{1} (0) < x_{d} (0) + f_{1} (0)

and other conditions

x_{1} \in Π_{x_{1}} : = {x_{1} \in R :

y_{d} (t) - f_{1} (t) < x_{1} (t) < y_{d} (t) + f_{1} (t)}

with

t > 0

Remark 10.

The quality of the FDSC is judged by the value of

e_{1}

. Hence, it is of great importance to choose parameters based on the selection principle. The value of

e_{1}

depends on

φ_{1}

and

b_{0}

. When

φ_{1}

decreases and

b_{0}

increases,

e_{1}

will approach zero. Furthermore, we can increase

ψ_{M}, Γ_{1}, \sum_{i = 1}^{4} μ_{i}, \sum_{i = 2}^{3} {\bar{M}}_{i}

and decrease

ƛ, ε_{2}, 1 / λ_{2}, 1 / λ_{3}, k_{i}, γ_{i}, i = 1, \dots, 4

in (87) to achieve excellent control performance of the system.

5. Simulation and Comparison Results

5.1. Design of Controllers

To verify the effectiveness of the FDSC method, this subsection provides (3) with two simulation cases: (I) The case 1 contains time delays; (II) The case 2 ignores time delays (

Δ f_{i} = 0, i = 1, 2, 3, 4

). PID and NDSC approaches are used as comparison substrates in each case to more visually illustrate the superiority of the FDSC solution.

Select the inherent PMSM parameters as:

J = 0.003798 Kg \cdot m^{2}, B = 0.001158 N \cdot m / (rad / s),

T_{L} = 1.5, φ = 0.1245 Wb, L_{d} = 0.00285 H, n_{p} = 3, L_{q} = 0.00315 H, R_{s} = 0.68 Ω .

The desired signal and the disturbance function are chosen as

x_{d} = 0.02 \sin (2 t) + 0.1

and

Δ E = 40 x_{2} \sin (2 t)

, respectively.

5.1.1. FDSC

Based on Remark 10 and trial and error, the detailed FDSC’s design parameters as selected and shown in Table 2 to ensure the tracking error

e_{1}

small enough:

5.1.2. NDSC

By virtual of [48], select the following variables of NDSC:

{\begin{cases} e_{1} = x_{1} - x_{d} e_{2} = x_{3} - ϑ_{3 c} e_{2} = x_{2} - ϑ_{2 c} e_{4} = x_{4} \\ {\dot{u}}_{2 c} = 1 / ε_{2} (e_{2} - e_{2 c}) {\dot{u}}_{3 c} = 1 / ε_{3} (u_{3} - u_{3 c}) \\ u_{2} = - k_{1} e_{1} + {\dot{x}}_{d} u_{3} = J / a_{1} (- k_{2} ℓ_{2} + {\dot{u}}_{2 c} - {\hat{β}}_{2}^{T} P_{2} (X_{2})) \\ u_{q} = 1 / b_{4} (- k_{3} e_{3} + {\dot{ϑ}}_{3 c} - {\hat{β}}_{3}^{T} P_{3} (X_{3})) u_{d} = J / c_{3} (- k_{4} e_{4} - {\hat{β}}_{4}^{T} P_{4} (X_{4})) \\ {\dot{\hat{β}}}_{2} = {\underline{χ}}_{2} [P_{2} (X_{2}) e_{2} - γ_{2} {\hat{β}}_{2}] {\dot{\hat{β}}}_{3} = {\underline{χ}}_{3} [P_{3} (X_{3}) e_{3} - γ_{3} {\hat{β}}_{3}] \\ {\dot{\hat{β}}}_{4} = {\underline{χ}}_{4} [P_{4} (X_{4}) e_{4} - γ_{4} {\hat{β}}_{4}], \end{cases}

(92)

based on Remark 10 and trial and error, the design parameters are chosen to guarantee the tracking error small enough:

k_{1} = 30, k_{2} = k_{3} = k_{4} = 80, {\underline{χ}}_{2} = {\underline{χ}}_{3} = {\underline{χ}}_{4} = 10, γ_{2} = γ_{3} = γ_{4} = 0.09, λ_{2} =

λ_{3} = 0.01

5.1.3. PID

According to [12], the following PID controller was chosen with

u_{d} = 0

u_{q} = k_{p} s_{1} + k_{i} \int_{0}^{t} s_{1} d τ + k_{d} d (s_{1}) / d t,

(93)

based on Remark 10 and trial and error, the design parameters are chosen to guarantee the tracking error small enough:

k_{p} = 20, k_{i} = 0.05, k_{d} = 1.5

Meanwhile, Table 3 lists three quantitative indicators to compare FDSC scheme with the NDSC and PID methods.

(a): Integration over the absolute value of the error (IAE):

J_{I A E} = \int_{0}^{t} | ℓ (τ) | d τ .

(94)

(b): Integration over time and the absolute value of the error (ITAE):

J_{I T A E} = \int_{0}^{t} τ | ℓ (τ) | d τ .

(95)

(c): Integration over squared error (ISE):

J_{I S E} = \int_{0}^{t} ℓ^{2} (τ) d τ .

(96)

Upon the above discussions, consider the two cases as:

Case 1: Select the time-delay terms as:

{\begin{cases} Δ f_{1} (x (t - τ_{1})) = 10 x_{1}^{4} (t - τ_{1}) x_{2} (t - τ_{1}) x_{3} (t - τ_{1}) x_{4} (t - τ_{1}) \\ Δ f_{2} (x (t - τ_{2})) = 12 x_{1}^{4} (t - τ_{2}) x_{2}^{4} (t - τ_{2}) x_{3} (t - τ_{2}) x_{4} (t - τ_{2}) \\ Δ f_{3} (x (t - τ_{3})) = 14 x_{1}^{4} (t - τ_{3}) x_{2}^{4} (t - τ_{3}) x_{3}^{4} (t - τ_{3}) x_{4} (t - τ_{3}) \\ Δ f_{4} (x (t - τ_{4})) = 16 x_{1}^{4} (t - τ_{4}) x_{2}^{4} (t - τ_{4}) x_{3}^{4} (t - τ_{4}) x_{4}^{4} (t - τ_{4}) . \end{cases}

(97)

Design the time-varying funnel type boundaries as

f_{1} (t) = e^{- 2 t} + 0.1 t / (2 t + 2),

and select the original values of the state variables as

x_{i} (0) = 0.01, i = 1, \dots, 4 .

Each RBFNN consists of 11 nodes with the center positioned in the interval

[- 11, 11]

, and the width is 10.

Case 2: The time-delay terms are

Δ f_{i} = 0, i = 1, 2, 3, 4 .

In other words, the time-delays are not considered.

5.2. Simulation Comparison Results

The simulation comparison results are shown in Figure 3, Figure 4, Figure 5, Figure 6 and Figure 7, every figure contains two sub-figures: Case 1 and Case 2. The simulation results will prove the superiority of the FDSC scheme, and justify whether the Lyapunov-Krasovskii functional is effective.

It can be seen from Figure 3 that the FDSC method ensures the best convergence of output signal

x_{1}

when compared with PID and NDSC methods no matter in Case 1 or Case 2. And the output signal

x_{1}

based on the FDSC method is limited with the funnel prescribed boundaries when the output signal

x_{1}

based on the NDSC method is out of the boundary. The curve of output signal

x_{1}

in case 1 is almost the same as it in case 2, which means that Lyapunov-Krasovskii functional can effectively address the time delays.

It is demonstrated from Figure 4 that the FDSC method has the smallest tracking error

s_{1}

among three control schemes no matter in Case 1 or Case 2. And the tracking error

s_{1}

based on the FDSC method is limited by the funnel prescribed boundaries when compared with the NDSC method. The curves of tracking error

s_{1}

are almost the same in the two cases, showing the effectiveness of the Lyapunov-Krasovskii functional.

It is shown from Figure 5 that the curve of the state variable

x_{2}

in Case 1 is almost the same as it in Case 2. It means that the Lyapunov-Krasovskii functional can effectively address the time delays. The controller designed in is paper guarantees that the state variable

x_{2}

in the closed-loop is controllable. The same conclusion can be drawn in Figure 6.

It can be seen from Figure 6 that the curves of the state variables

i_{d}

and

i_{q}

in case 1 are almost the same as them in case 2. It means that the Lyapunov-Krasovskii functionals can effectively address the time delays. The controller designed in is paper guarantees the stable operation of the PMSM systems with time delays. And the

d -

axis current

i_{d}

and the

q -

axis current

i_{q}

in the closed-loop are controllable. The same conclusion can be drawn in the Figure 7.

Within the funnel-type boundaries, Figure 3 gives the curves of output signal

x_{1}

and ideal signal

x_{d}

under three schemes (FDSC, PID, and NDSC). Figure 4 exhibits the curves of tracking error

s_{1}

for different scenarios. It can be concluded from Figure 3 and Figure 4 that FDSC has a better convergency rate than NDSC, and the tracking error

s_{1}

of FDSC method is minimal compared to the other two options without breaking through funnel-type boundaries even considering the time delays. Figure 5 displays the curves of state

x_{2}

in the FDSC approach. Figure 6 illustrates the trajectories of states

i_{d}

and

i_{q}

, and Figure 7 shows the curves of controllers

u_{d}

and

u_{q}

. According to Figure 3, Figure 4, Figure 5, Figure 6 and Figure 7, the states of the FDSC have extremely rapid response capability compared to NDSC and PID with time delays.

6. Conclusions

The FDSC with disturbance-observer has been investigated and suggested to resolve the position tracking control challenge for the PMSM with time delays here. At the beginning of this paper, an assumption and several lemmas are employed to uphold the control thesis and the design procedure of the funnel controller. With the aid of them, the inequalities in Section 3 can be simplified. Secondly, this paper introduces promising tools to eliminate the nonlinear uncertainties in PMSMs. For instance, this paper introduces RBFNNs to estimate the unknown functions because they can approximate the unknown hard-to-calculated functions that are in a closed set at any precision. For the unknown mismatched and matched external disturbance, this paper utilizes a finite-time second-order DO. The finite-time DO can approximate the unknown interference with a finite time speed. To remove the limitations of the output variable, a funnel variable is designed in this paper without the demand for precise initial values of the state variable. This paper uses the first-order filters to filtrate the multiple derivatives of the virtual controllers so that the “explosion of complexity” can be addressed. The Lyapunov-Krasovskii functionals are devised in the controller design procedure of this paper to deal with time delays and guarantee the stable operation of the PMSMs. Thirdly, with the aid of RBFNNs, the finite-time DO, first-order filters, and the Lyapunov-Krasovskii functionals, the controller design procedure can be finished based on the backstepping method. Despite that the devised FDSC method can provide outstanding tracking performance shown in Figure 3, Figure 4, Figure 5, Figure 6 and Figure 7 for the output-constrained PMSM, the finite-time response rates need to be further improved by combining it with finite-time stability theory. In further research later, we will further study how to improve the finite time convergence of tracking error based on the FDSC scheme. For PMSMs are widely applied in many significant fields, the neural adaptive FDSC scheme can be extended to vehicles, electric elevators, and machine tools.

Author Contributions

Conceptualization, M.L. and J.Z.; methodology, M.L.; software, F.W.; validation, T.Z., formal analysis, J.Z.; investigation, T.Z.; resources, S.L.; writing—original draft preparation, M.L.; writing—review and editing, M.L.; supervision, J.Z.; project administration, S.L.; funding acquisition, S.L. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported in part by the National Key technologies Research and Development Program of China under Grants No. 32020YFB1713300 and No. 2018AAA0101803, in part by the Natural Science Foundation of China No. 51635003 and No. 61863005, in part by the Guizhou Province Postgraduate Innovation Fund No. YJSKYJJ(2021)030, in part by Guizhou Provincial Science and Technology Projects No. (ZK [2022] 142), in part by the Guizhou optoelectronic information and intelligent application International Joint Research Center No. Qiankehe platform talents No. 5802[2019], in part by the Foundation of Key Laboratory of Advanced Manufacturing Technology, Ministry of Education, Guizhou University No. GZUAMT2021KF [11], in part by the Science and Technology Incubation Planning Project of Guizhou University No. [2020]75.

Conflicts of Interest

The authors declare no conflict of interest.

Nomenclature

Acronyms
PMSM	Permanent magnetic synchronous motor
SVPWM	Space vector pulse width modulation
NN	Neural network
RBFNN	Radial basis function neural network
DO	Disturbance observer
PPF	Prescribed performance function
FDSC	Funnel dynamic surface control
NDSC	Neural dynamic surface control
PID	Proportion integral differential
IAE	Integration over the absolute value of the error
ITAE	Integration over time and the absolute value of the error
ISE	Integration over squared error
Variables
$x_{i}, i = 1, \dots, 4$	State variables
$Δ E$	External disturbance term
$Δ l_{i} (x (t - τ_{i})), i = 1, \dots, 4$	Time delay terms
$℘^{*} \in R^{l}$	Desired weight vector of RBFNN
$\hat{℘}$	Updated weight vector
$ψ (X)$	Estimation error
$ν_{i} = {[ν_{i 1}, \dots, ν_{i m}]}^{T}$	Receptive field centers
$β_{i}, i = 1, \dots, 4$	Unknown variables
$u_{i c}, i = 2, 3$	The output of first-order filters
$u_{i}$	Virtual controllers, the input of first-order filters
$ϕ_{i}, i = 2, 3$	First-order filter errors
$η_{1}$	Modified funnel variable
Functions
$f_{1} (t)$	Positive funnel prescribed performance function
$F_{k} (\cdot)$	Gain function over time
$G (t)$	Scaling function
$ξ_{i j} > 0$	$Smooth functions in Lyapunov - Krasovskii functional V_{T}$
Parameters
$l > 1$	Node numbers of RBFNN
$χ_{i}$	Widths of the Gaussian functions

References

Sirimanna, S.; Thanatheepan, B.; Lee, D.; Agrawal, S.; Yu, Y.; Wang, Y.; Anderson, A.; Banerjee, A.; Haran, K. Comparison of electrified aircraft propulsion drive systems with different electric motor topologies. J. Propul. Power 2021, 37, 733–747. [Google Scholar] [CrossRef]
Foitzik, S.; Doppelbauer, M. Fault tolerant control of a three-phase PMSM by limiting the heat of an inter-turn fault. IET Electr. Power Appl. 2022, 16, 158–168. [Google Scholar] [CrossRef]
Qiu, Z.; Chen, Y.; Guan, Y.; Kang, Y. Vibration noise suppression algorithm of permanent magnet synchronous motor for new energy vehicles. Int. J. Heavy Veh. Syst. 2022, 29, 48–70. [Google Scholar] [CrossRef]
Hu, D.; Yang, L.; Yi, F.; Hu, L.; Yang, Q.; Zhou, J. Optimization of speed response of super-high-speed electric air compressor for hydrogen fuel cell vehicle considering the transient current. Int. J. Hydrogen Energy 2021, 46, 27183–27192. [Google Scholar] [CrossRef]
Zhang, J.; Wang, S.; Li, S.; Zhou, P. Adaptive neural dynamic surface control for the chaotic PMSM system with external disturbances and constrained output. Recent Adv. Electr. Electron. Eng. 2020, 13, 894–905. [Google Scholar]
Swaroop, D.; Hedrick, J.K.; Yip, P.P.; Gerdes, J.C. Dynamic surface control for a class of nonlinear systems. IEEE Trans. Autom. Control 2000, 45, 1893–1899. [Google Scholar] [CrossRef] [Green Version]
Lv, Z.; Ma, Y.; Liu, J.; Yu, J. Full-state constrained adaptive fuzzy finite-time dynamic surface control for pmsm drive systems. Int. J. Fuzzy Syst. 2021, 23, 804–815. [Google Scholar] [CrossRef]
Jiang, Y.; Xu, W.; Mu, C.; Zhu, J.; Dian, R. An improved third-order generalized integral flux observer for sensorless drive of PMSMs. IEEE Trans. Ind. Electron. 2019, 66, 9149–9160. [Google Scholar] [CrossRef]
Ke, D.; Wang, F.; He, L.; Li, Z. Predictive current control for PMSM systems using extended sliding mode observer with hurwitz-based power reaching law. IEEE Trans. Power Electron. 2021, 36, 7223–7232. [Google Scholar] [CrossRef]
Lian, C.; Xiao, F.; Liu, J.; Gao, S. Analysis and compensation of the rotor position offset error and time delay in field-oriented-controlled PMSM drives. IET Power Electron. 2020, 13, 1911–1918. [Google Scholar] [CrossRef]
Ali, N.; Rehman, A.U.; Alam, W.; Maqsood, H. Disturbance observer based robust sliding mode control of permanent magnet synchronous motor. J. Electr. Eng. Technol. 2019, 14, 2531–2538. [Google Scholar] [CrossRef]
Raj, F.V.A.; Kannan, V.K. Particle swarm optimized deep convolutional neural sugeno-takagi fuzzy PID controller in permanent magnet synchronous motor. Int. J. Fuzzy Syst. 2022, 24, 180–201. [Google Scholar] [CrossRef]
Liu, M.; Lu, B.; Li, Z.; Jiang, H.; Hu, C. Fixed-time synchronization control of delayed dynamical complex networks. Entropy 2021, 23, 1610. [Google Scholar] [CrossRef] [PubMed]
Xu, Y.; Li, S.; Zou, J. Integral sliding mode control based deadbeat predictive current control for PMSM drives with disturbance rejection. IEEE Trans. Power Electron. 2022, 37, 2845–2856. [Google Scholar] [CrossRef]
Li, R.; Zhu, Q.; Yang, J.; Narayan, P.; Yue, X. Disturbance-Observer-Based U-Control (DOBUC) for nonlinear dynamic systems. Entropy 2021, 23, 1625. [Google Scholar] [CrossRef]
Wang, Y.; Yu, H.; Liu, Y. Speed-current single-loop control with overcurrent protection for PMSM based on time-varying nonlinear disturbance observer. IEEE Trans. Ind. Electron. 2022, 69, 179–189. [Google Scholar] [CrossRef]
Cho, H.J.; Kwon, Y.C.; Sul, S.K. Time-optimal voltage vector transition scheme for six-step operation of PMSM. IEEE Trans. Power Electron. 2021, 36, 5724–5735. [Google Scholar] [CrossRef]
Dai, C.; Guo, T.; Yang, J.; Li, S. A disturbance observer-based current-constrained controller for speed regulation of PMSM systems subject to unmatched disturbances. IEEE Trans. Ind. Electron. 2021, 68, 767–775. [Google Scholar] [CrossRef]
Ge, Y.; Yang, L.; Ma, X. HGO and neural network based integral sliding mode control for PMSMs with uncertainty. J. Power Electron. 2020, 20, 1206–1221. [Google Scholar] [CrossRef]
Xu, M.; Huo, Z.; Li, S.; Jiang, L. Active steering PMSM speed control with wavelet neural network. Int. J. Veh. Des. 2020, 82, 64–74. [Google Scholar] [CrossRef]
Xu, B.; Zhang, L.; Ji, W. Improved non-singular fast terminal sliding mode control with disturbance observer for PMSM drives. IEEE Trans. Transp. Electrif. 2021, 7, 2753–2762. [Google Scholar] [CrossRef]
Hou, Q.; Ding, S.; Yu, X. Composite super-twisting sliding mode control design for PMSM speed regulation problem based on a novel disturbance observer. IEEE Trans. Energy Convers. 2021, 36, 2591–2599. [Google Scholar] [CrossRef]
Shao, X.; Si, H.; Zhang, W. Event-triggered neural intelligent control for uncertain nonlinear systems with specified-time guaranteed behaviors. Neural Comput. Appl. 2021, 33, 5771–5791. [Google Scholar] [CrossRef]
Zheng, Z.; Lau, G.-K.; Xie, L. Event-triggered control for a saturated nonlinear system with prescribed performance and finite-time convergence. Int. J. Robust Nonlinear Control 2018, 28, 5312–5325. [Google Scholar] [CrossRef]
Shi, Y.; Shao, X. Neural adaptive appointed-time control for flexible air-breathing hypersonic vehicles: An event-triggered case. Neural Comput. Appl. 2021, 33, 9545–9563. [Google Scholar] [CrossRef]
Zeng, Q.; Zhao, J. Event-triggered adaptive finite-time control for active suspension systems with prescribed performance. IEEE Trans. Ind. Inf. 2021. [Google Scholar] [CrossRef]
Shao, X.; Si, H.; Zhang, W. Fuzzy wavelet neural control with improved prescribed performance for MEMS gyroscope subject to input quantization. Fuzzy Sets Syst. 2020, 411, 136–154. [Google Scholar] [CrossRef]
Wang, S.; Li, S.; Chen, Q.; Ren, X.; Yu, H. Funnel tracking control for nonlinear servo drive systems with unknown disturbances. ISA Trans. 2021. [Google Scholar] [CrossRef]
Liu, C.; Wang, H.; Liu, X.; Zhou, Y. Adaptive finite-time fuzzy funnel control for nonaffine nonlinear systems. IEEE Trans. Syst. Man Cybern. Syst. 2021, 51, 2894–2903. [Google Scholar] [CrossRef]
Wang, S.; Ren, X.; Na, J.; Zeng, T. Extended-state-observer-based funnel control for nonlinear servomechanisms with prescribed tracking performance. IEEE Trans. Autom. Sci. Eng. 2017, 14, 98–108. [Google Scholar] [CrossRef]
Bao, J.; Wang, H.; Xiaoping Liu, P. Adaptive finite-time tracking control for robotic manipulators with funnel boundary. Int. J. Adapt. Control Signal Process. 2020, 34, 575–589. [Google Scholar] [CrossRef]
Zhu, Q.; Xiong, L.; Liu, H. A robust speed controller with smith predictor for a PMSM drive system with time delay. Int. J. Control Autom. Syst. 2017, 15, 2448–2454. [Google Scholar] [CrossRef]
Xing, H.; Ploeg, J.; Nijmeijer, H. Smith predictor compensating for vehicle actuator delays in cooperative ACC systems. IEEE Trans. Veh. Technol. 2019, 68, 1106–1115. [Google Scholar] [CrossRef]
Vadivel, R.; Joo, Y.H. Reliable fuzzy H_∞ control for permanent magnet synchronous motor against stochastic actuator faults. IEEE Trans. Syst. Man Cybern. Syst. 2021, 51, 2232–2245. [Google Scholar] [CrossRef]
Shanmugam, L.; Joo, Y.H. Design of interval type-2 fuzzy-based sampled-data controller for nonlinear systems using novel fuzzy Lyapunov functional and its application to PMSM. IEEE Trans. Syst. Man Cybern. Syst. 2021, 51, 542–551. [Google Scholar] [CrossRef]
Gong, C.; Hu, Y.; Gao, J.; Wang, Y.; Yan, L. An improved delay-suppressed sliding-mode observer for sensorless vector-controlled PMSM. IEEE Trans. Ind. Electron. 2020, 67, 5913–5923. [Google Scholar] [CrossRef]
Si, W.; Qi, L.; Hou, N.; Dong, X. Finite-time adaptive neural control for uncertain nonlinear time-delay systems with actuator delay and full-state constraints. Int. J. Syst. Sci. 2019, 50, 726–738. [Google Scholar] [CrossRef]
Wei, L.; Qian, C. Adaptive control of nonlinearly parameterized systems: The smooth feedback case. IEEE Trans. Autom. Control 2001, 47, 1249–1266. [Google Scholar]
Qian, C.; Wei, L. Non-Lipschitz continuous stabilizers for nonlinear systems with uncontrollable unstable linearization. Syst. Control Lett. 2015, 42, 185–200. [Google Scholar] [CrossRef]
Fang, W.; Bing, C.; Chong, L.; Jing, Z.; Meng, X. Adaptive neural network finite-time output feedback control of quantized nonlinear systems. IEEE Trans. Cybern. 2018, 48, 1839–1848. [Google Scholar]
Wang, H.; Sun, W.; Liu, P.X. Adaptive intelligent control of nonaffine nonlinear time-delay systems with dynamic uncertainties. IEEE Trans. Syst. Man Cybern. Syst. 2017, 47, 1474–1485. [Google Scholar] [CrossRef]
Yang, X.; Zheng, X. Adaptive nn backstepping control design for a 3-dof helicopter: Theory and experiments. IEEE Trans. Ind. Electron. 2020, 67, 3967–3979. [Google Scholar] [CrossRef]
Liu, H.; Pan, Y.; Cao, J.; Wang, H.; Zhou, Y. Adaptive neural network backstepping control of fractional-order nonlinear systems with actuator faults. IEEE Trans. Neural Networks Learn. Syst. 2020, 31, 5166–5177. [Google Scholar] [CrossRef]
Wang, H.; Li, M.; Zhang, C.; Shao, X. Event-based prescribed performance control for dynamic positioning vessels. IEEE Trans. Circuits Syst. II Express Briefs 2021, 68, 2548–2552. [Google Scholar] [CrossRef]
Li, F.; Liu, Y. Prescribed-performance control design for pure-feedback nonlinear systems with most severe uncertainties. SIAM J. Control Optim. 2018, 56, 517–537. [Google Scholar] [CrossRef]
Levant, A.; Livne, M. Globally convergent differentiators with variable gains. Int. J. Control 2018, 91, 1994–2008. [Google Scholar] [CrossRef]
Zhao, K.; Song, Y.; Zhang, Z. Tracking control of MIMO nonlinear systems under full state constraints: A single-parameter adaptation approach free from feasibility conditions. Automatica 2019, 107, 52–60. [Google Scholar] [CrossRef]
Yue, H.; Wei, Z.; Chen, Q.; Zhang, X. Dynamic surface control for a class of nonlinearly parameterized systems with input time delay using neural network. J. Franklin Inst. 2020, 357, 1961–1986. [Google Scholar] [CrossRef]

Figure 1. The curves of PPF and improved PPF (

π = 3, 5, 10

Figure 1. The curves of PPF and improved PPF (

π = 3, 5, 10

Figure 2. Overview of the control architecture for the PMSM system.

Figure 3. The output signal

x_{1}

and the ideal signal

x_{d}

curves for the FDSC, PID, and NDSC.

Figure 3. The output signal

x_{1}

and the ideal signal

x_{d}

curves for the FDSC, PID, and NDSC.

Figure 4. The tracking error responses

s_{1}

for the FDSC, PID, and NDSC.

Figure 4. The tracking error responses

s_{1}

for the FDSC, PID, and NDSC.

Figure 5. The state variable

x_{2}

responses.

Figure 5. The state variable

x_{2}

responses.

Figure 6. The state variables

i_{d}

and

i_{q}

responses.

Figure 6. The state variables

i_{d}

and

i_{q}

responses.

Figure 7. The controllers

u_{d}

and

u_{q}

trajectories.

Figure 7. The controllers

u_{d}

and

u_{q}

trajectories.

Table 1. Parameters of PMSM.

Parameters	Descriptions	Units
$θ$	Rotor angular	$rad$
$ω$	Rotor angular velocity, $\dot{θ}$	$rad / s$
$\dot{ω}$	First-order derivative of rotor angular velocity	${rad / s}^{2}$
$i_{d}, i_{q}$	Currents of $d - q$ axis	$A$
$u_{d}, u_{q}$	Voltages of $d - q$ axis	$V$
$L_{d}, L_{q}$	Stator inductances of $d - q$ axis	$H$
$J$	Inertia rotor moment	$kg \cdot m^{2}$
$B$	Friction coefficient	$N \cdot m / (rad / s)$
$φ$	Inertia magnet flux linkage	$Wb$
$R_{s}$	Armature resistance	$Ω$
$n_{p}$	Pole pairs
$T_{L}$	Load torque	$N \cdot m$

Table 2. FDSC’s design parameters.

Parameters	Values	Parameters	Values	Parameters	Values
$u_{2 c} (0)$	0	${\hat{β}}_{1} (0)$	−0.05	$d_{1}$	0.65
$u_{3 c} (0)$	0.5	${\hat{β}}_{2} (0)$	0	$d_{2}$	0.95
$ε_{2}$	0.1	${\hat{β}}_{3} (0)$	−0.5	$d_{3}$	0.75
$ε_{3}$	0.01	${\hat{β}}_{4} (0)$	0	$d_{4}$	35
$k_{1}$	10	$γ_{1}$	60	$μ_{1}$	0.06
$k_{2}$	20	$γ_{2}$	4	$μ_{2}$	0.3
$k_{3}$	20	$γ_{3}$	60	$μ_{3}$	0.1
$k_{4}$	1200	$γ_{4}$	0.4	$μ_{4}$	0.01

Table 3. Comparative numerical results of performance indicators.

Indicators		FDSC	PID	NDSC
Case 1	ISE	0.000661	0.079390	0.244500
	ITAE	0.005941	8.171000	2.801000
	IAE	0.012980	1.091000	0.978600
Case 2	ISE	0.000661	0.079010	0.247000
	ITAE	0.005894	8.151000	2.764000
	IAE	0.012980	1.089000	0.970700

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Li, M.; Li, S.; Zhang, J.; Wu, F.; Zhang, T. Neural Adaptive Funnel Dynamic Surface Control with Disturbance-Observer for the PMSM with Time Delays. Entropy 2022, 24, 1028. https://doi.org/10.3390/e24081028

AMA Style

Li M, Li S, Zhang J, Wu F, Zhang T. Neural Adaptive Funnel Dynamic Surface Control with Disturbance-Observer for the PMSM with Time Delays. Entropy. 2022; 24(8):1028. https://doi.org/10.3390/e24081028

Chicago/Turabian Style

Li, Menghan, Shaobo Li, Junxing Zhang, Fengbin Wu, and Tao Zhang. 2022. "Neural Adaptive Funnel Dynamic Surface Control with Disturbance-Observer for the PMSM with Time Delays" Entropy 24, no. 8: 1028. https://doi.org/10.3390/e24081028

APA Style

Li, M., Li, S., Zhang, J., Wu, F., & Zhang, T. (2022). Neural Adaptive Funnel Dynamic Surface Control with Disturbance-Observer for the PMSM with Time Delays. Entropy, 24(8), 1028. https://doi.org/10.3390/e24081028

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

Neural Adaptive Funnel Dynamic Surface Control with Disturbance-Observer for the PMSM with Time Delays

Abstract

1. Introduction

2. System Formulation and Preliminaries

2.1. System Statement

2.2. Neural Network Systems and Function Approximation

3. Design of Neural Adaptive Funnel Control

3.1. Funnel Control with Improved Prescribed Performance Function

3.2. Disturbance Observer

3.3. Neural Adaptive Funnel Controller Design

4. Stability Analysis

4.1. Verification of the Boundedness for All Variables

4.2. Deduction of the Binding Ranges for Variables

5. Simulation and Comparison Results

5.1. Design of Controllers

5.1.1. FDSC

5.1.2. NDSC

5.1.3. PID

5.2. Simulation Comparison Results

6. Conclusions

Author Contributions

Funding

Conflicts of Interest

Nomenclature

References

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI