Considering the backlash nonlinearity and parameter time-varying characteristics in electromechanical actuators, a chattering-free sliding-mode control strategy is proposed in this paper to regulate the rudder angle and suppress unknown external disturbances. Different from most existing backlash compensation methods, a special continuous function is addressed to approximate the backlash nonlinear dead-zone model. Regarding the approximation error, unmodeled dynamics, and unknown external disturbances as a disturbance-like term, a strict feedback nonlinear model is established. Based on this nonlinear model, a chattering-free nonsingular terminal sliding-mode controller is proposed to achieve the rudder angle tracking with a chattering elimination and tracking dynamic performance improvement. A Lyapunov-based proof ensures the asymptotic stability and finite-time convergence of the closed-loop system. Experimental results have verified the effectiveness of the proposed method.
National Natural Science Foundation of China514071431. Introduction
Electromechanical actuator is an important part of classical servo control systems [1], and it is widely used in the aerospace, military, transportation, and some other fields, such as aircraft servo systems [2], missile seeker servo platforms [3], and aircraft and vehicle braking systems [4]. As important techniques of electromechanical actuators, high performance servo motor and advanced stable controller are research hotspots in recent years. Permanent Magnet Synchronous Motor (PMSM) has many advantages like small torque and speed ripple, high torque-inertia ratio, and wide speed range. PMSM now is widely used in industries, especially in the aerospace [5].
Backlash is an important nonlinearity that limits the dynamic performance and steady precision of speed and position control in industrial, automation, and other applications. It exists in every mechanical system where a driving subsystem is not directly connected with the driven subsystem. Different from traditional hydraulic and pneumatic actuators, EMA has gear reducer which leads to the backlash nonlinearity. Backlash nonlinearity causes delays, noise, and oscillations which affect the system dynamic performance and steady precision. Due to the dynamic, nondifferentiable backlash nonlinearity and because it is difficult to be accurately measured, the compensation control is difficult to be designed.
Recently, studies of nonlinear systems with backlash have been the research hotspot. The noncontinuous transfer relationship caused by the backlash can be described from different perspectives. A number of mathematical models have been presented, such as hysteresis model [6, 7], dead-zone model [8], and impact-damper model [9]. The hysteresis model describes the relationship between the output angle of backlash and the input angle under the assumption that the shaft is stiff [10]. Dead-zone model describes the torque transitive relationship between the driving and driven subsystem [11]. Impact-damper model reflects the mechanism in the process of impaction caused by backlash.
Building a backlash inverse model at the control input and designing the feed forward compensation to offset the impact of backlash are most widely used in the current control compensation strategy. In practical engineering, PID control is the most commonly used algorithm due to its simple structure, but it is difficult to deal with nonlinear systems. So many intelligent methods have been studied including robust control [12], adaptive backstepping control [6], model predictive control [13], fuzzy control [14], and sliding-mode control [15]. This strategy is often applied in the system in which the backlash locates at the input or output. However, for the EMA, backlash cannot be simply converted to the control input. Inverse model compensation is no longer applicable. It is more reasonable to use the backlash dead-zone model for describing the force transfer relationship. Then the nonlinear system containing backlash can be regarded as Nonsmooth Sandwich System [16, 17].
Reference [8] uses the optimal control approach and adopts different control strategies at the backlash working time and normal condition, which can both make the backlash compensation and reduce the impact of backlash on the system. But the controller structure is too complex to be implemented. Reference [18] takes the differentiable function to approximate the dead-zone model and design a backstepping controller. The design requires an accurate system model while the system parameters are time-variant and are difficult to be accurately obtained, such as the stiffness coefficient and the width of the backlash. Reference [19] uses a fuzzy function approaching nonlinear function created by the backstepping strategy to simplify the design. The disadvantage is that they use an approximating function and its derivative term as the state variables. The states of the system do not have a clear physical significance and cannot be directly measured. And the control quality depends on the control parameter settings. So the backlash compensation effect cannot be guaranteed. Reference [20] designs an adaptive sliding-mode control (ASMC) based on extended state observer (ESO). ESO is employed to estimate the system states and an adaptive law is adopted to compensate backlash. The selected sliding surface and the controller can only guarantee the convergence but not terminal. And the chattering phenomenon has not been well solved.
Sliding-mode control (SMC) has self-adaptability to the system uncertainties and disturbance. But the singularity, chattering, and the convergence speed limit the application of SMC. In this paper, a chattering-free nonsingular fast terminal sliding-mode control (CNFTSMC) is presented based on backstepping method in order to compensate the backlash nonlinearity of the EMA and reduce the influence of the unknown external disturbance and parameter variation on the system. Compared to the existing methods, it solved the singularity problem by the design of the sliding surface and the control input. And a smooth control method via the low-pass filter is developed. This method can reduce the impact caused by the backlash effectively.
2. The Mathematical Model of EMA with Backlash
Several hypotheses are considered as follows before establishing the model:
Assume the stator core is not saturated.
Ignore the hysteresis losses and vortex losses.
The three phases are symmetric.
The air gap magnetic field is a sine wave.
The mathematical model of PMSM in dq axis is displayed as follows:(1)dθdt=ωdωdt=3pφf2Jiq-BJω-TLJdiqdt=-RLiq-pωid-pφfLω+uqLdiddt=-RLid+pωiq+udL,where θ and ω are the rotor angle and speed, respectively; φf is the permanent magnet flux; p is the pole pairs; J is the moment of inertia of the PMSM; B is the viscous friction coefficient; ud and uq are the voltages in dq axis; id and iq are the currents in dq axis; R is the phase winding resistance; L is the inductance; TL is the load torque.
The dynamic equation of the driven subsystem is shown by(2)dθsdt=ωsJsdωsdt=ηTL-Bsωs-Ts,where θs and ωs are the angle and speed of the load mechanism, respectively; Js and Bs are the moment of inertia and the friction coefficient of the load mechanism; η is the reduction ratio; Ts is the disturbance torque which includes the coupling torque and the unknown external disturbance.
There is a time-delay in torque transfer from the driving subsystem to the driven subsystem because of the backlash which exists in the reducer. The dead-zone model of the backlash is described by(3)TL=kTΔθ-αif Δθ>α0if Δθ≤αkTΔθ+αif Δθ<-α,where Δθ=θ-ηθs is the relative angular displacement; 2α is the width of the backlash; kT is the gear stiffness coefficient. In order to overcome the nonsmooth property of the model in (3), a differentiable function is adopted to approximate TL, given by (4)Tf=kTΔθ+kT2hlnehΔθ-α+e-hΔθ-αehΔθ+α+e-hΔθ+α,where h is a positive parameter (refer to as smooth degree). And the approximation error can be derived as (5)EΔθ=TL-Tf=-kTα-kTτΔθ2hif Δθ>α-kTΔθ-kTτΔθ2hif Δθ≤αkTα-kTτΔθ2hif Δθ<-α,where τΔθ=lneh(Δθ-α)+e-h(Δθ-α)/eh(Δθ+α)+e-h(Δθ+α).
Remark 1.
We can obtain from [18] that EΔθ<kTln2/2h and limh→+∞EΔθ=0, which means the nonsmooth property of backlash can be smoothed to any arbitrary precision by model (4). Thus, h in (4) is referred to as “smooth degree,” as illustrated in Figure 1.
Approximation of dead-zone with different smooth degrees: kT=160, α=0.02. (a) h=40, (b) h=100, and (c) h=500.
Then we can get from (1), (2), (3), (4), and (5):(6)dωsdt=Js-1-η2kTθs+ηkTθ-Bsωs+kT2hτΔθ+d1dωdt=J-1ηkTθs-kTθ-Bω+1.5pφfiq-kT2hτΔθ+d2,where d1=Js-1ηEΔθ-Ts and d2=-J-1EΔθ which are called the disturbance-like terms.
Define the system state variables as x=θs,ωs,θ,ω,iq,id; then the mathematical model of EMA with backlash is shown as(7)dθsdt=ωsdωsdt=Js-1-η2kTθs-Bsωs+kT2hτΔθ+Js-1ηkTθ+d1dθdt=ωdωdt=J-1ηkTθs-kTθ-Bω-kT2hτΔθ+1.5J-1pφfiq+d2y=θs,where y is the system output.
Two assumptions are listed as follows for the subsequent analysis.
Assumption 2.
The given signals θst, θ¨s∗t, and θ⃛s∗t are bounded and continuous.
Assumption 3.
The disturbances dii=1,2 and their derivatives are bounded which means that there are constants ldi and ρi satisfying the inequations: di≤ldi, d˙i≤ρi.
3. Design of the Control3.1. Design the Control for the Driven Subsystem
Define the error variables e1=θs-θs∗ and e2=e˙1=ωs-θ˙s∗. The sliding-mode surface can be written as(8)σ1=e1+λ1e2s1=σ1+β11sgnσ1σ1γ11+β12sgnσ˙1σ˙1γ12.And the virtual control θ∗ is shown:(9)θ∗=Jsη-1kT-1θeq∗+θsw∗θeq∗=-Js-1-η2kTθs-Bsωs+4ηkTατΔθ+θ¨s∗+λ1θ⃛s∗θsw∗+λ1θ˙sw∗=υ1υ1=-β12γ12-1sgnσ˙1σ˙12-γ121+β11γ11σ1γ11-1+k1s1+kd1+kl1+kn1sgns1.Select the Lyapunov function as follows:(10)V1=12s12.We can obtain(11)V˙1=s1s˙1=β12γ12σ˙1γ12-1s1-k1s1-kd1+kl1+kn1sgns1+d1+λ1d˙1≤β12γ12σ˙1γ12-1s1-k1s1-kn1sgns1=-K11s12-K12s1=-2K11V1-2K12V11/2≤0.And V˙1<0 for s1≠0, where K11=k1β12γ12σ˙1γ12-1 and K12=kn1β12γ12σ˙1γ12-1.
3.2. Design the Control for the Driving Subsystem
Define the error variables e3=θ-θ∗ and e4=e˙3=ω-θ˙∗; the sliding-mode surface and the controller are obtained as(12)σ2=e3+λ2e4(13)s2=σ2+β21sgnσ2σ2γ21+β22sgnσ˙2σ˙2γ22(14)iq∗=23Jp-1φf-1iqeq∗+iqsw∗iqeq∗=-J-1ηkTθs-kTθ-Bω-4ηkTατΔθ+θ¨∗+λ2θ⃛∗iqsw∗+λ2i˙qsw∗=υ2υ2=-β22γ22-1sgnσ˙2σ˙22-γ221+β21γ21σ2γ21-1+k2s2+kd2+kl2+kn2sgns2.
In order to realize the decoupling of current and speed and simplify the controller, the control strategy id∗=0 has been used. Now the control block diagram of the system is displayed in Figure 2.
Control block diagram of the system.
3.3. Stability AnalysisLemma 4.
The equilibrium point x=0 is globally finite-time stable for any given initial condition x0=x0 if a Lyapunov description can be obtained as [21](15)V˙+m1V+m2Vκ≤0.And then the setting time can be given by(16)t≤t0+1m11-κlnm1V1-κt0+m2m2,where m1,m2>0, 0<κ<1.
Lemma 5.
Assume a1,a2,…,an and δ∈0,2 are all positive numbers; then the following inequality holds [22]:(17)a1δ+a2δ+⋯+anδ≥a12+a22+⋯+an2δ/2.
Theorem 6.
For system (7), if the controllers are designed as (9) and (14), then the system will converge in finite-time. And the settling time is given by (21).
Proof.
Consider the following Lyapunov function:(18)VM=V1+12s22.
Then the derivative of VM is(19)V˙M=V˙1+V˙2=s1s˙1+s2s˙2=β12γ12σ˙1γ12-1s1-k1s1-kd1+kl1+kn1sgns1+d1+λ1d˙1+β22γ22σ˙2γ22-1s2-k2s2-kd2+kl2+kn2sgns2+d2+λ2d˙2≤β12γ12σ˙1γ12-1s1-k1s1-kn1sgns1+β22γ22σ˙2γ22-1s2-k2s2-kn2sgns2=-K11s12-K12s1-K21s22-K22s2≤-∑i=12Ki1si2-∑i=12Ki2si,where K21=k2β22γ22σ˙2γ22-1, K22=kn2β22γ22σ˙2γ22-1, K-1=minKi1, and K-2=minKi2.
From Lemma 5, (19) can be rewritten as(20)V˙M≤-K-1∑i=12si2-K-2∑i=12si≤-2K-1VM-2K-2VM1/2.According to Theorem 6 and Lemma 4, the system will converge in finite-time:(21)tM≤tM0+1K-1ln2K-1VMtM01/2+2K-22K-2.
Remark 7.
In Theorem 6, because of the switching function sgn·, υ is nonsmooth. Let υ pass a low-pass filter, and usw is the output of the filter. So the switching control term usw is smooth. If the switching control term usw is designed as u˙sw=υ, it is also a smooth function. But it is more difficult to be implemented in engineering applications.
4. Experimental Studies
In order to demonstrate the effectiveness of the proposed method, the platform shown in Figure 3 has been built which is controlled by TMS320F28335 of TI.
The experimental platform.
The platform is mainly composed of power supply, controller, drive motor, reducer, and load simulator. The parameters of the system are displayed in Table 1.
Parameters of the system.
Para
Description
Value
Unit
p
Number of pole pairs
5
—
φf
Permanent magnet flux linkage
0.143
Wb
J
Moment of inertia of PMSM
2.8×10-4
kg⋅m2
B
Viscous friction coefficient
1.0×10-4
N⋅m/rad/s
R
Stator resistance
1.73
Ω
L
Inductance
7
mH
TL
Rated torque
4.7
N⋅m
Js
Moment of inertia of flap
0.4
kg⋅m2
Bs
Friction coefficient of flaps
0.12
N⋅m/rad/s
η
Gear reduction ratio
30
—
Vdc
DC link voltage
270
V
kT
Stiffness coefficient
160
N⋅m/rad/s
2α
Width of the backlash
0.04
rad
The parameters of the sliding-mode surface λ1=λ2=0.1, β11=β21=10, β12=β22=0.4, γ11=γ21=2, γ12=γ22=1.5, and the parameters of the control k1=120, k2=400, kd1+kl1+kn1=20, kd2+kl2+kn2=35, and the smooth degree h=200. The experiments are under PID control and the proposed method, respectively.
4.1. Set the Reference Signal θs∗t=1rad
First, the step response of the closed-loop system has been considered and the performance is shown as Figures 4–7.
The phase trajectory under PID-controller.
The phase trajectory under SMC-controller.
The tracking trajectory.
The tracking error.
4.2. Set the Reference Signal θs∗t=sin4πtrad
The sinusoidal response is presented in Figures 8–11.
The tracking trajectory under PID-controller.
The tracking error under PID-controller.
The tracking trajectory under SMC-controller.
The tracking error under SMC-controller.
Figures 4 and 5 show the phase plane of the system. We can obtain that the proposed control can eliminate the limit cycle phenomenon. Figure 6 shows that the response under CNFTSMC has a smaller overshoot than that under PID-controller. And we can see from Figure 7 that the proposed controller has a higher tracking accuracy. Figures 8–11 illustrate that the proposed controller can reduce the phase delay and the tracking error.
5. Conclusion
A chattering-free nonsingular terminal sliding-mode control based on backstepping has been proposed to achieve the rudder angle tracking for EMA with considering the backlash nonlinearity. Our contributions are as follows:
Some deficiencies of the backlash compensation control in the current studies have been overcome. The proposed method can reduce the impact caused by the backlash. The experimental results demonstrate the effectiveness of the proposed method.
The singularity and chattering of conventional terminal sliding-mode control are effectively solved. The parameters can be set flexibly according to the actual demand.
The system can converge in finite-time under the chattering-free nonsingular fast terminal sliding-mode control.
Competing Interests
The authors declare that there is no conflict of interests regarding the publication of this article.
Acknowledgments
This research was supported by the National Natural Science Fund of China (Grant 51407143).
GuoH.XingW.Development of electromechanical actuators20072836206272-s2.0-34347248720QiH.FuY.QiX.LangY.Architecture optimization of more electric aircraft actuation system201124450651310.1016/S1000-9361(11)60058-72-s2.0-80052193245LeeS.-Y.LeeJ.-S.A control for 2-axis gimbaled millimeter wave seeker using space vector PWM of PMSM201160122386239110.5370/KIEE.2011.60.12.23862-s2.0-83755187937ZhangQ. H.LiY. R.Design and analysis on aircraft brake system200333397100WenZ.WangZ.QiX.BaoL.Artificial-neural-network adaptive control of compound actuator20073311129512982-s2.0-37749042254ZhouJ.ZhangC. J.WenC. Y.Robust adaptive output control of uncertain nonlinear plants with unknown backlash nonlinearity200752350350910.1109/tac.2006.890473MR23004812-s2.0-34548512302MerzoukiR.DavilaJ. A.FridmanL.CadiouJ. C.Backlash phenomenon observation and identification in electromechanical system200715444745710.1016/j.conengprac.2006.09.0012-s2.0-33846310812TaoG.MaX.LingY.Optimal and nonlinear decoupling control of systems with sandwiched backlash200137216517610.1016/s0005-1098(00)00153-9MR18320222-s2.0-0035157197Mata-JimenezM. T.BrogliatoB.GoswamiA.On the control of mechanical systems with dynamic backlashProceedings of the 36th Conference on Decision and Control1997Washington, DC, USAIEEE19901995NordinM.GutmanP.-O.Controlling mechanical systems with backlash—a survey200238101633164910.1016/s0005-1098(02)00047-xMR21335232-s2.0-0036779201ZuoZ.LiX.ShiZ.L 1 adaptive control of uncertain gear transmission servo systems with deadzone nonlinearity201558677510.1016/j.isatra.2015.07.0092-s2.0-84943584059TarbouriechS.QueinnecI.PrieurC.Stability analysis and stabilization of systems with input backlash201459248849410.1109/tac.2013.2273279MR31648942-s2.0-84893601081PhilippP.ThomasB.MiroslavB.Hybrid approach to modeling, control and state estimation of mechanical systems with backlash2007801117291740LiuY.-J.TongS.Adaptive fuzzy control for a class of nonlinear discrete-time systems with backlash20142251359136510.1109/TFUZZ.2013.22868372-s2.0-84902164335CorradiniM. L.OrlandoG.Robust stabilization of nonlinear uncertain plants with backlash or dead zone in the actuator200210115816610.1109/87.9743492-s2.0-0036210799DongR.TanY.On-line identification algorithm and convergence analysis for sandwich systems with backlash20119358859410.1007/s12555-011-0320-22-s2.0-80052687194TanY.DengL.Modeling the dynamic sandwich system with hysteresis using NARMAX model20149716218810.1016/j.matcom.2013.04.029MR31379142-s2.0-84885391795ShiZ. G.ZuoZ. Y.Backstepping control for gear transmission servo systems with backlash nonlinearity201512275275710.1109/tase.2014.23694302-s2.0-84927636031DuR.-H.WuY.-F.ChenW.ChenQ.-W.Adaptive backstepping fuzzy control for servo systems with backlash201330225426010.7641/cta.2013.205222-s2.0-84876037752PengbingZ.YaoyaoS.Adaptive sliding mode control of the A-axis used for blisk manufacturing201427370871510.1016/j.cja.2014.04.0132-s2.0-84903575963NekoukarV.ErfanianA.Adaptive fuzzy terminal sliding mode control for a class of MIMO uncertain nonlinear systems2011179344910.1016/j.fss.2011.05.009MR28181962-s2.0-79959971947YuS.YuX.ShirinzadehB.ManZ.Continuous finite-time control for robotic manipulators with terminal sliding mode200541111957196410.1016/j.automatica.2005.07.001MR21686602-s2.0-24944498043