Deviation-free and model-free filtering device and method for servo driving system

Abstract

According to the non-deviation and non-model filtering device and method for the servo driving system, all filter gains are determined only by three scalar parameters (k1, k2 and k3), complex matrix operation or online gain adjustment is not needed, and rapid and efficient field debugging is achieved. The method can be efficiently realized in an embedded DSP environment, the hardware cost is remarkably reduced, and the control performance of the system is averagely improved by about 27% through experimental verification.

Description

Technical Field

[0001] This invention belongs to the field of servo system control technology, and in particular relates to a filter device and method for a deviation-free and model-free servo drive system. Background Technology

[0002] In existing servo systems, three common methods are used to estimate states such as position, velocity, and acceleration: low-pass filter (LPF), Luenberger observer, and extended state observer (ESO). However, these methods all have the following technical limitations:

[0003] 1) Low-pass filter (LPF)

[0004] LPF is one of the most commonly used filtering methods in industrial servo drive systems. It filters out high-frequency noise in encoder measurement signals by setting a cutoff frequency. Its first-order structure is simple, computationally inexpensive, and easy to implement in embedded systems.

[0005] However, the filtering performance of LPF is extremely sensitive to bandwidth selection: if the bandwidth is set too narrow, it will cause signal response lag; if the bandwidth is too wide, it will not be able to effectively suppress high-frequency interference.

[0006] Furthermore, due to its inherent phase delay and amplitude attenuation characteristics, the signal is prone to distortion under high-speed operation or rapid load changes, which affects the feedback accuracy and stability of the system.

[0007] 2) Luenberger Observer

[0008] The Luenberger observer is based on a state-space model to design an error feedback matrix, which is used to estimate the system state.

[0009] Compared to the LPF, the Luenberger observer performs better in high-frequency response and has smaller phase error. However, it is highly dependent on the accurate mathematical model of the system, especially physical parameters such as the motor's torque constant, moment of inertia, and coefficient of friction.

[0010] In practical engineering, these parameters are often difficult to obtain accurately and can change under different working conditions, thus limiting the Luenberger observer's application in the field.

[0011] To ensure the desired response speed and system stability, it is necessary to repeatedly design characteristic equations and perform matrix operations, which makes the debugging process complex and the implementation difficult.

[0012] 3) Extended State Observer (ESO)

[0013] ESO is an advanced observer structure proposed to enhance the robustness of the system. It considers model errors and external disturbances as "extended states" and dynamically estimates and compensates them, achieving error suppression and state reconstruction.

[0014] However, ESO still relies on the basic model structure of the system, especially the physical models of inertia, friction, load torque, etc.

[0015] When the load changes frequently or contains high-frequency components, ESO is sensitive to the rate of disturbance change, which may lead to increased estimation error, even oscillation, delay or system instability.

[0016] To solve the above problems, an online gain scheduler or system identification module often needs to be designed, which will further increase the complexity of system implementation.

[0017] Although the above methods can achieve a certain degree of state estimation under certain conditions, there are three key technical problems in actual engineering applications:

[0018] 1. Poor performance consistency

[0019] Existing filters are very sensitive to system parameter changes and load disturbances, especially under high-speed operation, large parameter uncertainty or sudden external disturbances. It is easy to cause estimation error to increase sharply or the controller to be unstable, resulting in large fluctuations in closed-loop control performance, and the reliability and predictability of system operation decrease.

[0020] 2. Strong dependence on parameter identification and database

[0021] In particular, Luenberger and ESO type filters are highly dependent on accurate physical parameters of the motor, such as inductance, resistance, moment of inertia, friction coefficient, etc. To achieve the desired performance, a gain database is usually constructed, and a real-time parameter estimation algorithm is designed, which significantly increases the implementation complexity, computational resource consumption and development cost of the system.

[0022] 3. Complex debugging process, difficult engineering deployment

[0023] The filter system needs multiple rounds of iterative experiments, matrix eigenvalue configuration, and even manual construction of characteristic polynomials in actual deployment. Engineers need to have strong control theory and numerical analysis capabilities, which is not conducive to quick online or non-professional operation, and the engineering implementation efficiency is low.

[0024] In summary, existing servo filtering technology cannot meet the requirements of model independence, robustness and debugging convenience at the same time. SUMMARY

[0025] In view of the problem that the existing servo filtering technology is difficult to meet the requirements of model independence, robustness and debugging convenience, the present application provides a filter device and method for servo driving system without deviation and model, all filter gains are determined by only three scalar parameters (k1, k2, k3), without complex matrix operation or online gain adjustment, realizing fast and efficient on-site debugging. It can be efficiently realized in embedded DSP environment, significantly reducing hardware cost, and through experimental verification, the average control performance of the system is improved by about 27%.

[0026] The technical solutions of the present application are as follows:

[0027] The filter device for servo driving system without deviation and model comprises:

[0028] A signal acquisition module for acquiring rotor position θ of a test motor m and stator current i a .

[0029] A filter module for receiving rotor position θ sent by the signal acquisition module m , and obtaining speed estimation value ω and acceleration estimation value α through three layers of filtering of position filtering, speed filtering and acceleration filtering

[0030] A speed and current controller for receiving stator current i, speed estimation value ω and acceleration estimation value α and calculating output voltage command v according to them a . a

[0031] A pulse width modulator for generating driving signal of three-phase inverter according to voltage command v a .

[0032] And the three-phase inverter for driving load motor according to the driving signal of the three-phase inverter.

[0033] Preferably, in the filter device for servo driving system without deviation and model,

[0034] The filter module comprises a position filter, a speed filter and an acceleration filter;

[0035] The position filter takes rotor position θ as input m , outputs position estimation value θ through two series connected integrators and feedback paths

[0036]

[0037] Gain:​​​​

[0038] Where, x θ,1 =θ m , k1 and γ1 are design parameters used to determine the gain value;

[0039] The speed filter is As input, the output velocity estimate is obtained by passing through a dual-integral filter of the same structure.

[0040]

[0041] Gain:k P，j :=k j +γ j k I，j =k j γ j (14)

[0042]

[0043] in, k2 and γ2 are design parameters used to determine the gain value;

[0044] The acceleration filter uses velocity estimates. Input: Estimated acceleration value

[0045]

[0046] in, γ and γ3 are design parameters used to determine the gain value.

[0047] A filtering method for unbiased, model-free servo drive systems includes the following steps:

[0048] Collect the rotor position θ of the test motor m With stator current i a ;

[0049] According to the rotor position θ m The velocity estimate is obtained through three layers of filtering: position filtering, velocity filtering, and acceleration filtering. and acceleration estimator

[0050] According to the stator current i a The speed estimate and acceleration estimator Calculate the output voltage command v a ;

[0051] According to the voltage command v a Generate drive signals to drive the load motor.

[0052] The present invention has the following beneficial effects:

[0053] 1. No modeling required, applicable to various types of motors.

[0054] This invention does not rely on any electrical parameters, mechanical models, or external disturbance models, and can be widely applied to various servo control platforms such as permanent magnet synchronous motors (PMSM), brushless DC motors (BLDC), and induction motors (IM).

[0055] 2. Excellent dynamic performance and filtering accuracy

[0056] This invention effectively suppresses estimation errors caused by phase delay and response lag through pole-zero cancellation and dynamic error diagonalization, and can maintain stable and fast convergence characteristics even under high-speed operation or sudden load conditions.

[0057] 3. Simple structure and easy debugging

[0058] All filtering parameters consist of only **three scalar gains (k1, k2, k3)**, eliminating the need for complex matrix calculations or eigenvalue design, facilitating rapid on-site application and adjustment. Furthermore, this architecture can operate efficiently in embedded systems, minimizing computational load and program overhead.

[0059] 4. Experimentally proven performance improvement

[0060] In actual servo system experiments, the proposed filter improved the RMSE index of position and velocity estimation by an average of 27.4% compared with the traditional LPF and ESO structures, while significantly suppressing high-frequency oscillations in the acceleration estimation process.

[0061] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0062] Figure 1 A structural block diagram of the filter module in the filter device for the unbiased, model-free servo drive system provided by the present invention;

[0063] Figure 2 A flowchart illustrating the filtering method for the unbiased, model-free servo drive system provided by this invention.

[0064] Figure 3 This is a schematic diagram of the experimental environment for a servo system based on a dual-motor test platform.

[0065] Figure 4 To ensure that the speed reference value remains constant, for a frequency of... The time-varying load, the measured control speed and stator current;

[0066] Figure 5 To achieve a constant speed steady-state mode, for a frequency of... The time-varying load, the filtering error of position, velocity and acceleration. Detailed Implementation

[0067] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0068] This invention provides a bias-free, model-free filtering device for servo drive systems, comprising:

[0069] The signal acquisition module is used to acquire the rotor position θ of the motor. m With stator current i a ;

[0070] The filter module is used to receive the rotor position θ sent by the signal acquisition module. m The velocity estimate is obtained through three layers of filtering: position filtering, velocity filtering, and acceleration filtering. and acceleration estimator

[0071] Speed ​​and current controller, which is used to receive the stator current i a The speed estimate and acceleration estimator And calculate the output voltage command v based on it. a ;

[0072] A pulse width modulator, which is used to adjust the voltage command v according to the voltage command v a Generate the drive signal for the three-phase inverter;

[0073] and the three-phase inverter, which is used to drive the motor according to the drive signal of the three-phase inverter.

[0074] like Figure 1 As shown, the filter module includes a position filter, a velocity filter, and an acceleration filter;

[0075] The position filter is based on the rotor position θ. m As input, through two cascaded integrators and a feedback path, a stable and unbiased position estimate is output.

[0076] This part only requires adjustment of one proportional gain parameter k1, and its structure is simple.

[0077]

[0078] Gain:

[0079] x θ,1 =θ m ,x θ,2 =ω m ,x θ,3 =a m

[0080] As above, by substituting formula (10) into (9) and integrating both sides, the calculation can be performed.

[0081] k1 and γ1 are design parameters in positive constant form, used to determine the gain values ​​mentioned above.

[0082] The speed filter is As input, the output velocity estimate is obtained by passing through a dual-integral filter of the same structure. Only one gain parameter, k2, needs to be adjusted;

[0083]

[0084] Gain:k P，j :=k j +γ j k I，j =k j γ j (14)

[0085] This value can be calculated using formula (9), substituted into formula (12), and integrated over both sides of (12). Then, it can be obtained using formula (13). calculate Calculation. The formula is shown in the lower left image. The subsequent calculation can be performed in two steps.

[0086] Numerical solution of differential equations: (Achieved through formulas (12) and (13))

[0087] calculate: (Achieved through formula (13))

[0088] k2 and γ2 are design parameters in positive constant form, used to determine the gain values ​​mentioned above.

[0089] The acceleration filter uses velocity estimates. Input: Estimated acceleration value The structure is the same as the previous two stages, and only one proportional gain k3 needs to be set;

[0090] This value can be calculated using formulas (12)-(14), substituted into formula (12), and integrated on both sides of the equation. Then, it can be obtained using formula (13). calculate Calculation. The subsequent calculation can be performed in two steps.

[0091] Numerical solution of differential equations: (Based on formulas (12)(13))

[0092] calculate (Based on formula (13))

[0093] k3 and γ3 are design parameters in positive constant form, used to determine the gain values ​​mentioned above.

[0094] like Figure 2 As shown, the present invention also provides a bias-free, model-free filtering method for servo drive systems, which includes the following steps:

[0095] System Initialization: This refers to the system startup and parameter setting phase.

[0096] Measuring (measurement module): Acquires rotor position θ m With stator current i a .

[0097] Proposed Filter: Input θ m Output estimate This module consists of three layers of filters, which respectively realize position, velocity and acceleration estimation.

[0098] Speed ​​& Current Controller: Uses estimated results for closed-loop control, outputting a voltage command (V). a .

[0099] PWM (Pulse Width Modulator): Generates the drive signal for a three-phase inverter.

[0100] Three-Phase Inverter: Drives the motor according to the PWM signal.

[0101] To verify the feasibility and practical performance of the offset-free, model-independent state estimation filter structure proposed in this invention, a servo system experimental environment based on a dual-motor test platform was constructed, such as... Figure 3As shown. The system uses a 500W brushless DC motor (BLDCM) as the test motor and a 700W permanent magnet synchronous motor (PMSM) as the load motor. The real-time deployment of the controller and filter is achieved using a high-speed digital signal processor (TI LAUNCHXL-F28379D) and a three-phase inverter (TIDRV8305, DC bus voltage of 24V).

[0102] The entire system is programmed in C language, with a service cycle of 0.1ms. It communicates with the host computer via a controller area network (CAN) to achieve experimental data acquisition and graphical visualization. Two closed-loop controllers are used as a comparison basis in the experiment:

[0103] Controller A: Speed-based and acceleration Feedback, independent of current loop.

[0104] Controller B: Speed-dependent With current loop feedback.

[0105] The controller bandwidth is set to the speed loop ω sc =2π5rad / s, current loop ω cc = 2π100rad / s.

[0106] The filter described in this invention comprises three cascaded integral structures, which are used to estimate position, velocity and acceleration, respectively. The filter parameters only require three scalars k1, k2 and k3, making it simple to implement and easy to adjust online and embed.

[0107] Compared to traditional low-pass filters (LPFs) and extended state observers (ESOs), the filter of this invention performs better in the following aspects:

[0108] It has strong noise suppression capabilities and provides a smoother estimated response.

[0109] It has a faster response speed, less jitter, and good dynamic consistency;

[0110] Even under high-frequency load disturbance conditions, it can still maintain stable output.

[0111] The parameters of the test motor used in the experiment were set as nominal values ​​as follows, and then applied to the actual parameters using correction factors:

[0112] J m,0 =1.45J m B m，0 =1.25B m k T,0 =1.2k T (=k e,0 ), L a,0 =1.3La ,and R a,0 =1.4R a .

[0113] Figure 4 To ensure that the speed reference value remains constant, for a frequency of... The time-varying load, the measured control speed and stator current.

[0114] With a speed reference value set at 1000 rpm, the closed-loop response results of the system under sinusoidal loads at frequencies f_{T_L,ref} = 6, 12, and 18 Hz show that the proposed filtering method can significantly reduce the oscillations and fluctuations in the speed channel. During sudden load changes, traditional filtering methods exhibit significant speed overshoot and persistent oscillations, while the proposed method maintains smooth and rapid speed convergence characteristics. This is mainly due to the systematic introduction of a low-pass filter in the filtering stage, the dynamic diagonalization effect of error resulting from order reduction, and the model-independent perturbation suppression mechanism. Furthermore, Figure 4 The waveform results of the stator current are also presented. As can be seen from the figure, the current ripple amplitude is significantly reduced after adopting the method proposed in this invention, which greatly reduces the stress on the three-phase inverter and DC bus capacitors compared with the traditional method. This improvement not only enhances the stability of closed-loop control but also effectively extends the service life of hardware components, demonstrating the practical advantages of the proposed solution in industrial scenarios.

[0115] Figure 5 To achieve a constant speed steady-state mode, for a frequency of... The time-varying load, the filtering error of position, velocity and acceleration.

[0116] Under the same time-varying load, the method proposed in this invention exhibits significant differences from traditional filtering methods in the velocity and acceleration estimation stages. Traditional filtering methods generate substantial ripple and oscillations in the velocity and acceleration estimation signals at higher load frequencies, leading to a deterioration in closed-loop performance. In contrast, the method proposed in this invention effectively suppresses this instability, ensuring that the filtered output remains stable across the entire frequency band.

[0117] Specifically, the velocity estimation results show that the method of this invention can significantly reduce the high-frequency vibration level during the transition process, avoiding the resulting velocity feedback jitter. The acceleration estimation results further illustrate that while traditional methods rapidly amplify errors as the disturbance frequency increases, the proposed method maintains a lower error amplitude and better smoothness. In summary, Figure 4 The results demonstrate that the filter structure proposed in this invention can provide consistent dynamic performance and strong robustness under various load disturbance frequencies.

[0118] Table 1 is a comparative analysis of the frequency response of the present invention with that of the traditional low-pass filter (LPF) and extended state observer (ESO) based on Bode plots.

[0119] Table 1

[0120]

[0121] ※Sce.#1:k T.0 =1.2k T →1.5k T →1.2k T

[0122] Sce#2:k T.0 =1.2k T →1.8k T →1.2k T

[0123] Sce.#3:k T.0 =1.2k T →2.1k T →12k T

[0124] As shown in the table above, the filter proposed in this invention effectively overcomes the phase delay and gain instability problems in traditional methods, and can maintain stable response performance over a wide frequency band. Meanwhile, a quantitative comparative analysis of the root mean square error (RMSE) of position and velocity estimation was conducted.

[0125] Compared with traditional low-pass filters (LPF) and extended state observers (ESO), the proposed filter reduces the average RMSE of position and velocity estimation by 27.4% and exhibits the smallest amplitude oscillation in acceleration estimation, demonstrating excellent dynamic performance.

[0126] This invention has the following characteristics:

[0127] A. Achieving model independence

[0128] The structural design, which combines a dual integrator with a disturbance observer, enables state estimation without relying on a motor parameter model, and is highly versatile.

[0129] B. Enhance performance consistency

[0130] By utilizing pole-zero cancellation and dynamic error diagonalization techniques, stable and consistent filtering performance can be maintained even under conditions such as high-frequency vibration or sudden load changes.

[0131] C. Simplify the debugging process

[0132] All filter gains are determined by only three scalar parameters (k1, k2, k3), eliminating the need for complex matrix operations or online gain adjustments, thus enabling fast and efficient on-site commissioning.

[0133] D. Improve industrial site applicability

[0134] It can be efficiently implemented in an embedded DSP environment, significantly reducing hardware costs, and experimental verification shows that the system control performance is improved by an average of about 27%.

[0135] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details.

Claims

1. Filter device for bias-free model-free servo drive systems, characterized in that comprising: a signal acquisition module for acquiring rotor position θ of the test motor m and stator current i a ; a filter module for receiving the rotor position θ from the signal acquisition module m and obtaining a speed estimate by position filtering, speed filtering and acceleration filtering and an acceleration estimate sub-module a speed and current controller for receiving the stator current i a , the speed estimate and acceleration estimate and calculating therefrom an output voltage command v a ; a pulse width modulator for generating driving signals for a three-phase inverter in dependence of the voltage command v a generating driving signals for a three-phase inverter; and the three-phase inverter for driving a load motor according to a drive signal of the three-phase inverter.

2. The filter device for a bias-free, model-free servo drive system according to claim 1, characterized in that The filter module comprises a position filter, a speed filter and an acceleration filter; The position filter has as input the rotor position θ m and outputs a position estimate value Gain: k P，1 : = 2k1+ γ1, where x θ,1 = θ m , k1 and γ1 are design parameters for determining the gain value; The velocity filter outputs a velocity estimate value from the same structure of double integration filter Gain: k P，j : = k j + γ j , k I，j = k j γ j , (14) wherein, k2 and γ2 are design parameters used to determine the gain value; the acceleration filter takes as input the velocity estimate and outputs an acceleration estimate wherein, and γ3 is a design parameter used to determine the gain value.

3. Filter method for bias-free model-free servo drive systems, characterized in that comprising the steps of: Collecting the rotor position θ of the test motor m With the stator current i a ; According to the rotor position θ m And through the three layers of filtering of position filtering, velocity filtering and acceleration filtering, the velocity estimation value is obtained And acceleration estimation sub According to the stator current i a , the speed estimate and the acceleration estimate the output voltage command v a is calculated; According to the voltage command v a A drive signal is generated to drive a load motor.

Images