A contactless permanent magnet vernier motor mechanical parameter identification method and system

Abstract

The application discloses a kind of contactless permanent magnet vernier motor mechanical parameter identification method, system, belongs to servo motor control field.The application proposes a kind of parallel cascade extension sliding mode identification strategy, only needs to rely on the electromagnetic torque and mechanical speed of vernier motor end inner rotor as original data, by establishing integrated equivalent dynamic model considering the influence of magnetic gear drive, introduce simplified LuGre friction model to establish coulomb friction dynamic process, the multiple mechanical parameters to be identified are grouped and parallel cascade extension sliding mode observer is designed, and high, low acceleration differential excitation signal is alternately identified each group parameter, gradually eliminate the coupling between parameters;Solve the problem that traditional motor parameter identification method ignores friction dynamic process, multiple parameters are strongly coupled and difficult to decouple, and the identification precision is low.

Description

Technical Field

[0001] This invention belongs to the field of permanent magnet synchronous motor control technology, specifically relating to a method for identifying multiple mechanical parameters in a contactless permanent magnet vernier motor driven servo system. Background Technology

[0002] Permanent magnet synchronous motors are widely used in servo drives due to their high power density and excellent dynamic performance. However, the traditional permanent magnet motor + mechanical reducer structure has problems such as slow response, high noise, difficult maintenance, and reliance on imported precision reducers. The contactless integrated permanent magnet vernier motor deeply integrates magnetic gears and vernier motors. While eliminating the mechanical reducer, it also has the advantages of high torque density, high dynamic response, and low vibration and noise. It is particularly suitable for robot joint servo applications with compact space, low speed and high torque, and high positioning accuracy requirements.

[0003] The dynamic control performance of servo systems, such as PI parameter self-tuning, model predictive control, and feedforward control, heavily relies on the accuracy of the system's mechanical parameters. These parameters mainly include equivalent moment of inertia, viscous friction coefficient, load torque, and LuGre model parameters that can accurately describe complex frictional behavior. However, these parameters are difficult to measure directly in actual operation and are strongly coupled with each other. Especially when considering the dynamic process of Coulomb friction, parameter identification faces significant challenges.

[0004] Currently, methods for identifying mechanical parameters of motor systems are mainly classified into four categories: observer-based methods, recursive methods, adaptive methods, and intelligent methods. However, existing methods either neglect Coulomb friction or simplify the process using only sign functions, ignoring its dynamic process near the zero-crossing velocity point, thus limiting the identification accuracy. For example, some studies identify parameters by constructing full-rank equations, but assume that Coulomb friction can be ignored. Other studies, while considering friction models, only identify the friction parameters themselves without decoupling them from parameters such as inertia and load. Recursive methods, such as Kalman filtering and recursive least squares, can be used for online estimation, but have high computational complexity and are sensitive to initial values. Adaptive methods struggle to guarantee the convergence of all parameters when dealing with multi-parameter coupling. Intelligent methods, such as particle swarm optimization, offer high accuracy but require numerous iterations, resulting in high computational and storage overhead, making them difficult to deploy in general-purpose drive systems. Observer-based methods, especially sliding mode observers, have attracted widespread attention due to their robustness and computational simplicity.

[0005] Therefore, there is an urgent need for a mechanical parameter identification method that can simultaneously consider the dynamic process of Coulomb friction, effectively decouple multiple parameters, and has low computational complexity and is easy to implement in engineering. Summary of the Invention

[0006] In view of the shortcomings of the prior art, the present invention provides a parallel cascade identification method for mechanical parameters of a contactless permanent magnet vernier motor driven servo system. It aims to solve the problems of neglecting friction dynamics, difficulty in decoupling parameter coupling, and high computational complexity in the prior art. It is particularly suitable for online or offline identification of rotational inertia, viscous friction coefficient, load torque, and LuGre friction model parameters during the system's self-tuning process.

[0007] To solve the above technical problems, the present invention is achieved through the following technical solution:

[0008] First, this invention proposes a parallel cascade identification method for mechanical parameters of a permanent magnet synchronous motor driven servo system, comprising the following steps:

[0009] Step 1: Establish a dynamic model of the permanent magnet synchronous motor driven servo system, and introduce a simplified LuGre friction model considering the dynamic process of Coulomb friction; the dynamic model equates the servo system to a single-mass rigid system, and its equation of motion is:

[0010] ,

[0011] ,

[0012] Among them, J eq For the equivalent moment of inertia, ω m This refers to the motor speed. Let T be the angular acceleration. e For electromagnetic torque, T f For the total frictional torque, T l This is the load torque;

[0013] ,

[0014] Where T f σ0 represents the total frictional torque; σ0 is the bristle stiffness coefficient; z is the average bristle deformation, which is an internal state variable of the model; σ1 is the bristle damping coefficient; σ2 is the rate of change of deformation; B eq ω is the coefficient of viscous friction. m ω represents the relative angular velocity between the contact surfaces, i.e., the mechanical speed of the outer rotor of the motor; g(ω) is the Stribeck effect function, where T c Let T be the Coulomb friction torque. st For static friction torque, ω s For Stribeck speed.

[0015] To reduce the complexity of parameter identification, this invention will use ω s Take a sufficiently small value so that g(ω) degenerates to g≈Tc; at the same time, ignore the effect of σ1 and let T clb=σ0z, thus simplifying the original model's six parameters into three equivalent parameters, facilitating engineering applications. The simplified LuGre friction model is expressed as:

[0016] .

[0017] Step 2: Divide the mechanical parameters to be identified into two groups: The first group of parameters includes the equivalent moment of inertia J. eq Equivalent viscous friction coefficient B eq and load torque T l The second set of parameters includes the stiffness coefficient σ0 and the Coulomb friction torque T from the simplified LuGre model. c ;make =1 / J eq τ=σ0 / T c Then the parameter to be identified becomes B eq T l T clb , σ0 and τ.

[0018] Step 3: Design a parallel cascaded extended sliding mode observer, which includes a first parallel observer structure and a second parallel observer structure; the first parallel observer structure includes three parallel extended sliding mode observers Ω1, Ω2 and Ω3, which are used for synchronous identification. B eq and T l The second parallel observer structure includes a component for estimating the Coulomb friction dynamic component T. clb An extended sliding mode observer Ω4, and two parallel extended sliding mode observers Ω5 and Ω6, are used to simultaneously identify σ0 and τ, respectively.

[0019] Furthermore, the specific design of the first parallel observer structure in step 3 is as follows:

[0020] Let T clb 'T' is the value of T calculated from the second set of parameter estimates using a simplified LuGre model. clb The values, and the expressions for each observer, are as follows:

[0021] The expression for observer Ω1 is:

[0022] ,

[0023] The expression for observer Ω2 is:

[0024] ,

[0025] The expression for observer Ω3 is:

[0026] .

[0027] Furthermore, the specific design of the second parallel observer structure in step 3 is as follows:

[0028] ,

[0029] Where, ω m4 This refers to the motor speed. , , These are the estimated values ​​of the reciprocal of the equivalent moment of inertia, the coefficient of viscous friction, and the load torque, respectively, identified by the first parallel observer structure. For T clb The estimated value, p4 is the switching gain, S4= -ω m4 Let f4 be the sliding surface, f4 be the feedback gain coefficient, and G4 = - , and Let σ0 and τ be the estimated values, respectively, and sgn(·) be the sign function.

[0030] The observer Ω5 is constructed based on a simplified LuGre model and is used to identify σ0. Its expression is:

[0031] ,

[0032] in, For Ω5 pairs T clb The estimate, Here is the estimated value for Ω4, p5 is the switching gain, and S... 5= -T clb f5 is the sliding surface and f5 is the feedback gain coefficient.

[0033] The observer Ω6 is constructed based on a simplified LuGre model and is used to identify τ. Its expression is:

[0034] ,

[0035] in, For Ω6 pairs T clb The estimated p6 switch gain, S6= -T clb f6 is the sliding surface, and f6 is the feedback gain coefficient.

[0036] Furthermore, the gain coefficient of the observer must satisfy the following stability and convergence conditions:

[0037] Stability is analyzed using the observer Ω1 as an example, and the sliding surface S1 is defined as... -ω m1 Choose the Lyapunov function V1=0.5S12 Its derivative is:

[0038] ,

[0039] in, for The estimation error, the residual term R1 of the observer Ω1 is the comprehensive error correction term after considering the estimation deviation of the Coulomb friction dynamic component, the estimation error of the viscous friction coefficient, and the estimation error of the load torque. For residual terms, observation error , ΔT clb =T clb -T clb ' This represents the Coulomb friction estimation error. To ensure... The switching gain p1 must satisfy:

[0040] ,

[0041] in, , , These are the upper bounds of G1, state estimation error e1, and residual term R1 in the corresponding observer, respectively.

[0042] Once the sliding motion is established, there is The error dynamic equation can be obtained as follows:

[0043] ,

[0044] When f1G1R1 is controlled to be near zero The exponential convergence has a solution that is: Therefore, f1 > 0 is required; the value of f1 affects the convergence speed and steady-state accuracy. The larger the f1, the faster the convergence but the greater the chattering. The smaller the f1, the higher the steady-state accuracy but the slower the convergence.

[0045] Similarly, for observer Ω2, the switching gain p2 must satisfy... Feedback gain f2 > 0; for observer Ω3, switching gain p3 must satisfy... Feedback gain f3 > 0;

[0046] For observer Ω4, the switching gain p4 must satisfy... Feedback gain f4 > 0; after sliding mode motion is established, T clb estimation error e clb satisfy:

[0047] ,

[0048] in When f4G4R4 and F are controlled near zero, e clb Exponential convergence;

[0049] For observer Ω5, the switching gain p5 must satisfy... Feedback gain f5 > 0; for observer Ω6, switching gain p6 must satisfy... Feedback gain f6 > 0;

[0050] In practical engineering applications, the typical range of values ​​for each gain coefficient is as follows: switching gain p i ∈ [-10000, -100] (i=1,…,6), feedback gain f1∈ [0.5, 10], f2∈ [1e -8 , 1e -5 ],f3∈ [1e -5 , 1e -3 f4∈[0.001, 0.1], f5∈[1, 50], f6∈[5, 100]; the specific values ​​need to be adjusted according to factors such as system power level and sampling time.

[0051] Furthermore, the design of the cascaded structure is based on parameter error boundary analysis: in step 4.1, due to ΔT clb ≠0, e B e Tl Restricted to ΔT clb Within the relevant boundaries, the boundary expression is:

[0052] ,

[0053] ,

[0054] ,

[0055] By increasing |G1| and |G2| using high acceleration commands, these boundaries can be effectively compressed; in step 4.2, due to the use of low acceleration commands, and ω m Small enough to make , Therefore, the identification error in step 4.1 has a negligible impact on the identification of the second set of parameters, ensuring the identification accuracy of the second set of parameters.

[0056] The method of this invention only needs to utilize the phase current and rotor position information of the servo system, and through the designed parallel cascaded extended sliding mode observer, it can accurately identify all the key mechanical parameters including the LuGre model during the self-tuning process.

[0057] Furthermore, the present invention also proposes an electronic system comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the method proposed in the present invention.

[0058] On the other hand, the present invention also proposes a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the method proposed in the present invention.

[0059] Compared with existing mechanical parameter identification strategies, the present invention has the following beneficial technical effects:

[0060] 1. This invention can achieve simultaneous and accurate identification of up to six mechanical parameters, which is the most precise solution for friction modeling in the prior art. Compared with the traditional strategy of identifying only 2-3 basic mechanical parameters, it can more comprehensively and realistically reflect the mechanical characteristics of the motor system and provide more complete parameter support for high-precision control.

[0061] 2. This invention introduces a simplified LuGre friction model in parameter identification, which considers the dynamic process of Coulomb friction near the zero-crossing point of velocity. This model is closer to the physical reality than the traditional method of simulating friction using only sign functions, and can effectively improve the control accuracy at low speeds and during commutation.

[0062] 3. This invention is the first to realize the mechanical parameter identification of an equivalent multi-inertia system, breaking through the limitation of existing strategies that are only applicable to single-inertia systems. It can effectively cover complex servo system scenarios with multiple transmission chains and multiple load couplings, greatly expanding the application scope of parameter identification technology and providing a brand-new solution for high-precision control of complex electromechanical systems.

[0063] 4. This invention effectively solves the strong coupling problem between rotational inertia, viscous friction coefficient, and Coulomb friction dynamic parameters by grouping parameters and designing a parallel cascade structure for alternating identification under different excitation signals, thus ensuring the convergence and accuracy of all parameter identifications. Theoretical analysis provides parameter error boundaries and convergence conditions, providing a basis for gain selection. By continuously executing multiple motion cycles in each identification step, the parameter estimates are ensured to converge sufficiently, improving the stability and reliability of the identification results.

[0064] 5. The entire identification scheme of this invention is based on an extended sliding mode observer, which has low computational complexity. It only requires motor phase current and position signals, and is easy to deploy in existing servo drive systems without the need for additional sensors or complex computing resources. The specific value ranges of the twelve gain coefficients of the six observers are given, which facilitates rapid tuning by engineers. Attached Figure Description

[0065] The invention will now be further described with reference to the accompanying drawings.

[0066] Figure 1 This is a block diagram of the overall structure for parameter identification of the parallel cascaded extended sliding mode observer proposed in this invention.

[0067] Figure 2 This is a schematic diagram of the structure of the contactless integrated permanent magnet vernier servo motor of the present invention.

[0068] Figure 3 This is an overall block diagram of the control system of the permanent magnet vernier motor based on the parameter identification strategy of parallel cascade extended sliding mode observer of the present invention.

[0069] Figure 4 This is a complete implementation flowchart of the sliding mode identification method for parallel cascaded expansion of mechanical parameters of permanent magnet vernier motor of the present invention.

[0070] Figure 5 This is a simulation verification result of the parallel cascade extended sliding mode identification method proposed in this invention on a permanent magnet vernier motor for identifying multiple mechanical parameters.

[0071] Figure 6 This is a simulation comparison of the speed step response of a permanent magnet vernier motor using the parallel cascade extended sliding mode parameter identification method of this invention.

[0072] Explanation of reference numerals in the attached diagram: 1—Stator of permanent magnet vernier motor; 2—Armature winding; 3—Permanent magnet of inner rotor of magnetic gear; 4—Permanent magnet of outer rotor of magnetic gear; 5—Integrated rotor; 6—Magnetic isolation block; 7—Permanent magnet of permanent magnet vernier motor; 8—Magnetic adjustment ring. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] Figure 1 This is a block diagram of the overall structure of the parallel cascaded extended sliding mode observer proposed in this invention. Addressing the problem of strong coupling of multiple mechanical parameters and difficulty in accurately identifying nonlinear friction in permanent magnet vernier motors, two sets of synchronous mechanical parameter identification strategies are constructed to achieve synchronous and high-precision online identification of all mechanical parameters. The A-group mechanical parameter identification module targets the equivalent moment of inertia J. eq viscous friction coefficient B eq Equivalent load torque T lThree types of high-speed sensitive parameters are used, and three parallel extended sliding mode observers Ω1, Ω2, and Ω3 are set up. The electromagnetic torque T output from the vernier motor is integrated through a multiplexer. e The mechanical angular velocity ω of the rotor inside the vernier clicking end m and the Coulomb dynamic friction torque T identified by Group B. clb As input; where observer Ω1 identifies the equivalent moment of inertia. Ω2 synchronously identifies the coefficient of viscous friction Ω3 represents the equivalent load torque After noise reduction via EMA exponential moving average filtering, the identification results of group A are output through a multiplexer and fed back to module B. The group B mechanical parameter identification module identifies the Coulomb dynamic friction torque T. clb Three types of low-speed nonlinear friction parameters, namely static friction stiffness σ0, relaxation coefficient τ, etc., are analyzed. Three parallel extended sliding mode observers Ω4, Ω5, and Ω6 are designed to receive the A-group identification parameters and the mechanical angular velocity ω. m As input, Ω4 identifies Coulomb dynamic friction. Ω5 Identification of Static Friction Stiffness Ω6 identifies relaxation coefficient Through the closed-loop cascading linkage and iterative calculation of two-level modules, the coupling between the inner and outer rotors and the strong correlation interference between parameters are gradually decoupled, and finally the accurate convergence and stable output of all mechanical parameters are achieved.

[0075] Figure 2 This is a schematic diagram of the non-contact integrated permanent magnet vernier servo motor of the present invention. The motor consists of a vernier motor part and a magnetic gear part, both of which adopt a coaxial integrated rotor structure. The vernier motor part includes: a motor stator 1, an armature winding 2, and a motor permanent magnet 7; the magnetic gear part includes: an inner rotor of the magnetic gear (including an inner rotor permanent magnet 3), an outer rotor of the magnetic gear (including an outer rotor permanent magnet 4), and a magnetic adjustment ring 8. Among them, the integrated rotor 5 serves as both the motor rotor of the vernier motor and the inner rotor of the magnetic gear, and the two are coaxially and rigidly connected to form a rigid integrated structure; the outer rotor of the magnetic gear is the terminal output rotor. This structure achieves non-contact magnetic force transmission through the magnetic gear, and the terminal output torque and the electromagnetic torque on the motor side satisfy a fixed reduction ratio relationship. Since the rotational inertia of the motor side and the load side are coupled through the magnetic gear, it is impossible to directly identify their independent inertia. Therefore, an equivalent mechanical parameter model is introduced: the inertia of the motor rotor, the inertia of the inner rotor of the magnetic gear, and the load inertia are converted to the motor side according to the square of the reduction ratio to form an equivalent rotational inertia; at the same time, the load torque is converted to an equivalent load torque according to the reduction ratio. Based on this equivalent model, an integrated motor with an internal and external rotor coupling structure can be equivalent to a single-mass rigid system, thus providing a unified model basis for subsequent mechanical parameter identification.

[0076] Figure 3This is the overall block diagram of the control system for a permanent magnet vernier motor based on a parallel cascaded extended sliding mode observer parameter identification strategy. The system adopts a dual closed-loop vector control architecture, with the inner rotor mechanical angular velocity given by ω. m ∗ With actual angular velocity ω m The deviation is taken as input, and the PI speed regulator outputs the basic q-axis current command; simultaneously, a parallel cascaded extended sliding mode observer is based on the actual angular velocity ω of the inner rotor. m Online estimation of equivalent inertia of motor Equivalent viscous friction coefficient Load torque Dynamic friction component Equal total disturbance, generating composite compensation current This is superimposed on the speed loop output to achieve disturbance feedforward compensation. The compensated q-axis current command is then compared with the actual q-axis current i. q The difference is calculated, with the d-axis using the i-axis. d ∗ =0 control and actual d-axis current i d The difference between the d and q axis voltages is output via a PI current regulator, which then generates a three-phase PWM drive signal via an SVPWM module to drive a three-phase inverter to power the motor. The three-phase stator currents of the motor are transformed by Park to obtain the actual d and q axis currents, and the inner rotor position angle θ is differentiated to obtain the actual angular velocity ω. m This forms a complete closed-loop control, enabling the permanent magnet vernier motor to operate with high precision and strong anti-disturbance capability.

[0077] Figure 4 This is a complete implementation flowchart of the parallel cascade extended sliding mode identification method for mechanical parameters of permanent magnet vernier motors proposed in this invention. It constructs a full-process identification system to address identification problems such as the coupling of internal and external rotors, strong correlation of multiple parameters, and frictional nonlinearity of permanent magnet vernier motors.

[0078] Step 1: In response to the coupling characteristics of the inner and outer rotors caused by the magnetic gear transmission of the permanent magnet vernier motor, Step 1 establishes an integrated equivalent dynamic model, which incorporates the dynamic characteristics of the inner and outer rotors and the magnetic gear reduction ratio into the single mass equivalent model. At the same time, a simplified LuGre friction model is introduced to fully characterize the dynamic transition process of Coulomb friction, providing a model basis for the accurate identification of nonlinear friction parameters.

[0079] Step 2: Based on the problem of easy coupling in the identification of multiple mechanical parameters, the mechanical parameters to be identified are divided into two groups: Group A is the equivalent moment of inertia J. eq viscous friction coefficient B eq Load torque T l It can be fully excited under high-speed and high-acceleration conditions; Group B consists of friction-related stiffness coefficient σ0 and Coulomb friction torque T. cIt has higher recognition accuracy under low speed and low acceleration conditions;

[0080] Step 3: A parallel cascaded extended sliding mode observer architecture was designed: The first parallel observer group Ω1, Ω2, and Ω3 are designed for the parameters of group A, and the parallel structure enables synchronous identification of multiple parameters; The second parallel observer group includes the Coulomb dynamic friction observer Ω4 and the extended sliding mode observers Ω5 and Ω6, which are specifically designed for the nonlinear characteristics of the friction parameters of group B, solving the problem of insufficient identification accuracy of traditional observers for low-speed friction parameters, and achieving decoupled identification of the two groups of parameters through the cascaded structure.

[0081] Step 4: For the excitation characteristics of the two sets of parameters, apply a first excitation signal with high acceleration and high speed to fully excite the dominant characteristics of the inertial torque of the A set of parameters under high speed, and use the first parallel observer to complete the identification; apply a second excitation signal with low acceleration and low speed to identify the frictional nonlinear characteristics and complete the identification of the B set of parameters, thereby improving the identification accuracy and convergence speed.

[0082] Step 5: Achieve full parameter convergence through a cascaded iteration mechanism: Using the identification results of the previous cascaded steps, update the model parameters of the subsequent cascaded observer in real time, gradually decouple the coupling effects of the inner and outer rotors, eliminate cross interference between parameters, until all mechanical parameters converge to accurate values ​​and are directly substituted into the control system.

[0083] Figure 5 The Simulink simulation results of the parallel cascaded extended sliding mode identification method proposed in this invention on a permanent magnet vernier motor comprehensively verify the accuracy, convergence, and robustness of online identification of all mechanical parameters under all operating conditions, providing sufficient simulation support for the effectiveness of the method. The entire simulation process is performed under square wave speed commands for parameter identification. Figure 5 (a) shows the identification results of inertial and viscous friction parameters in group A, corresponding to the identification performance of the first parallel observer under high acceleration and high velocity excitation, and the equivalent moment of inertia identification value. Only minor transient fluctuations occurred during the start-up phase and under sudden changes in mechanical speed, which then rapidly converged to 0.0142 kg⋅m. 2 The true value, the steady-state error is suppressed within ±3.7%, and the identified value of the viscous friction coefficient. The parameter remained stable at the set value throughout the entire process, with fluctuations of less than 0.5%. The embedded magnified local image further verified the steady-state accuracy of parameter identification, proving that the observer has strong robustness and high-precision identification capability for large inertia and high-speed parameters.

[0084] Figure 5(b) shows the results of load torque and speed tracking identification, which verifies the method's ability to estimate load disturbances in real time. The observed speed of the inner rotor can be completely synchronized with the actual speed command. There is no significant drop or overshoot under five periodic load change conditions, and the dynamic response time is less than 10ms. The load torque of the outer rotor is 10N⋅m. The equivalent load torque identification value can track load changes in real time without lag. The identification value converges to 1.69N⋅m, and the identification deviation is less than 3%, achieving accurate online estimation of load disturbances.

[0085] Figure 5 (c) shows the identification results of the static friction parameters in group B, corresponding to the identification performance of the second parallel observer under low acceleration and low velocity excitation, and the identified static friction stiffness value. The torque converges rapidly to the true value, with a convergence time of less than 0.2 s and a steady-state stability of 15.4 N⋅m / rad without significant fluctuations. The relative error between this value and the set value of 15 N⋅m / rad is approximately 2.67%. The steady-state value of the Coulomb friction torque is also shown. The parameter was maintained stably at 1.52 N⋅m throughout the entire process, with the set value at 1.5 N⋅m. The embedded magnified diagram clearly shows the dynamic process of the parameter from the initial value to convergence, verifying the observer's ability to accurately identify low-speed nonlinear friction parameters.

[0086] Figure 5 (d) shows the identification results of dynamic friction parameters, verifying the identification performance of dynamic friction parameters in the simplified LuGre model, and the identified relaxation coefficient values. Fast convergence to the true value, smooth dynamic response with no overshoot, and Coulomb dynamic friction torque identification value. The dynamic transition process of Coulomb friction was accurately tracked, and the nonlinear characteristics of friction from static friction to dynamic friction were fully characterized. The parallel cascade extended sliding mode identification method of this invention can realize the online accurate identification of all mechanical parameters of permanent magnet vernier motor, such as equivalent inertia, viscous friction, load torque, static / dynamic friction. The steady-state error of all parameters is less than 5%, and all parameters converge in a short time, which meets the real-time requirements of motor control system and provides accurate parameter support for subsequent high-precision composite disturbance rejection control.

[0087] Figure 6 The mechanical parameter identification method based on parallel cascaded extended sliding mode observer proposed in this invention is applied to the speed step response comparison results of a permanent magnet vernier motor servo control system. Figure 6 (a) shows the speed response curve under the control scheme without incorporating the identification results of this invention. Figure 6(b) shows the speed response curve under the control scheme incorporating the identification results of this invention, verifying the significant improvement in the system's dynamic control performance after using the precise mechanical parameters obtained from the identification for speed loop PI parameter tuning and current loop disturbance feedforward compensation. The motor reference speeds were set to a high speed of 200 rpm (20.94 rad / s) and a low speed of 125 rpm (13.09 rad / s). A step response test was performed from a standstill with a sudden speed increase command. The speed response characteristics, speed error, and dynamic indicators were compared under the two control conditions: without and with the identification results of this invention. The control scheme without the identification results uses fixed PI parameters and has no friction or load torque feedforward compensation, while the scheme with the identification results uses the identified equivalent moment of inertia J. eq Equivalent viscous friction coefficient B eq Coulomb friction torque T c These parameters are used for speed loop PI parameter self-tuning, and the load torque T is also used. l Coulomb friction dynamic component T clb The estimated value is used for current loop feedforward compensation.

[0088] Simulation results show that without the identification results, the maximum speed error of the system's step response reaches 2.95 rad / s, the steady-state speed error is 0.17 rad / s, and the settling time is long. However, after introducing the identification results of this invention, the maximum speed error is only 2.07 rad / s, which is 29.83% lower than the traditional solution; the steady-state speed error is 0.04 rad / s, and the settling time is only 85 ms, which is nearly 50.58% shorter than the 172 ms of the traditional solution. The speed response has no obvious oscillations, and the dynamic tracking performance is excellent.

[0089] This embodiment also proposes an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method steps of the present invention.

[0090] Finally, this embodiment also discloses a computer-readable storage medium, which stores a computer program, wherein when the computer program is executed by a processor, it employs the steps of any of the methods described in the above embodiments.

[0091] The computer program can be stored in a computer-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The computer-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the computer-readable medium includes, but is not limited to, the above-mentioned components.

[0092] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0093] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0094] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0095] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0096] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for identifying the mechanical parameters of a contactless permanent magnet vernier motor based on a parallel cascaded extended sliding mode observer, characterized in that, Includes the following steps: An integrated equivalent dynamic model is established on the permanent magnet vernier motor side, and a simplified LuGre friction model considering the Coulomb friction dynamic process is introduced. The mechanical parameters to be identified are divided into two groups: the first group of parameters includes the equivalent moment of inertia J. eq Equivalent viscous friction coefficient B eq and load torque T l The second set of parameters includes the stiffness coefficient σ0 and the Coulomb friction torque T from the simplified LuGre model. c ; Design a parallel cascaded extended sliding mode observer. Construct a first parallel observer structure to simultaneously identify the equivalent parameters on the motor side of the first set of parameters. Construct a second parallel observer structure to simultaneously identify the terminal output friction characteristics of the second set of parameters and estimate the Coulomb friction dynamic component T online. clb ; The two parallel observer structures are connected in a cascaded manner; During the self-tuning process, the following identification process is performed: a first excitation signal is applied, and the first parallel observer structure is activated so that the inertial torque and viscous friction torque dominate the electromagnetic torque, in order to identify the first set of parameters; A second excitation signal is applied to make the Coulomb friction torque dominate the electromagnetic torque, and the second parallel observer structure is activated to identify the second set of parameters; By using a cascaded structure, the above identification process is executed alternately, and the observer parameters of the subsequent cascaded step are updated using the identification results of the previous cascaded step, thereby gradually eliminating the coupling between parameters and obtaining the accurate values ​​of all mechanical parameters.

2. The method according to claim 1, characterized in that, The contactless integrated permanent magnet vernier servo motor includes a vernier motor section and a magnetic gear section, both adopting a coaxial rotor structure. The vernier motor rotor is coaxially connected to the inner rotor of the magnetic gear, while the outer rotor of the magnetic gear is the terminal output rotor. Considering the magnetic gear transmission ratio N, the system is equivalent to a single-mass rigid system. A simplified LuGre friction model is introduced to describe the dynamic process of Coulomb friction, establishing a simplified equivalent model for the joint motor as follows: , , Among them: J m J is the moment of inertia of the permanent magnet vernier motor. HD For a harmonic reducer, the moment of inertia of a magnetic gear is N, where N is the reduction ratio of the harmonic reducer, and J is J. eq J is the equivalent rotational inertia of the system. L Let ω be the moment of inertia of the load. m B represents the speed at the vernier motor end. eq The coefficient of viscous friction is... For ω m The first derivative, T clb T is the dynamic component of Coulomb friction. l This represents the load torque.

3. The method according to claim 2, characterized in that, The simplified LuGre friction model is expressed as follows: , , Where σ0 is the stiffness coefficient, T c Let be the Coulomb friction torque.

4. The method according to claim 1, characterized in that, The first parallel observer structure includes three parallel extended sliding mode observers Ω1, Ω2 and Ω3, which are used for synchronous identification. B eq and T l The identification values ​​are respectively ; The expression for observer Ω1 is: , The expression for observer Ω2 is: , The expression for observer Ω3 is: , in, It is the reciprocal of the moment of inertia. =1 / J eq T ei T is the electromagnetic torque output from the vernier motor terminal of the corresponding observer. ' clb T is calculated from the second set of parameter estimates. clb Value, the superscript ^ indicates the observed value, S i For the sliding surface of the corresponding observer, i.e. p i For the switching gain, f i For the feedback gain coefficient, G1= G2= G3= .

5. The method according to claim 1, characterized in that, The second parallel observer structure includes a component for estimating the Coulomb friction dynamic component T. clb An extended sliding mode observer Ω4, and two parallel extended sliding mode observers Ω5 and Ω6, are used to simultaneously identify the stiffness coefficient σ0 and relaxation coefficient τ=σ0 / T in the simplified LuGre model, respectively. c T c The frictional torque is the Coulomb friction torque. The observer Ω4 is constructed based on the system's equations of motion and LuGre dynamic equations, and its expression is as follows: , Where, ω m4 T is the motor speed. e4 For electromagnetic torque, , , These are the estimated values ​​of the reciprocal of the equivalent moment of inertia, the coefficient of viscous friction, and the load torque, respectively, identified by the first parallel observer structure. For T clb The estimated value, p4 is the switching gain, S4= -ω m4 Let f4 be the sliding surface, f4 be the feedback gain coefficient, and G4 = - , and Let σ0 and τ be the identification values ​​respectively, and sgn(·) be the sign function; The observer Ω5 is constructed based on a simplified LuGre model and is used to identify σ0. Its expression is: , in, For Ω5 pairs T clb The estimate, Here is the estimated value for Ω4, p5 is the switching gain, and S... 5= -T clb f5 is the sliding surface, and f5 is the feedback gain coefficient. The observer Ω6 is constructed based on a simplified LuGre model and is used to identify τ. Its expression is: , in, For Ω6 pairs T clb The estimate is p6, where p6 is the switching gain, and S6 = -T clb f6 is the sliding surface, and f6 is the feedback gain coefficient.

6. The method according to claim 1, characterized in that, The first excitation signal is a high-acceleration, high-speed command signal, used to increase the proportion of inertial torque and viscous friction torque in the electromagnetic torque; the second excitation signal is a low-acceleration, low-speed command signal, used to increase the proportion of Coulomb friction torque in the electromagnetic torque; both the first and second excitation signals are generated by the CNC system, using position commands with S-shaped speed and acceleration curves to avoid impacting the actuator.

7. The method according to claim 1, characterized in that, The alternating execution process of the cascaded structure includes: Step 4.1: Apply the first excitation signal, execute multiple motion cycles continuously, enable the first parallel observer structure, and set the initial estimate of the second set of parameters to zero, i.e., T. clb ' =0, synchronously identify the first set of parameters in each cycle to obtain the preliminary estimate J. eq (1) B eq (1) T l (1) ; Step 4.2: Apply the second excitation signal, execute multiple motion cycles continuously, activate the second parallel observer structure, and use the first set of parameter estimates obtained in Step 4.1 to identify the second set of parameters, thereby obtaining the stiffness coefficient σ0. (2) Coulomb friction torque T c (2) In this step, observer Ω4 estimates the dynamic component T of Coulomb friction in real time. clb The observer Ω6 identifies σ0 and τ in parallel, and uses τ=σ0 / T c T is obtained by conversion c ; Step 4.3: Apply the first excitation signal again, execute multiple motion cycles continuously, activate the first parallel observer structure, and reconstruct the Coulomb friction dynamic component T by simplifying the LuGre model. clb ' Re-identify the first set of parameters to obtain the precise value J. eq B eq T l ; Step 4.4: Apply the second excitation signal again, execute multiple motion cycles continuously, activate the second parallel observer structure, and re-identify the second set of parameters using the accurate first set of parameters obtained in Step 4.3 to obtain the accurate values ​​σ0 and T. c ; Steps 4.1 to 4.4 above constitute a complete cascaded identification loop. Multiple loops can be repeated according to the accuracy requirements until all parameters converge.

8. The method according to any one of claims 4 or 5, characterized in that, The switching gain p i and feedback gain coefficient f i The value of must satisfy the stability and convergence conditions of the observer, specifically: For each extended sliding mode observer Ω i Its switching gain p i The following conditions must be met to ensure that the sliding surface S i It is possible to reach and maintain a position near zero within a finite amount of time, i.e., satisfy the following inequality: ， in, , , G in the corresponding observer i State estimation error e i Residual term R i The upper bound of the absolute value condition; i=1,…,6, is determined by selecting negative values ​​p that satisfy the absolute value condition. i To ensure the accessibility of sliding mode motion; The feedback gain coefficient f i It must be greater than zero to ensure that once the sliding mode motion is established, the parameter estimation error can converge exponentially to zero, i.e., satisfy: ， At the same time, to ensure the non-homogeneous term f in the error equation i G i R i When the parameter is controlled to near zero in steady state, the design of the cascaded structure and excitation signal is required to ensure that the error term, which is independent of the parameter to be identified, is small enough in each step of identification to avoid parameter coupling and thus meet the actual convergence condition.

9. An electronic system comprising: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, characterized in that the instructions are executed by the at least one processor to enable the at least one processor to perform the method steps of any one of claims 1-8.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-8.

Images