High performance servo drive and control method

Abstract

The application provides a high-performance servo driver and a control method, which are applied to the technical field of high-performance servo driver control and include that a servo control loop module is arranged in a digital signal processor, and an AI learning algorithm module is arranged in a field programmable logic array; the AI learning algorithm module: based on operation parameters generated by a servo system where the high-performance servo driver is located, the AI learning algorithm is trained and learned to obtain a characteristic model and an output compensation current component; the servo control loop module: used for performing vibration suppression on the servo system according to the compensation current component. Compared with the prior art, the application can effectively suppress vibrations of arbitrary amplitudes, vibrations of arbitrary time lengths and vibrations caused by arbitrary factors through vibration suppression of the compensation current component; the vibration suppression is performed at the cycle pace of a current loop, so that there is no delay or the delay can be completely ignored relative to a mechanical system, and high-speed or super-high-speed operation occasions can be completely adapted.

Description

Technical Field

[0001] This application relates to the field of high-performance servo drive control technology, specifically to a high-performance servo drive and control method. Background Technology

[0002] High-performance servo drives are widely used in high-end, high-precision intelligent manufacturing equipment, such as machine tools, robots, aerospace, and medical devices. High-performance servo drives require high positioning accuracy and high stability. A servo system is itself an electromechanical system; ensuring positioning accuracy and preventing vibration of the mechanical system, or even promptly and quickly suppressing vibration if it does occur, is a significant challenge when a servo motor drives a mechanical system to perform rotational or feed motions.

[0003] In practical systems, even if a servo system is equipped with a high-resolution motor encoder and has high-precision drive control hardware and software configuration and control performance, vibration or even resonance in the mechanical system during operation can still negatively impact servo positioning accuracy, leading to increased positioning errors, failure to meet requirements, and even fatal consequences for the mechanical system itself. Many factors can cause servo system vibration, such as inappropriate control parameter settings, external disturbances to the mechanical system, changes in servo system load, system resonance, etc. Even brief, slight vibrations can degrade servo system performance. Therefore, if a high-performance servo driver cannot effectively suppress vibration, even with high-level hardware and software configuration, it cannot fully guarantee high performance and high precision.

[0004] Under current technological conditions, vibration suppression methods generally fall into two categories: offline and online suppression. Offline methods require pre-calculation of the mechanical system's resonant frequency, followed by setting a filter with the corresponding frequency value in the servo control system to eliminate vibrations at that frequency. The disadvantages of this method are: it requires pre-calculation of the system's resonant frequency; it is ineffective for vibrations at frequencies other than the resonant frequency; and it cannot eliminate minor or short-term vibrations caused by other factors. Online vibration suppression utilizes adaptive algorithms, online resonant frequency identification, or external vibration monitoring sensors to suppress mechanical vibrations in real time. This method is superior to offline methods, but the technology is still far from mature, and different companies may adopt different technical measures. In general, this method has certain threshold requirements for vibration amplitude and duration to facilitate identification and calculation by the control system; therefore, it is ineffective for short-term vibrations below the threshold amplitude. Furthermore, most of these methods, given the current technological level, are based on the principle of identifying and tracking changes in vibration frequency parameters such as vibration frequency and amplitude through algorithms, thereby calculating the vibration frequency value in real time and setting corresponding filters. Because this method requires complex algorithms, it places certain demands on the computing power of the control core, and the resulting vibration suppression effect has a certain lag effect. Therefore, it has certain limitations for high-speed operating systems and may not achieve a good vibration suppression effect, or even fail. Under certain special load conditions, it may not work at all.

[0005] As for systems with external vibration monitoring sensors, they increase system cost, complexity, and reliability, and are not recommended unless absolutely necessary.

[0006] Therefore, under current technological conditions, there is no effective and high-performance vibration suppression and control method to address the various vibration problems faced by high-performance servo drive systems.

[0007] Therefore, a new, adaptable, and high-performance vibration suppression method is needed to address the various vibration problems faced by high-performance servo drive systems. Summary of the Invention

[0008] In view of this, the present invention provides a high-performance servo driver and control method. It utilizes AI learning algorithms to establish a feature model of the high-performance servo driver, continuously trains and iterates the feature model, and identifies vibration-related compensation current components from the feature model. These compensation current components are then fed into the current loop controller in the servo control loop module to suppress vibration. Any slight vibration, short-term vibration, mechanical resonance, etc., can be identified and compensated for, thereby achieving high-performance servo performance.

[0009] This invention provides the following technical solutions:

[0010] This invention provides a high-performance servo driver, comprising: a servo control loop module and an AI learning algorithm module;

[0011] The servo control loop module is located in the digital signal processor, and the AI ​​learning algorithm module is located in the field-programmable logic array.

[0012] AI learning algorithm module: Used to train and learn the AI ​​learning algorithm based on the operating parameters of the servo system where the high-performance servo driver is located, to obtain the feature model and the compensation current component output by the feature model.

[0013] Servo control loop module: Used to suppress vibration in the servo system based on the compensation current component output by the AI ​​learning algorithm module.

[0014] Preferably, the operating parameters include: the command current value of the motor, the angle signal output by the motor encoder, and the timing of the pulse width modulation signal generation.

[0015] Preferably, the servo control loop module includes: a position loop controller, a speed loop controller, and a current loop controller.

[0016] Preferably, the servo control loop module inputs the compensation current component into the current loop controller for vibration suppression.

[0017] Preferably, the digital signal processor and the field-programmable logic array exchange data via a data bus.

[0018] Preferably, the field-programmable logic array has an external random access memory and a flash memory.

[0019] Preferably, the field-programmable logic array includes: an AI learning algorithm module, a data acquisition and processing module, a pulse width modulation generation module, and an encoder interface module.

[0020] Preferably, the data acquisition and processing module simultaneously acquires the operating parameters and stores them in the flash memory; the AI ​​learning algorithm module loads the operating parameters into the random access memory, processes them, and outputs the compensation current component.

[0021] The present invention also provides a high-performance servo driver control method, comprising:

[0022] Step 1: Based on the operating parameters of the servo system where the high-performance servo driver is located, the learning algorithm is trained to obtain the feature model and the compensation current component output by the feature model.

[0023] Step 2: Suppress vibration in the servo system based on the compensation current component.

[0024] Preferably, the operating parameters include: the command current value of the motor, the angle signal output by the motor encoder, and the timing of the pulse width modulation signal generation.

[0025] Compared with the prior art, the beneficial effects that the above-mentioned at least one technical solution adopted by the present invention can achieve include at least the following: it can effectively suppress vibrations of arbitrary amplitude, vibrations of arbitrary duration, and vibrations caused by any factors; it executes in the periodic step of the current loop controller, therefore, there is no delay, or the delay is completely negligible relative to the mechanical system, and it can be fully adapted to high-speed or ultra-high-speed operating occasions; it has virtually no restrictions on the type of mechanical load, can adapt to any mechanical load, and can train a real-time model through the AI ​​learning algorithm module and continuously correct it, therefore, the high-performance servo drive has high adaptability. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic diagram of the control structure of a high-performance servo driver according to this application;

[0028] Figure 2 This is a schematic diagram of the hardware configuration of a high-performance servo driver according to this application;

[0029] Figure 3 This is a schematic diagram illustrating an input parameter implementation method for an AI learning algorithm module in this application;

[0030] Figure 4 This is a block diagram of an AI learning algorithm module in this application;

[0031] Figure 5 This is a schematic diagram of the control structure of another high-performance servo driver in this application;

[0032] Figure 6This is a schematic diagram of the vibration suppression result of a synchronous belt mechanical load position control according to this application. Detailed Implementation

[0033] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0034] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this application, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0036] It should also be noted that the illustrations provided in the following embodiments are only for illustrating the basic concept of this application and do not constitute any limitation on the content of this application.

[0037] Additionally, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that practice can be carried out without these specific details.

[0038] In existing technologies, methods for suppressing vibrations in high-performance servo drives are generally divided into two types: offline setting and online suppression.

[0039] The offline setup method requires pre-calculating the resonant frequency of the mechanical system, and then setting a filter with the corresponding frequency value in the servo control system to eliminate vibrations at that frequency. The disadvantages of this method are: it requires pre-calculating the system's resonant frequency; it is ineffective for vibrations of other frequency components besides the resonant frequency; and it cannot eliminate minor or short-term vibrations caused by other reasons.

[0040] Online vibration suppression methods utilize adaptive algorithms, online resonant frequency identification, or external vibration monitoring sensors to suppress mechanical vibrations in real time. This method is superior to offline methods, but the technology is still far from mature, and different companies may adopt different technical measures. Generally speaking, this method has certain threshold requirements for vibration amplitude and duration to facilitate identification and calculation by the control system. Therefore, it is ineffective for short-duration vibrations below the threshold amplitude. Furthermore, most of these methods, at the current technological level, are based on algorithms that identify and track changes in vibration frequency and amplitude, thereby calculating the vibration frequency value in real time and setting corresponding filters. Because this method uses complex algorithms, it places certain demands on the computing power of the control core, and the resulting vibration suppression effect has a certain lag effect. Therefore, it has limitations for high-speed operating systems and may not achieve good vibration suppression results, or even fail; under certain special load conditions, it may not work at all.

[0041] As for systems with external vibration monitoring sensors, they increase system cost, complexity, and reliability, and are not recommended unless absolutely necessary.

[0042] In view of this, the inventors found that there is a need for a more adaptable and high-performance vibration suppression and control method to deal with various vibration problems faced by high-performance servo drive systems.

[0043] Based on this, the present invention proposes a processing scheme: using AI learning algorithms to establish a feature model of the servo system in which the high-performance servo driver is located, continuously training, learning and iteratively calculating the feature model, extracting the current component related to the vibration of the servo system from the feature model, and using this current component as a compensation current component, feeding the compensation current component into the current loop in the servo control loop module to achieve the purpose of suppressing vibration.

[0044] The technical solutions provided by the various embodiments of this application are described below with reference to the accompanying drawings.

[0045] This invention provides a high-performance servo drive, comprising: a servo control loop module and an artificial intelligence (AI) learning algorithm module; wherein, the servo control loop module is disposed in a digital signal processor (DSP), and the AI ​​learning algorithm module is disposed in a field-programmable gate array (FPGA); the AI ​​learning algorithm module is used to train and learn based on the operating parameters of the servo system in which the high-performance servo drive is located, to obtain a feature model and a compensation current component output by the feature model; the servo control loop module is used to suppress vibration of the servo system according to the compensation current component output by the AI ​​learning algorithm module.

[0046] The operating parameters include: the command current value of the motor, the angle signal output by the motor encoder, and the timing of the pulse width modulation (PWM) signal generation.

[0047] Specifically, the high-performance servo driver adopts a hardware architecture combining DSP and FPGA. The algorithm in the servo control loop module mainly runs in the DSP, while the AI ​​learning algorithm in the AI ​​learning algorithm module mainly runs in the FPGA. By training and learning parameters such as the current command based on the actual motor, the output signal of the motor encoder combined with the timing of the signal generated by Pulse Width Modulation (PWM), and other parameters, a feature model of the servo system is established using the AI ​​learning algorithm. The feature model is continuously trained, learned, and iteratively calculated, and the compensation current component related to the vibration of the servo system is extracted from the feature model. Then, by feeding the compensation current component into the current loop controller in the servo control loop module, the vibration is suppressed. Thus, any slight vibration, short-term vibration, and mechanical resonance can be identified and compensated for and eliminated.

[0048] Specifically, the command current value of the motor is obtained based on the actual current command of the motor, and the output angle signal of the motor encoder is obtained based on the output signal of the motor encoder.

[0049] This application can effectively suppress vibrations of any amplitude, any duration, and caused by any factor. It executes in cycle-by-cycle mode according to the current loop controller, therefore, there is no delay or the delay is negligible relative to the mechanical system, making it fully adaptable to high-speed or ultra-high-speed operating environments. This application has virtually no restrictions on the type of mechanical load and can adapt to any mechanical load. It can train a real-time model through AI learning and continuously correct it, thus providing high adaptability for high-performance servo drives.

[0050] This application employs an AI learning algorithm running in an FPGA. By training and learning several operating parameters, a feature model of the servo system is obtained, thereby extracting the compensation current component related to vibration factors. This compensation current component is then directed to a current loop controller running in a DSP for compensation, thereby achieving the purpose of vibration suppression.

[0051] The high-performance servo driver provided by this invention will be described below.

[0052] like Figure 1 As shown, the high-performance servo driver provided in this application includes a servo control loop module and an AI learning algorithm module. The servo control loop module includes a position loop controller, a speed loop controller, and a current loop controller. The servo control loop module also includes a differentiator and a PWM generation module. The servo control loop module inputs the compensation current component to the current loop controller to suppress vibration in the motor and load system. Wherein, θ cmd Kp_p represents the position command of the position loop; Kv_p represents the gain of the position loop controller; Kv_p represents the proportional gain of the speed loop controller; Kv_i represents the integral gain of the speed loop controller; S represents the Laplace operator; Ki_p represents the proportional gain of the current loop controller; Ki_i represents the integral gain of the current loop controller; i comp This represents the current compensation component output by the AI ​​learning algorithm module, i.e., the compensation current component; cmd This indicates the input current command to the current loop controller, i.e., the commanded current value for the motor; t p This represents the pulse train at the edge of the PWM waveform taken from the PWM generation module; θ represents the output angle of the motor encoder; i fdb This represents the motor feedback current. The servo control loop module consists of a position loop controller, a speed loop controller, and a current loop controller. The input parameters for the AI ​​learning algorithm module are the command current value, the motor encoder angle θ, and the PWM waveform edge timing pulse train t taken from the PWM generation module. p The AI ​​learning algorithm module outputs the current compensation component i. comp The data is fed into the current loop controller input command.

[0053] Furthermore, the digital signal processor (DSP) and the field-programmable logic array (FPGA) exchange data via a data bus.

[0054] Specifically, this application adopts a DSP+FPGA hardware configuration mode. The position loop controller, speed loop controller, and current loop controller are executed in the DSP; the AI ​​learning algorithm module is mainly completed in the FPGA; at the same time, the DSP and FPGA exchange data through a data bus. The FPGA has a large-capacity RAM and a large-capacity Flash Disk to store the running data for the algorithm in the FPGA to call and learn.

[0055] In this invention, the field-programmable logic array (FPGA) is externally equipped with random access memory (RAM) and flash memory. Further, the FPGA includes: an AI learning algorithm module, a data acquisition and processing module, a pulse width modulation (PWM) generation module, and an encoder interface module. The data acquisition and processing module simultaneously acquires operating parameters and stores them in the flash memory; the AI ​​learning algorithm module loads the operating parameters into the RAM, processes them, and outputs a compensation current component. Specific configurations are as follows... Figure 2 As shown, the FPGA contains the following logic block areas: an AI learning algorithm module, a data acquisition and processing module, a PWM generation module, and an encoder interface module. The data acquisition and processing module primarily handles the i... cmd The DSP performs data acquisition and processing. cmd The data is input to the FPGA's data acquisition and processing module via the data bus for further processing, resulting in i cmd * The processing result i cmd * On one hand, the signal is sent to the AI ​​learning algorithm module, and on the other hand, it is sent to the PWM generation module. The PWM generation module generates PWM trigger pulses, namely Space Vector Pulse Width Modulation (SVPWM) based on space vector control, and outputs them to the external power switch trigger circuit. Simultaneously, the PWM generation module generates a pulse train t based on the edge time of each PWM pulse. p , pulse train t p After standardized processing, the signal is sent to the AI ​​learning algorithm module as its operating clock source. The PWM generation module generates the trigger pulses for the six switching transistors required by the three-phase bridge topology circuit. The motor encoder output signal is directly sent to the encoder interface module in the FPGA, where it is decoded into an angle value θ. θ can be simultaneously sent to both the AI ​​learning algorithm module and the DSP. The FPGA has external large-capacity RAM and Flash Disk to store operating data for AI learning and training. The AI ​​learning algorithm module outputs current compensation data i. compTo the current loop controller in the DSP.

[0056] The theoretical basis of this application will be explained below, which will provide a foundation for the AI ​​learning algorithm of this application.

[0057] The kinematic equation of the electric motor can be expressed by formula (1):

[0058]

[0059] Among them, T e T represents the electromagnetic torque of the motor. L J represents the motor load; J represents the inertia of the motor and load, which is considered a constant in this application; ω represents the motor speed; and B represents the viscosity coefficient.

[0060] For the sake of simplicity in this application, B is ignored as 0; therefore, formula (1) becomes formula (2):

[0061]

[0062] Furthermore, assume the motor's torque coefficient is K. t Then the electromagnetic torque T e This can be expressed as formula (3):

[0063] T e =i*k t (3)

[0064] Where i represents the current of the motor, the kinematic equation (1) can be written as formula (4):

[0065]

[0066] Taking the difference between both sides of equation (4) yields equation (5):

[0067]

[0068] Calculating formula (5) yields formula (6):

[0069]

[0070] Furthermore, due to formula (7):

[0071]

[0072] Then we can obtain formula (8):

[0073]

[0074] When Δt is sufficiently small, we consider ΔT LIf ≈0, that is, assuming the load remains constant, then we can finally obtain formula (9).

[0075]

[0076] Equation (9) shows that when the motor system is examined within a sufficiently small unit time period, the dynamic characteristics of the motor system are only related to three parameters: the change in current, the change in motor angle, and the unit time. Therefore, by using an AI learning algorithm to learn and train the system features in real time based on these three parameters, the feature model of the servo system can be obtained, which can be represented by Equation (10).

[0077] f(Δi, Δθ, Δt); (10)

[0078] Where Δi represents the change in current, i here refers to... cmd The change; Δθ represents the change in motor angle;

[0079] The value here is taken from the output change value of the motor encoder; Δt represents the unit time length used; the pulse train generated based on the PWM pulse edge is used as the regular clock source.

[0080] The AI ​​learning algorithm module provided in this application is as follows: Figure 3 As shown, the AI ​​learning module has three input parameters and one output parameter; the three input parameters are i cmd Δi represents the command current value taken from the current control loop running in the DSP, which is actually decomposed into Δi after entering the AI ​​learning algorithm module; θ represents the angle signal taken from the motor encoder output, which is actually decomposed into Δθ after entering the AI ​​learning algorithm module; Δt represents the clock source for the AI ​​learning algorithm module to perform learning and training. Figure 3 It can be seen that in the SVPWM waveform of vector control, the moment of edge transition of each SVPWM waveform is captured as a pulse train t. p After standardization, a regular clock source with Δt as the time unit is generated, which serves as the clock basis for sampling Δi and Δt within the AI ​​module. Based on this principle, Δt can be calculated using formula (11).

[0081] Δt = PWM period / (n+1); (11)

[0082] Here, n represents the SVPWM generation mode. When SVPWM is 5-segment, n = 5; when SVPWM is 7-segment, n = 7. Therefore, from the perspective of operating speed, the AI ​​learning algorithm module's algorithm operation speed is 5 to 7 times that of the PWM frequency. From the perspective of the edge time of the Δt pulse train, it occurs at the moment of each power switch switching. Theoretically, each power switch switching will be accompanied by or cause fluctuations in the current or voltage applied to the motor armature. These fluctuations will inevitably be reflected in the slight vibration of the motor's rotating shaft, and these slight vibrations will inevitably manifest in the output signal of the high-resolution encoder. The encoder signal output is directly interfaced into the FPGA, and is synchronously sampled at high speed at Δt time intervals within the FPGA. Therefore, theoretically, the maximum sampling delay of the AI ​​learning algorithm module from the switch state to the encoder output state change is Δt time. As the PWM switching frequency increases, Δt further decreases, thus further improving the synchronization, real-time performance, and execution speed of the AI ​​module's learning algorithm. In theory, any minute change caused on the motor shaft, whether regular or irregular, such as mechanical vibration, can be captured by the FPGA on the high-resolution encoder. This is a prerequisite for AI learning algorithms to extract the dynamic model features of the system through iterative learning, and then extract the current change value related to the vibration, even slight vibration, as a feedback quantity to eliminate the vibration.

[0083] One output parameter of the AI ​​learning algorithm module is i comp Once the AI ​​learning algorithm module has trained and established a feature model that fully reflects the dynamic characteristics of the system, the compensation current component is extracted and fed into the actual current loop controller, thereby eliminating the vibration of the actual servo system. It should be noted that the i output by the AI ​​learning algorithm module... comp The parameters are used for feedback correction with the time period of the current control loop as the step size, thereby suppressing vibration. The step size is 1 / fc, where fc represents the update frequency of the current loop controller.

[0084] The AI ​​learning algorithm in the AI ​​learning algorithm module is mainly based on a training model with three parameters: Δt, Δi, and Δθ. Through continuous iteration, learning, training, and correction, the AI ​​learning algorithm can optimally find the intrinsic relationship between these three parameters, i.e., obtain a feature model. Then, based on the irregular law (vibration) of Δθ, it can find i related to this irregular vibration. comp This is then fed into the current loop controller to eliminate vibration.

[0085] Furthermore, the operating parameters are stored in a large-capacity Flash Disk and then loaded into RAM sequentially for use by the learning, iteration, and model training algorithms.

[0086] The workflow of the AI ​​learning algorithm module in this application is described below.

[0087] like Figure 4 As shown, the workflow of the AI ​​learning module is as follows: The data acquisition and processing module in the FPGA simultaneously acquires three parameters, Δt, Δi, and Δθ, with the acquisition cycle based on Δt, and stores them in the Flash disk for subsequent use by the AI ​​learning algorithm module. The AI ​​learning algorithm module loads the acquired data (Δt, Δi, and Δθ) into the RAM space in batches for feature data processing and extraction. The optimization iteration criterion of the AI ​​learning algorithm module is to minimize the correlation between the change of Δi and the irregular Δθ within the k*Δt time period, i.e., to minimize vibration. The so-called irregular Δθ refers to angle changes other than the regular angle value increments caused by normal motor operation, which are all classified as irregular values; for example, sudden changes or anomalies in angle caused by vibration. Here, k is a positive integer and can be set to a certain value. Because, according to the principle of Δt generation, in actual systems, the higher the PWM switching frequency, the smaller Δt becomes. This may result in Δθ being 0 for a long period of time, which is equivalent to a lot of invalid data. This wastes resources for the iterative calculation of the AI ​​learning algorithm. Therefore, it is appropriate to increase the unit time length by multiplying it by a positive integer k to adjust the unit time length in order to improve the speed and efficiency of AI operation, but the principle is to avoid causing adverse results.

[0088] Theoretically, the Δθ of the rule can be calculated using formula (12).

[0089]

[0090] Where ω represents the motor speed, measured in radians per second (rad / sec); P encoder Indicates encoder resolution; Δθ is in radians (rad).

[0091] Furthermore, after setting the constraints, the feature model f(Δi, Δθ, Δt) is continuously optimized through iterative learning. Based on the iterative learning, the value of Δi most relevant to the irregular Δθ is obtained, and this value is fed into the current loop controller to correct the irregular Δθ until Δθ is minimized, thereby suppressing the corresponding vibration component. The constraints are: Δi is sampled within the k*Δt time period; Δθ is differentiable twice consecutively within k*Δt; k is a positive integer greater than or equal to 1, which can be determined by the algorithm's autonomous learning or set.

[0092] Example 1

[0093] The embodiments of this application are also used to suppress end vibrations of mechanical mechanisms.

[0094] like Figure 5 As shown, a servo motor drives a synchronous belt mechanical load. The AI ​​learning algorithm module not only takes the original three parameters as input, but also the output θ of the linear encoder installed on the synchronous belt side. L It is also integrated into the AI ​​learning algorithm module. The AI ​​learning algorithm module then learns, trains, and models a model with four input parameters, which can be represented by formula (13).

[0095] f(Δi, Δθ, Δθ) L ,Δt); (13)

[0096] like Figure 5 In the configuration shown in formula (13), the high-performance servo drive needs to suppress the mechanical vibration at the mechanical load end, which is derived from the parsing of the load encoder information. At this time, the output Δθ of the motor encoder is considered a regular quantity, and Δ(Δθ) L -Δθ) is used as an irregular quantity.

[0097] like Figure 6 As shown, the position curves without vibration compensation and those with vibration compensation via the AI ​​learning algorithm module are displayed. The horizontal axis represents time (t) in units of k*Δt, and the vertical axis represents the normalized angle value θ. It can be seen that before compensation, in the time period from 0 to t1 without vibration compensation, the position curve of the synchronous belt mechanical load exhibits a certain degree of vibration spikes. The curve containing θ represents the curve of the motor and load system, which is relatively free of spikes. L The graph of the synchronous belt mechanical load shows a certain degree of vibration spikes. This vibration is mainly high-frequency, small-amplitude. Analysis of the vibration frequency band reveals that the vibration frequency changes constantly with the operation of the synchronous belt mechanical load. When the AI ​​learning algorithm module f(Δi, Δθ, Δθ)... L After compensation (Δt), the vibration spikes on the position curve are visibly eliminated, demonstrating a significant improvement. Furthermore, because this compensation is based on the frequency fc of the current control loop's execution cycle, the delay effect is negligible, allowing the synchronous belt to operate at high speeds without vibration. This improves the system's instantaneous positioning accuracy, further enhances system stability and reliability, and strengthens the servo system's performance.

[0098] The present invention also provides a high-performance servo driver control method, comprising:

[0099] Step 1: Based on the operating parameters of the servo system where the high-performance servo driver is located, the AI ​​learning algorithm is trained to obtain the feature model and the compensation current component output by the feature model.

[0100] Step 2: Suppress vibration in the servo system based on the compensation current component.

[0101] In one alternative implementation, the operating parameters include: the commanded current value of the motor, the angle signal output by the motor encoder, and the timing of the pulse width modulation signal generation.

[0102] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the product embodiments described later are relatively simple since they correspond to the methods; relevant parts can be referred to the descriptions in the system embodiments.

[0103] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A high-performance servo driver, characterized in that, include: Servo control loop module and AI learning algorithm module; The servo control loop module is located in the digital signal processor and includes at least a current loop controller. The AI ​​learning algorithm module is located in the field programmable logic array (FPGA), which includes at least a PWM generation module, an encoder interface module, and a data acquisition and processing module. The PWM generation module is used to generate PWM trigger pulses and generate pulse trains based on the edge transition times of the PWM trigger pulse waveforms. The processed pulse trains are used as the clock source for the AI ​​learning algorithm module. The encoder interface is used to acquire the angle signal output by the decoded motor encoder and send the angle signal to the AI ​​learning algorithm module to obtain the angle signal change Δθ. The data acquisition and processing module is used to acquire and process the motor command current value and send the motor command current value to the AI ​​learning algorithm module to decompose it into the change Δi of the motor command current value. The AI ​​learning algorithm module is used to acquire the operating parameters of the servo system where the high-performance servo driver is located. The operating parameters include the change in motor command current value Δi, the change in angle signal output by the motor encoder Δθ, and the clock pulse train Δt generated based on the power switch switching time of the PWM generation module. The AI ​​learning algorithm module uses the clock pulse train Δt as the clock reference and performs online training based on Δi, Δθ, and Δt to obtain a feature model f(Δi, Δθ, Δt). It outputs a compensation current component related to the irregular change in Δθ in the servo system from the feature model. The optimization iteration criterion of the AI ​​learning algorithm module is that the change in Δi has the least correlation with the irregular Δθ within the k*Δt time period, i.e., the vibration is minimal. The irregular Δθ refers to the irregular value of all angle changes except for the regular angle value increment caused by the normal operation of the motor. The AI ​​learning algorithm module also includes constraints for optimizing the feature model f(Δi, Δθ, Δt). The constraints are: Δi collects Δθ within the k*Δt time interval; Δθ is differentiable twice consecutively within k*Δt; and k is a positive integer greater than or equal to 1. The servo control loop module is used to feed the compensation current component output by the AI ​​learning algorithm module into the current loop controller with the current loop control cycle as the step size, so as to suppress vibration of the servo system.

2. The high-performance servo driver according to claim 1, characterized in that, The servo control loop module includes: a position loop controller and a speed loop controller.

3. The high-performance servo driver according to claim 1, characterized in that, The digital signal processor and the field-programmable logic array exchange data via a data bus.

4. The high-performance servo driver according to claim 1, characterized in that, The field-programmable logic array has an external random access memory and a flash memory.

5. The high-performance servo driver according to claim 4, characterized in that, The field-programmable logic array includes: the AI ​​learning algorithm module, the data acquisition and processing module, the pulse width modulation generation module, and the encoder interface module.

6. The high-performance servo driver according to claim 4, characterized in that, The data acquisition and processing module simultaneously acquires the operating parameters and stores them in the flash memory; the AI ​​learning algorithm module loads the operating parameters into the random access memory, processes them, and outputs the compensation current component.

7. A high-performance servo driver control method, characterized in that, Applied to any one of the high-performance servo drives according to claims 1-6, the method includes: Step 1: The AI ​​learning algorithm module obtains the operating parameters of the servo system where the high-performance servo driver is located. The operating parameters include the change in the command current value of the motor Δi, the change in the angle signal output by the motor encoder Δθ, and the clock pulse train Δt generated based on the power switch switching time of the pulse width modulation generation module. The AI ​​learning algorithm module uses the clock pulse train Δt as the clock reference and performs online training based on Δi, Δθ, and Δt to obtain the feature model f(Δi, Δθ, Δt). From the feature model, the compensation current component related to the irregular change of Δθ in the servo system is output. The optimization iteration criterion of the AI ​​learning algorithm module is that within the k*Δt time period, the change of Δi has the least correlation with the irregular Δθ, that is, the vibration is minimal. The irregular Δθ refers to the angle changes other than the regular angle value increment caused by the normal operation of the motor, which are all classified as irregular values. Step 2: Using the current loop control cycle as the step size, feed the compensation current component output by the AI ​​learning algorithm module into the current loop controller to suppress vibration in the servo system.

Images