Feedforward control method and system for electric actuator

Abstract

The invention relates to the technical field of electric actuator motion control, in particular to a feedforward control method and system for an electric actuator, and the method comprises the steps: obtaining the operation data of the whole stroke of the electric actuator at a preset speed; carrying out data normalization; selecting any curve fitting algorithm from the algorithm library, selecting any regression model from the model library to obtain algorithm model combinations, performing curve fitting to obtain fitted curves, and repeating selection and fitting operations to obtain fitted curves corresponding to all the algorithm model combinations; calculating scores of fitting curves corresponding to the algorithm model combinations, and determining an optimal algorithm model combination; calculating a feed-forward current value of the electric actuator at the current position based on the optimal algorithm model combination; and determining a control current value according to the feed-forward current value and the target current value. According to the embodiment of the invention, the influence of control lag of the electric actuator is eliminated in a feedforward mode, so that the response speed of the electric actuator is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of electric actuator motion control, and in particular to a feedforward control method and system for an electric actuator. Background Art

[0002] Voice coil electric actuators have the advantages of simple structure, fast dynamic response speed, no cogging torque, high linearity, long life, fast response, etc. They are widely used in high-speed and high-precision positioning systems such as high-precision positioning platforms, image stabilization systems, precision medical equipment, and flexible robots. In practical applications, due to the influence of mechanical structure and load, there will be some external forces (such as gravity, elastic force, magnetic force, etc.) during the movement, which will cause the response speed of voice coil electric actuators to decrease.

[0003] At present, two methods are generally used to solve the above problems: table lookup and observer. The table lookup method can be used to compensate the current according to the position feedback, which can be well dealt with for simple scenes such as springs and gravity, but it is difficult to achieve a good compensation effect when there are cogging forces, magnetic springs, etc. When the current is compensated by the observer, the compensation current will have a certain lag and cannot meet the needs of fast-beat sports scenes. Summary of the invention

[0004] The object of the present invention is to provide a feedforward control method for an electric actuator, the method comprising: obtaining operating data of the electric actuator running all strokes at a preset speed; the operating data comprising position data and current data corresponding to the position data; performing data normalization on the position data and the current data corresponding to the position data; selecting any curve fitting algorithm from an algorithm library and any regression model from a model library to obtain an algorithm model combination, performing curve fitting on the normalized operating data based on the algorithm model combination to obtain a fitting curve, and repeating the selection and fitting operations to obtain fitting curves corresponding to all algorithm model combinations; calculating the scores of the fitting curves corresponding to each of the algorithm model combinations to determine the optimal algorithm model combination; calculating the feedforward current value of the electric actuator at the current position based on the optimal algorithm model combination; and determining the control current value according to the feedforward current value and the target current value.

[0005] The feedforward control method for an electric actuator provided in an embodiment of the present invention pre-sets an algorithm library and a model library. Before performing feedforward calculation on the electric actuator, a fitting curve is obtained by fitting each possible algorithm model combination, and each fitting curve is scored to select the optimal algorithm model combination. Based on the optimal algorithm model combination, the feedforward compensation can be improved, thereby eliminating the influence of the control lag of the electric actuator by means of feedforward, so that the response speed of the electric actuator is faster.

[0006] Optionally, the algorithm library includes: LM algorithm, Powell algorithm; the model library includes: polynomial regression model, Fourier regression model, sine regression model.

[0007] The embodiments of the present invention provide a preferred algorithm and model, based on which the algorithm and model can be applied to the fitting of most curve data of electric actuators, and the fitting effect is good.

[0008] Optionally, the algorithm library also includes least squares method, neural network fitting method, kernel method, spline interpolation method, and maximum likelihood method; the model library also includes: exponential model, logarithmic model, Gaussian model, power series model, and Weibull distribution model.

[0009] The embodiments of the present invention further provide more feasible algorithms and models, which can enrich the algorithm model combination, expand the optional range, facilitate the selection of the optimal algorithm model combination, and improve the fitting effect.

[0010] Optionally, the calculating of scores of fitting curves corresponding to each of the algorithm model combinations to determine the optimal algorithm model combination includes: calculating the goodness of fit a, formula complexity b, and calculation complexity c of the fitting curves corresponding to each of the algorithm model combinations; obtaining scores of each algorithm model combination according to the goodness of fit a, the formula complexity b, and the calculation complexity c to determine the optimal algorithm model combination.

[0011] The embodiment of the present invention provides a specific scoring method to obtain the optimal algorithm model combination, thereby improving the fitting effect.

[0012] Optionally, the score of the fitting curve corresponding to each algorithm model combination is calculated based on the following formula: Rating = -2ln(A)+(B+C)ln(n) Where A=1-R 2 , R 2 is the coefficient of determination of the model combination, B is the number of parameters, C is the time or number of operations required for calculation, and n is the number of samples.

[0013] The embodiment of the present invention provides a specific scoring method to obtain the optimal algorithm model combination, thereby improving the fitting effect.

[0014] Optionally, calculating the scores of the fitting curves corresponding to each of the algorithm model combinations to determine the optimal algorithm model combination includes: calculating the scores of the fitting curves corresponding to each of the algorithm model combinations according to the Akaike Information Criterion or the Bayesian Information Criterion to determine the optimal algorithm model combination.

[0015] The embodiment of the present invention provides a specific scoring method to obtain the optimal algorithm model combination, thereby improving the fitting effect.

[0016] Optionally, perform data normalization based on the following formula:

[0017] in, is the normalized position value, is the original position value, is the maximum position in the entire motion range, is the minimum position in the entire motion range, is the maximum current in the entire motion range, It is the minimum current value in the entire movement range.

[0018] The embodiment of the present invention provides a specific method for data normalization to improve the probability of successful fitting.

[0019] Optionally, the preset speed is less than 20 mm / min.

[0020] The embodiment of the present invention limits the specific range of the preset speed to ensure a stable and smooth motion process and avoid the influence of gain mismatch on the collected data.

[0021] An embodiment of the present invention provides a feedforward control system for an electric actuator, comprising a controller; the controller is used to execute any one of the methods described above.

[0022] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is read and executed by a processor, any of the above methods is implemented.

[0023] The feedforward control system for an electric actuator and the computer-readable storage medium provided in the embodiments of the present invention can achieve the same technical effects as the aforementioned feedforward control method for an electric actuator. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0025] Figure 1 A schematic flow chart of a feedforward control method for an electric actuator provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] The present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0028] The embodiment of the present invention provides a feedforward control method and system for an electric actuator, which eliminates the influence of control lag of the electric actuator by means of feedforward, thereby making the response speed of the electric actuator faster.

[0029] Figure 1 A flow chart of a feedforward control method for an electric actuator provided by an embodiment of the present invention is shown, the method comprising: S102, obtaining operation data of the electric actuator running at a preset speed for the entire stroke.

[0030] The operation data includes position data and current data corresponding to the position data.

[0031] Before the electric actuator is actually used, it is necessary to detect its data first, so as to make appropriate compensation according to the current state. For example, the electric actuator can be a voice coil motor, a rotary motor, a linear motor, etc.

[0032] When performing data detection, the electric actuator can run the entire stroke at a low speed, while recording the position and torque current information corresponding to each position. The low speed is to avoid the influence of gain mismatch on the collected data. As long as the movement process is stable and smooth, 1-20mm / min is generally used, preferably 10mm / min.

[0033] S104, normalizing the position data and the current data corresponding to the position data.

[0034] Considering that the curve fitting process may cause fitting failure due to large differences in the numerical ranges of position data and current data, it is necessary to normalize the data to increase the probability of successful fitting.

[0035] For example, the position data of existing voice coil motors is in units of encoder pulses, so the order of magnitude ranges from thousands to millions, while the current data is in units of amperes, so the order of magnitude is generally between a few tenths of an ampere and several amperes. It can be seen that the gap between the two is large, and there is a risk of fitting failure or inaccuracy. In order to minimize data loss, the above data can be normalized in this embodiment. Optionally, the normalization can be performed on the position data range ([ P max : P min ]) is proportionally reduced to the current data range ([ I max : I min ]).

[0036] Exemplarily, the present embodiment specifically adopts Min-Max normalization. Data normalization is performed based on the following formula:

[0037] in, is the normalized position value, is the original position value, is the maximum position in the entire motion range, is the minimum position in the entire motion range, is the maximum current in the entire motion range, It is the minimum current value in the entire movement range.

[0038] S106, select any curve fitting algorithm from the algorithm library and any regression model from the model library to obtain an algorithm model combination, perform curve fitting on the normalized operating data based on the algorithm model combination to obtain a fitting curve, and repeat the above selection and fitting operations to obtain fitting curves corresponding to all algorithm model combinations.

[0039] In this embodiment, an algorithm library and a model library are pre-constructed, and a plurality of curve fitting algorithms are stored in the algorithm library, and a plurality of regression models are stored in the model library. When performing feedforward correction on a certain electric actuator, the use of different algorithms and models will lead to different compensation effects. Therefore, in this embodiment, any curve fitting algorithm and any regression model are selected to obtain an algorithm model combination, and a curve fitting is performed based on the algorithm model combination to obtain a fitting curve, and then another curve fitting algorithm or another regression model is selected to obtain another algorithm model combination and a fitting curve is obtained by fitting, exhausting all possible algorithm model combinations, and fitting curves of each algorithm model combination are obtained by fitting.

[0040] Exemplarily, the algorithm library includes: LM (Levenberg-Marquardt) algorithm, Powell algorithm. The model library includes: polynomial regression model, Fourier regression model, sine sum regression model.

[0041] Specifically, the LM algorithm is an optimization algorithm for solving nonlinear least squares problems. It combines the advantages of the gradient descent method and the Newton method, and strikes a balance between the two by dynamically adjusting a damping factor. When the damping factor is very small, the algorithm is similar to the Newton method, with a larger step size and fast convergence speed. When the damping factor is very large, the algorithm is similar to the gradient descent method, with a smaller step size and good stability. The LM algorithm is particularly suitable for nonlinear least squares problems, that is, when the objective function is a nonlinear function of the parameters. The Powell algorithm, also known as the Powell conjugate direction method, is a direct search method for solving unconstrained optimization problems, especially when the objective function is non-differentiable or the derivative is difficult to calculate. The core idea of ​​the Powell algorithm is to use conjugate directions to accelerate convergence. In unconstrained optimization problems, the algorithm starts from a certain initial point and searches along a set of initial search directions to find the minimum point of the objective function.

[0042] The above two algorithms can cover the fitting of most curve data.

[0043] Specifically, although the polynomial model has high-order terms, each term is one, and the use of first-order and second-order polynomials can fit some slightly complex waveforms, which is sufficient to cope with most working conditions in actual feedforward control. Fourier series model, Fourier series expansion can fit any waveform, but in MCU (Microcontroller Unit, microcontroller unit), it cannot support high-order Fourier operations, so some accuracy can be discarded, and 1 to 4 orders can be used for complex working conditions. Although the sine and cosine operations consume a lot of resources in the MCU, the calculation process can be greatly optimized by looking up the table. In motor control, there are sine and cosine table lookup operations, so no additional resources are consumed. The sine sum model can be used as a simplified version of the Fourier series model. Under the same amount of calculation, this model can achieve higher-order curve fitting.

[0044] In summary, in actual production, the use of polynomial model, Fourier series model, and sine sum model can cope with almost all working conditions in actual production.

[0045] In order to speed up the calculation process, some data functions in the model can be processed by table lookup, such as sin, cos, tan, etc.

[0046] S108, calculating the scores of the fitting curves corresponding to each algorithm model combination, and determining the optimal algorithm model combination.

[0047] In this embodiment, the fitting curves of all possible algorithm model combinations are scored to select the combination with the best effect.

[0048] Optionally, the score for each fitted curve is calculated based on: First, the goodness of fit a, formula complexity b, and computational complexity c of the fitting curve corresponding to each algorithm model combination are calculated. In this embodiment, the fitting curve is evaluated from three aspects: goodness of fit, formula complexity, and computational complexity, and the weight values ​​of the three can be set based on expert experience. Then, the scores of each algorithm model combination are calculated based on the above goodness of fit a, formula complexity b, and computational complexity c to determine the optimal algorithm model combination.

[0049] For example, the scores of the fitting curves corresponding to each algorithm model combination are calculated based on the following formula: Rating = -2ln(A)+(B+C)ln(n) Where A=1-R 2 , R 2 is the coefficient of determination of the model combination, B is the number of parameters, C is the time or number of operations required for calculation, and n is the number of samples.

[0050] For example, the weights are set according to expert experience as shown in Table 1.

[0051]

[0052] Table 1 Optionally, the scores of the fitting curves corresponding to each algorithm model combination are calculated according to the Akaike information criterion (AIC) or the Bayesian Information Criterion (BIC) to determine the optimal algorithm model combination.

[0053] In this embodiment, the AIC or BIC of each algorithm model combination will be calculated. These criteria penalize model complexity, thereby helping to select a model that fits the data but does not overfit. Specifically, the AIC or BIC values ​​of each model can be compared, and the model with the lowest AIC or BIC value is considered to be the optimal model.

[0054] S110, calculating the feedforward current value of the electric actuator at the current position based on the above optimal algorithm model combination.

[0055] Among them, the feedforward current value can be calculated in real time according to the current position and the corresponding parameters of the optimal algorithm model combination.

[0056] S112, determining a control current value according to the feedforward current value and the target current value.

[0057] After the feedforward current value is calculated, the feedforward value can be added to the target value and input into the current loop regulator of the motor.

[0058] The feedforward control method for an electric actuator provided in an embodiment of the present invention pre-sets an algorithm library and a model library. Before performing feedforward calculation on the electric actuator, a fitting curve is obtained by fitting each possible algorithm model combination, and each fitting curve is scored to select the optimal algorithm model combination. Based on the optimal algorithm model combination, the feedforward compensation can be improved, thereby eliminating the influence of the control lag of the electric actuator by means of feedforward, so that the response speed of the electric actuator is faster.

[0059] Optionally, the data normalization may also be performed in the following manners, for example: Z-Score normalization, nonlinear normalization method.

[0060] Optionally, the above curve fitting algorithm can also adopt: least squares method, neural network fitting method, kernel method, spline interpolation method, maximum likelihood method; the above regression model can also adopt: exponential model, logarithmic model, Gaussian model, power series model, Weibull distribution model.

[0061] An embodiment of the present invention further provides a feedforward control system for an electric actuator, including a controller; the controller can implement the above-mentioned feedforward control method for an electric actuator.

[0062] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0063] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the control device through a computer, and the program can be stored in a computer-readable storage medium. When the program is executed, it may include the processes of the above-mentioned method embodiments, wherein the storage medium may be a memory, a disk, an optical disk, etc.

[0064] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.

[0065] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprises a..." do not exclude the presence of other identical elements in the process, method, article or device including the elements. The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0066] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A feedforward control method for an electric actuator, characterized in that: The method comprises: Acquire operation data of the electric actuator running all strokes at a preset speed; the operation data includes position data and current data corresponding to the position data; performing data normalization on the position data and the current data corresponding to the position data; Select any curve fitting algorithm from the algorithm library, select any regression model from the model library to obtain an algorithm model combination, perform curve fitting on the normalized operating data based on the algorithm model combination to obtain a fitting curve, and repeat the selection and fitting operations to obtain fitting curves corresponding to all algorithm model combinations; Calculate the scores of the fitting curves corresponding to the algorithm model combinations to determine the optimal algorithm model combination; Calculating a feedforward current value of the electric actuator at a current position based on the optimal algorithm model combination; A control current value is determined according to the feedforward current value and the target current value.

2. The method according to claim 1, characterized in that The algorithm library includes: LM algorithm and Powell algorithm; The model library includes: polynomial regression model, Fourier regression model, sine regression model.

3. The method according to claim 1, characterized in that The algorithm library also includes least square method, neural network fitting method, kernel method, spline interpolation method, and maximum likelihood method; The model library also includes: exponential model, logarithmic model, Gaussian model, power series model, and Weibull distribution model.

4. The method according to claim 1, characterized in that: The step of calculating the scores of the fitting curves corresponding to the algorithm model combinations to determine the optimal algorithm model combination includes: Calculate the goodness of fit a, formula complexity b, and calculation complexity c of the fitting curve corresponding to each algorithm model combination; The scores of each algorithm model combination are calculated according to the goodness of fit a, the formula complexity b, and the calculation complexity c, and the optimal algorithm model combination is determined.

5. The method according to claim 1, characterized in that The score of the fitting curve corresponding to each algorithm model combination is calculated based on the following formula: Rating = -2ln(A)+(B+C)ln(n) Where A=1-R 2 , R 2 is the coefficient of determination of the model combination, B is the number of parameters, C is the time or number of operations required for calculation, and n is the number of samples.

6. The method according to claim 1, characterized in that The step of calculating the scores of the fitting curves corresponding to the algorithm model combinations to determine the optimal algorithm model combination includes: The scores of the fitting curves corresponding to the algorithm model combinations are calculated according to the Akaike information criterion or the Bayesian information criterion to determine the optimal algorithm model combination.

7. The method according to claim 1, characterized in that Data normalization is performed based on the following formula: in, is the normalized position value, is the original position value, is the maximum position in the entire motion range, is the minimum position in the entire motion range, is the maximum current in the entire motion range, It is the minimum current value in the entire movement range.

8. The method according to any one of claims 1 to 7, characterized in that: The preset speed is less than 20 mm / min.

9. A feedforward control system for an electric actuator, characterized in that: The method comprises a controller; the controller is used to execute the method according to any one of claims 1 to 8.

Images