A method for controlling straightness based on a magnetoelectric transducer

Abstract

The present application relates to the technical field of precision machinery control, in particular to a kind of based on magnetoelectric transducer and straightness control method, comprising: obtaining original magnetic induction data by magnetoelectric transducer, three-dimensional magnetic field distribution is obtained by electromagnetic field space mapping and axial gradient vector field is extracted.Based on this, virtual magnetic streamline model is constructed to identify local disturbance area of magnetic field distortion exceeding threshold value.The data of this area is subjected to multi-scale time-frequency analysis, and high-frequency fluctuation caused by mechanical vibration and low-frequency drift component caused by thermal deformation are separated.Two components are input into cross-coupling compensation decision network to generate high-frequency suppression instruction and low-frequency compensation trajectory respectively, and the combined comprehensive driving signal is used to realize compensation.The method can realize accurate positioning of straightness disturbance source, and decouple control of errors of different physical sources, thereby improving control accuracy and stability under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of precision mechanical control technology, and in particular to a method based on a magnetoelectric converter and linearity control. Background Technology

[0002] In the fields of precision manufacturing and measurement, straightness control is crucial for ensuring the accuracy of motion axis systems. Current technologies primarily rely on geometric sensors such as laser interferometers and optical scales to directly measure displacement deviations, or on single magnetic gratings or inductive synchronizers to acquire position information. These methods compensate for and correct the overall measured deviations through a unified feedback control loop. However, under high-precision, long-stroke conditions, the straightness error of a mechanical system is the result of the combined effects of multiple physical factors.

[0003] Traditional solutions have drawbacks. Methods based on endpoint or finite-point measurements struggle to accurately locate localized, minute defects or disturbances on guide rails or load-bearing components, resulting in compensation actions acting on the entire system, leading to low efficiency and potential introduction of new errors. Existing control strategies typically treat detected error signals as a whole, employing uniform filtering algorithms or PID control. This approach fails to distinguish the dynamic characteristics of different physical sources within the error signal, resulting in a lack of specificity in the compensation strategy. Consequently, when faced with complex disturbances, it is difficult to further improve system response performance and steady-state accuracy.

[0004] There is a need for a technical solution that can accurately locate the disturbance source affecting straightness in space and separate and compensate the error signal based on the physical mechanism of the disturbance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a method based on a magnetoelectric converter and straightness control.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method based on a magnetoelectric converter and straightness control, comprising:

[0007] The raw magnetic induction data stream related to the target straightness is obtained through a magneto-electric converter;

[0008] The original magnetic induction data stream is input into a preset electromagnetic field space mapping model to calculate a three-dimensional magnetic field intensity distribution matrix. An axial gradient vector field reflecting the axis offset characteristics is extracted from the three-dimensional magnetic field intensity distribution matrix.

[0009] A virtual magnetic streamline model is constructed based on the axial gradient vector field, and a local disturbance region defined by magnetic field distortion exceeding a preset distortion threshold is identified in the virtual magnetic streamline model.

[0010] Multi-scale time-frequency feature extraction is performed on the original magnetic induction data stream corresponding to the local disturbance region to separate the fast fluctuation component caused by mechanical vibration and the slow drift component caused by thermal deformation.

[0011] The fast fluctuation component and the slow drift component are treated as independent variables and input into a compensation decision network based on cross-coupling. The compensation decision network outputs a high-frequency suppression command for the fast fluctuation component and a low-frequency compensation trajectory for the slow drift component, respectively.

[0012] The high-frequency suppression command and the low-frequency compensation trajectory are combined to generate a comprehensive drive signal.

[0013] As a further aspect of the present invention, the step of acquiring the original magnetic induction data stream related to the target straightness through a magneto-electric converter includes:

[0014] Multiple magnetoelectric converter array units are arranged at equal intervals on the straight axis to be controlled;

[0015] Simultaneously acquire the analog voltage signal output from each of the magneto-electric converter array units;

[0016] Each analog voltage signal is simultaneously sampled and held and converted from analog to digital to obtain the corresponding discrete voltage value sequence;

[0017] The discrete voltage value sequences of all magnetoelectric converter array units are arranged and combined according to the array spatial position to form a two-dimensional data matrix with time as the first dimension and spatial position as the second dimension, which serves as the original magnetic induction data stream.

[0018] As a further aspect of the present invention, the step of inputting the original magnetic induction data stream into a preset electromagnetic field spatial mapping model to calculate the three-dimensional magnetic field intensity distribution matrix includes:

[0019] Based on the arrangement geometric model and calibration conversion coefficients of the magneto-electric converter, a mapping function from the two-dimensional data matrix to the magnetic field strength at a spatial point is established. The electromagnetic field spatial mapping model is the mapping function.

[0020] Substitute each discrete voltage value sequence in the two-dimensional data matrix into the mapping function to calculate the three-dimensional magnetic field strength component of the corresponding spatial point at the corresponding time.

[0021] By integrating the three-dimensional magnetic field intensity components of all spatial points at all times, a three-dimensional matrix is ​​constructed with the spatial point index as the first dimension, the time index as the second dimension, and the magnetic field component direction as the third dimension, which serves as the three-dimensional magnetic field intensity distribution matrix.

[0022] As a further aspect of the present invention, extracting the axial gradient vector field reflecting the axis offset characteristics from the three-dimensional magnetic field intensity distribution matrix includes:

[0023] In the three-dimensional magnetic field intensity distribution matrix, a series of spatial points are selected as the axial observation chain along the axial direction of the preset reference line.

[0024] Calculate the difference in magnetic field intensity vector between adjacent spatial points on the axial observation chain at each time step to obtain the axial gradient vector at each observation location;

[0025] Arrange the axial gradient vectors at all observation locations in spatial and temporal order to form the axial gradient vector field.

[0026] The calculation of the magnetic field strength vector difference between adjacent spatial points on the axial observation chain at each time step yields the axial gradient vector at each observation location, including:

[0027] For each spatial point on the axial observation chain, determine its next adjacent spatial point on the axial direction of the reference line;

[0028] At each sampling time, extract the three-dimensional magnetic field intensity vector of the current spatial point and the three-dimensional magnetic field intensity vector of the adjacent spatial points;

[0029] Calculate the vector difference between the three-dimensional magnetic field strength vector of the adjacent spatial point and the three-dimensional magnetic field strength vector of the current spatial point;

[0030] The vector difference is normalized to eliminate the magnitude difference caused by the spacing between different spatial points;

[0031] The normalized vector difference is used as the axial gradient vector of the current spatial point at the current time.

[0032] Store the axial gradient vectors of all spatial points at all times, in order of spatial point order and temporal order.

[0033] As a further aspect of the present invention, the step of constructing a virtual magnetic streamline model based on the axial gradient vector field includes:

[0034] Within the spatial range defined by the axial gradient vector field, multiple starting points are selected at preset intervals.

[0035] Starting from each starting point, integral tracking is performed along the direction of the gradient vector axially at the starting point to generate a continuous trajectory line;

[0036] The set of all trajectory lines constitutes the virtual magnetic streamline model.

[0037] As a further aspect of the present invention, the virtual magnetic streamline model identifies local disturbance regions defined by magnetic field distortion exceeding a preset distortion threshold, including:

[0038] Calculate the local curvature of each trajectory line in the virtual magnetic streamline model;

[0039] The local curvature is compared with a preset curvature threshold, and trajectory segments with local curvature greater than the curvature threshold are selected.

[0040] The spatial ranges occupied by all the selected trajectory segments are merged to form the local disturbance region;

[0041] The calculation of the local curvature of each trajectory line in the virtual magnetic streamline model includes:

[0042] For each trajectory line in the virtual magnetic streamline model, three consecutive sampling points on the trajectory line are selected using a sliding window method;

[0043] Calculate the forward tangent vector and the backward tangent vector at the intermediate sampling point, wherein the forward tangent vector points from the intermediate sampling point to the next sampling point, and the backward tangent vector points from the previous sampling point to the intermediate sampling point;

[0044] The forward and backward tangent vectors are averaged to obtain the average tangent vector at the intermediate sampling point;

[0045] Calculate the cosine of the angle between the forward tangent vector and the backward tangent vector, and determine the curvature angle of the trajectory line at the intermediate sampling point based on the cosine of the angle.

[0046] By combining the spatial distance between adjacent sampling points, the local curvature value of the trajectory line at the intermediate sampling point is obtained through the curvature calculation formula;

[0047] Move the sliding window along the trajectory line and repeat the steps to obtain a sequence of local curvature values ​​at all sampling points on the trajectory line.

[0048] As a further aspect of the present invention, multi-scale time-frequency feature extraction is performed on the original magnetic induction data stream corresponding to the local disturbance region, including:

[0049] Extract segments of the original magnetic induction data stream corresponding to the local disturbance region to obtain the disturbance data sequence;

[0050] The perturbation data sequence is decomposed by wavelet packet transform to obtain a set of sub-band signals;

[0051] Based on the center frequency of the sub-band signal, identify the set of sub-band signals with frequencies higher than the frequency segmentation threshold, and reconstruct them into fast fluctuation components.

[0052] The set of sub-band signals with frequencies below the frequency segmentation threshold is identified and reconstructed into slowly drifting components.

[0053] As a further aspect of the present invention, the step of merging the high-frequency suppression command and the low-frequency compensation trajectory to generate a comprehensive driving signal includes:

[0054] The high-frequency suppression command is converted from digital to analog to obtain an analog high-frequency compensation signal;

[0055] The low-frequency compensation trajectory is subjected to cubic spline interpolation to generate a continuous low-frequency compensation signal.

[0056] The analog high-frequency compensation signal is superimposed with the continuous low-frequency compensation signal in the time domain to obtain the analog comprehensive driving signal;

[0057] The step of performing digital-to-analog conversion on the high-frequency suppression command to obtain an analog high-frequency compensation signal includes:

[0058] The high-frequency suppression command is subjected to sampling rate boosting processing to increase its sampling frequency to the target output frequency;

[0059] The high-frequency suppression command after the sampling rate is increased is subjected to anti-aliasing filtering through a multi-stage digital filter to eliminate high-frequency noise interference.

[0060] A high-precision digital-to-analog converter is invoked to perform digital-to-analog conversion on the filtered high-frequency suppression command to generate an initial analog signal.

[0061] The initial analog signal is subjected to analog filtering to remove harmonic components introduced during the conversion process, resulting in an analog high-frequency compensation signal.

[0062] As a further aspect of the present invention, the method further includes:

[0063] The integrated drive signal is converted into a pulse sequence to drive the straightness actuator to generate a corresponding corrective displacement.

[0064] Real-time acquisition of updated magnetic induction data stream after correction of displacement action fed back by the magnetoelectric converter;

[0065] The updated magnetic induction data stream is spatially correlated with the preset target straightness magnetic field template. Based on the correlation calculation results, the parameters of the electromagnetic field spatial mapping model and the cross-coupled compensation decision network are dynamically adjusted.

[0066] The process of converting the integrated drive signal into a pulse sequence to drive the straightness actuator to generate a corresponding corrective displacement includes:

[0067] The simulated composite drive signal is pulse-width modulated to generate a pulse sequence;

[0068] The electromagnetic actuator in the linearity actuator is driven by a power amplifier circuit.

[0069] The electromagnetic actuator generates a corresponding axial displacement output based on the change in the duty cycle of the pulse sequence, which acts as a corrective displacement on the straight axis to be controlled.

[0070] As a further aspect of the present invention, the step of comparing the spatial correlation between the updated magnetic induction data stream and the preset target straightness magnetic field template, and dynamically adjusting the parameters of the electromagnetic field spatial mapping model and the cross-coupled compensation decision network based on the correlation calculation results, includes:

[0071] The updated magnetic induction data stream is used to calculate the updated three-dimensional magnetic field intensity distribution matrix through the electromagnetic field space mapping model;

[0072] Calculate the spatial correlation coefficient between the updated three-dimensional magnetic field intensity distribution matrix and the preset target straightness magnetic field template;

[0073] If the spatial correlation coefficient is lower than the adaptive adjustment threshold, the mapping function parameters in the electromagnetic field spatial mapping model are updated by gradient based on the updated three-dimensional magnetic field intensity distribution matrix and the change in the axial gradient vector field, and the internal weight matrix of the cross-coupled compensation decision network is adjusted synchronously.

[0074] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0075] By constructing a virtual magnetic streamline model and identifying local disturbance regions based on a preset magnetic field distortion threshold, this method achieves a shift from overall magnetic field measurement to spatial location of local defects. It does not rely on external geometric sensors but utilizes the natural rheological properties of the magnetic field morphology in space to intuitively reveal the microscopic location of axis offset. Compared to conventional methods that only determine whether the magnetic field strength exceeds limits, this technology can accurately delineate the physical region where the magnetic field distortion occurs, shifting the control system's focus from the "amount of deviation across the entire axis" to the "specific defect location causing the deviation." This allows for precise allocation of subsequent compensation resources, avoiding error propagation and control lag caused by overall averaging or compensation.

[0076] Multi-scale time-frequency feature extraction is performed on the identified local disturbance data, and the signal is separated into two independent components—fast fluctuation and slow drift—based on the physical source. Then, independent suppression commands and compensation trajectories are generated through a cross-coupled compensation decision network. This process achieves decoupling of the error signal in the frequency domain and its physical source. Conventional unified processing methods often compromise between control bandwidth and response speed, while this scheme allows for the design of suppression strategies with fast response characteristics for high-frequency mechanical vibrations, while simultaneously planning smooth and forward-looking compensation paths for low-frequency thermal deformation. The two strategies are generated independently and then merged, effectively avoiding mutual interference caused by a single compensation loop when handling complex disturbances. This enables the system to simultaneously provide optimal responses to both fast and slow-changing errors, improving overall control accuracy and dynamic stability under complex operating conditions. Attached Figure Description

[0077] Figure 1This is a flowchart of the linearity control method based on a magnetoelectric converter described in this invention;

[0078] Figure 2 A flowchart for calculating the three-dimensional magnetic field intensity distribution matrix;

[0079] Figure 3 This is a virtual magnetic streamline model based on the axial gradient vector field and the local curvature space distribution;

[0080] Figure 4 This is a heatmap showing the spatiotemporal distribution of the original magnetic induction data stream.

[0081] Figure 5 The results are from the multi-scale time-frequency feature extraction. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0083] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0084] See Figure 1A method based on a magnetoelectric converter and straightness control is proposed, the overall implementation of which is as follows: The original magnetic induction data stream related to the target straightness is acquired through a magnetoelectric converter; this original magnetic induction data stream is input into a preset electromagnetic field spatial mapping model to calculate a three-dimensional magnetic field intensity distribution matrix; an axial gradient vector field reflecting the axis offset characteristics is extracted from this three-dimensional magnetic field intensity distribution matrix; a virtual magnetic streamline model is constructed based on the axial gradient vector field, and local disturbance regions defined by magnetic field distortion exceeding a preset distortion threshold are identified in this virtual magnetic streamline model; multi-scale time-frequency feature extraction is performed on the original magnetic induction data stream corresponding to the local disturbance region to separate the fast fluctuation component caused by mechanical vibration and the slow drift component caused by thermal deformation; the fast fluctuation component and the slow drift component are used as independent variables and input into a compensation decision network based on cross-coupling; this compensation decision network outputs a high-frequency suppression command for the fast fluctuation component and a low-frequency compensation trajectory for the slow drift component, respectively; the high-frequency suppression command and the low-frequency compensation trajectory are merged to generate a comprehensive driving signal.

[0085] In one embodiment of the present invention, see [reference] Figure 2 In practical implementation, taking the control of a precision linear motion platform as an example, the platform's full-stroke axis is set as the linear axis to be controlled. Along this axis, sixteen magnetoelectric converter array units are arranged at fixed intervals L along the platform base. Each magnetoelectric converter array unit contains a set of triaxial Hall sensors. At the start of the control cycle, the central controller sends a synchronous trigger signal to synchronously acquire the three analog voltage signals output by each magnetoelectric converter array unit. These analog voltage signals correspond to the magnetic field induction components in three orthogonal directions in space. In practice, the analog voltage signal of each channel is sent to an independent signal conditioning circuit, which includes a sample-and-hold circuit and an analog-to-digital converter. Under the control of a synchronous clock, each analog voltage signal undergoes synchronous sampling and holding and analog-to-digital conversion to obtain a series of discrete voltage values. For a sampling moment... Each magnetoelectric converter array unit outputs a vector containing three voltage values, denoted as... subscript Representing the The spatial location index of each array unit, and all sixteen magnetoelectric converter array units in The discrete voltage value vector at each moment is arranged according to the spatial position of the array elements from the start to the end, forming a row vector with forty-eight elements. The row vectors formed at each sampling time point are stacked vertically in chronological order, ultimately forming a vector with time index as the row and spatial location and channel number as the column. A two-dimensional data matrix with 48 rows and 48 columns; this two-dimensional data matrix is ​​the original magnetic induction data stream.

[0086] In some embodiments, a mapping function from a two-dimensional data matrix to the magnetic field strength at a spatial point is established based on the arrangement geometry model of the magnetoelectric converter and the calibration conversion coefficients. The arrangement geometry model predefines the three-dimensional coordinates of the center point of each magnetoelectric converter array unit. The calibration conversion coefficients are obtained through offline calibration experiments and are a coefficient matrix that maps the sensor output voltage to the actual magnetic field strength. For the first... The array elements have a calibration transformation coefficient matrix denoted as . The electromagnetic field spatial mapping model is a mapping function defined by the geometric positions and calibration coefficients of all array elements. Mapping function The expression is:

[0087]

[0088] in: Indicates spatial location place, time The three-dimensional magnetic field strength vector, It is the first One magnetoelectric converter array unit Calibrate the transformation coefficient matrix. It is the first Each array element at time... The measured discrete voltage vector depends on the sensitivity of the magnetoelectric transducer used, the actual magnetic field strength of the environment, and the gain of the signal conditioning circuit. It is the bias vector of the environmental background magnetic field.

[0089] In some embodiments, each discrete voltage value vector in the two-dimensional data matrix Substitute into the mapping function Calculate the corresponding spatial points At the corresponding time Three-dimensional magnetic field strength components For all sixteen spatial locations, in all At each sampling time, this calculation process is repeated to integrate the three-dimensional magnetic field strength components of all spatial points at all times, constructing a three-dimensional matrix. 3D matrix First dimension index Corresponding spatial point location, second-dimensional index The third-dimensional index corresponds to the sampling time. The direction of the corresponding magnetic field component, three-dimensional matrix elements The value is The Each component of this three-dimensional matrix This is the calculated three-dimensional magnetic field intensity distribution matrix. It can be understood that the above mapping process converts the voltage observation data of the sensor network into a digital field model describing the evolution of the magnetic field across the entire axis of space over time. It can also be understood that synchronous sampling and analog-to-digital conversion ensure data consistency across the time dimension. Optionally, a fixed interval... Adjustments can be made based on the platform's accuracy requirements, such as in ultra-high precision applications. It can be set to 10 mm. Optionally, the number of bits for analog-to-digital conversion can be selected to 16 bits or higher precision to meet dynamic range requirements.

[0090] In one embodiment of the present invention, specifically in a practical implementation, the three-dimensional magnetic field intensity distribution matrix is ​​calculated based on the aforementioned original magnetic induction data stream synchronously acquired by the magnetoelectric converter array and obtained through the electromagnetic field spatial mapping model. For the straightness control of a linear motor platform, subsequent processing is performed, including the three-dimensional magnetic field intensity distribution matrix. It contains information on the evolution of the magnetic field at sixteen spatial points over a period of time, in a three-dimensional magnetic field intensity distribution matrix. In this process, along the axial direction of a preset reference line, a series of spatial points located on the center line of the platform's motion axis are selected as the axial observation chain. This axial observation chain contains eight spatial points, and their indices are as follows: The difference in magnetic field strength vector between adjacent spatial points on the axial observation chain at each time step is calculated to obtain the axial gradient vector at each observation location.

[0091] In practice, the difference in magnetic field intensity vector between adjacent spatial points on the axial observation chain at each moment is calculated, and the axial gradient vector at each observation position is obtained. The process is as follows: For each spatial point on the axial observation chain... Determine its next spatial point adjacent to the reference line axis. Because the spatial point is in the matrix The data is stored sequentially at each sampling time. From the three-dimensional magnetic field intensity distribution matrix Extract the current spatial point Three-dimensional magnetic field strength vector and adjacent spatial points Three-dimensional magnetic field strength vector Calculate the vector difference between the three-dimensional magnetic field strength vector of the adjacent spatial point and the three-dimensional magnetic field strength vector of the current spatial point. .

[0092] In some embodiments, the vector difference Normalization is performed to eliminate the magnitude differences caused by different spatial point spacings. and The actual physical distance between them is Normalization is performed by dividing the vector difference by this distance, resulting in the normalized axial gradient vector. The calculation formula is as follows:

[0093]

[0094] in: Indicates a point in space place, time The axial gradient vector, and These are the three-dimensional magnetic field strength vectors of two adjacent points. The spacing of the magnetoelectric converter array units needs to be determined by a trade-off between spatial sampling density, sensor cost, installation complexity, and overall system performance requirements, based on the specific accuracy specifications and engineering constraints of the target platform. The normalized vector difference is then used. As the current spatial point At the present moment The axial gradient vector, in spatial point order and temporal order, stores all seven observation locations in all... The axial gradient vectors at each sampling time point eventually form a three-dimensional tensor structure, which is the axial gradient vector field. Its dimensions correspond to the observation position index, time index, and vector components, respectively.

[0095] In some embodiments, a virtual magnetic streamline model is constructed based on the axial gradient vector field. Within the spatial range defined by the axial gradient vector field, multiple starting points are selected at preset intervals, for example, uniformly selected on multiple cross-sections within the platform's motion range. Starting from each starting point... The process begins with integral tracking along the direction of the axial gradient vector at the starting point. This tracking is performed on a discrete time-space grid. The fourth-order Runge-Kutta method is used to solve the streamline differential equations, generating a continuous trajectory. The set of trajectories generated from all starting points constitutes the virtual magnetic streamline model, which is represented in computer memory as a set of spatial curves. The axial gradient vector reflects the rate of change of the magnetic field along the axis and is a sensitive indicator for identifying geometric deviations of the axis. The virtual magnetic streamline model visually represents the virtual force line distribution derived from measured magnetic field data, providing a spatial framework for identifying local distortions.

[0096] Optionally, normalization can also be performed using the ratio of the magnitude of the vector difference to the spacing for scalar normalization, and then expressed in combination with a unit vector of direction. Optionally, the density of the starting points can be adjusted according to the required streamline model resolution; a denser distribution of starting points can be used in regions with drastic magnetic field changes.

[0097] In one embodiment of the present invention, in a specific implementation, based on the aforementioned virtual magnetic streamline model constructed by extracting the axial gradient vector field from the three-dimensional magnetic field intensity distribution matrix and constructing it according to the vector field, local disturbance regions are identified in the control scenario of a precision linear motion platform. The virtual magnetic streamline model includes multiple trajectory lines generated by integral tracing from a series of starting points. Each trajectory line is represented by a sequence of spatial coordinate points connected in sequence. The process of identifying local disturbance regions defined by magnetic field distortion exceeding a preset distortion threshold in the virtual magnetic streamline model includes calculating the local curvature of each trajectory line in the virtual magnetic streamline model.

[0098] In practical implementation, the local curvature of each trajectory line in the virtual magnetic streamline model is calculated as follows: For each trajectory line in the virtual magnetic streamline model, a sliding window containing three consecutive sampling points is used. The sliding window moves point by point along the point sequence of the trajectory line. The three points within the window are denoted as follows: , and ,in These are intermediate sampling points; calculate intermediate sampling points. Forward and backward tangent vectors at the given location, forward tangent vector From intermediate sampling points Point to the next sampling point ,Right now Backward tangent vector From the previous sampling point Pointing to the middle sampling point That is For the forward tangent vector and back tangent vector Average the data to obtain intermediate sampling points. The average tangent vector at the point Calculate the forward tangent vector. With the backward tangent vector cosine value of the angle between The trajectory line is determined at the intermediate sampling point based on the cosine value of the included angle. The bending angle at a point, and the cosine of the included angle, are calculated using the vector dot product formula: .

[0099] In some embodiments, the trajectory line at the intermediate sampling point is obtained by combining the spatial distance between adjacent sampling points and using the curvature calculation formula. Local curvature value at The calculation formula used is:

[0100]

[0101] in: Indicates the trajectory line at point The local curvature value at that point, It is the forward tangent vector With the backward tangent vector The angle between them and These are the magnitudes of the two tangent vectors. Moving the sliding window along the trajectory line, the above calculation process is repeated for every sampling point on the trajectory line except for the beginning and end points, resulting in a sequence of local curvature values ​​at all sampling points on the trajectory line. In some embodiments, the local curvature is compared with a preset curvature threshold, and trajectory segments with local curvature greater than the curvature threshold are selected. The preset curvature threshold is denoted as... For a sequence of curvature values ​​calculated along a trajectory, multiple consecutive curvature values ​​exceeding... The trajectory segments formed by adjacent sampling points are marked, and the spatial range occupied by all the selected trajectory segments is merged to form a local disturbance region. The merging operation is achieved by calculating the bounding box of the coordinates of the sampling points in all the marked trajectory segments, that is, finding the minimum and maximum values ​​of the coordinate components in all points, and defining a cubic spatial range as the local disturbance region.

[0102] It is understandable that local curvature quantifies the rate of change of the local direction of virtual magnetic streamlines, making it an effective measure for detecting magnetic field distortion regions. It is also understandable that by merging the spatial range defined by high-curvature trajectory segments, discrete anomalies can be aggregated into continuous, spatially meaningful perturbation regions. Optionally, a curvature threshold... Offline calibration or online adaptive adjustment can be performed depending on the required straightness accuracy. Optionally, a three-dimensional convex hull algorithm can be used when merging spatial ranges to wrap the set of marked points and more accurately define the geometry of the disturbed region.

[0103] See Figure 3In the straightness control method based on magnetoelectric converters, the visualization of the virtual magnetic streamline model and local curvature distribution is the core step in spatially quantifying the axis offset characteristics and magnetic field distortion regions. Specifically, the raw magnetic induction data is converted into a three-dimensional magnetic field intensity distribution matrix through an electromagnetic field spatial mapping model. Then, the axial gradient vector field is extracted from this matrix, and based on this, an integral is used to generate a virtual magnetic streamline model composed of multiple spatial trajectory lines. In the figure, the dark blue dot set represents the magnetoelectric converter array units arranged at equal intervals along the reference straight axis. The colored dots and connecting lines constitute the virtual magnetic streamline trajectory obtained by integrating the axial gradient vector field. The color mapping corresponds to the color scale on the right, intuitively reflecting the local curvature values ​​at each trajectory point: the higher the curvature value, the closer the color is to red, indicating a more significant degree of magnetic field distortion and a more prominent axis offset characteristic in that region. In the process of identifying local disturbance regions, the local curvature of each virtual magnetic flux line is calculated using a sliding window with three consecutive sampling points: the curvature value sequence of each trajectory point is obtained by combining the angle and magnitude of the forward and backward tangent vectors with the spacing between adjacent sampling points; then, the curvature values ​​are compared with a preset threshold to filter out trajectory segments with excessive curvature, and their spatial ranges are merged to form local disturbance regions. In the figure, clusters of high curvature (red, orange, and yellow) points are concentrated in the central spatial region, which corresponds to the magnetic field distortion disturbance region caused by mechanical vibration or thermal deformation during the straightness control process, providing spatial positioning basis for subsequent multi-scale time-frequency feature extraction and compensation decision-making.

[0104] In one embodiment of the present invention, in a specific implementation, for the local disturbance region identified by calculating the local curvature of the virtual magnetic streamline model and comparing it with a preset threshold, the corresponding original magnetic induction data stream is extracted for multi-scale time-frequency feature extraction. The original magnetic induction data stream is a two-dimensional data matrix containing time and space information. The local disturbance region corresponds to a set of specific spatial location indices. The data columns of these spatial location indices at all sampling time points are extracted to obtain a two-dimensional disturbance data matrix as a disturbance data sequence.

[0105] In practice, the perturbation data sequence is decomposed by wavelet packet transform. The Daubechies4 wavelet is used as the basis function. Each spatial location data is independently decomposed into three complete wavelet packet decompositions to obtain a set of sub-band signals covering different frequency bands. Wavelet packet decomposition divides the frequency bands of the signal more finely. Each sub-band signal has a specific center frequency and bandwidth. Refer to Table 1 to show a typical wavelet packet decomposition sub-band frequency classification.

[0106] Table 1: Classification of Wavelet Packet Decomposition Subband Frequency

[0107]

[0108] In some embodiments, a set of sub-band signals with frequencies higher than a preset frequency segmentation threshold is identified based on the center frequency of the sub-band signal. The frequency segmentation threshold is set according to the dominant frequency range of the mechanical vibration of the controlled platform, for example, set to [value missing]. Hz, extract the sub-band signals with frequencies higher than 10Hz from the table, i.e., the sub-band signals corresponding to nodes (3,2) to (3,7), and merge and reconstruct them into fast oscillation components using the wavelet packet reconstruction algorithm. The mathematical expression for reconstructing the fast oscillation component is:

[0109]

[0110] in: This represents the reconstructed fast fluctuation component time series. This represents all frequencies above the frequency segmentation threshold. The sub-band index set, for example , This indicates that the wavelet packet decomposition is at the third level, the... The sub-band signal at each node has a value that depends on the strength of the original perturbation data. The weighting coefficients corresponding to this sub-band are typically initialized to 1. Frequency segments below the frequency segmentation threshold are identified. The sub-band signal sets, namely the sub-band signal sets corresponding to nodes (3,0) and (3,1), are merged and reconstructed into slowly drifting components using the wavelet packet reconstruction algorithm. .

[0111] In some embodiments, the high-frequency suppression command and the low-frequency compensation trajectory are combined to generate a comprehensive driving signal. The high-frequency suppression command is output by the compensation decision network and is a discrete-time series. The high-frequency suppression command is converted from digital to analog to obtain an analog high-frequency compensation signal. To increase the sampling rate, an interpolation filter is used to adjust the sampling frequency from the control frequency. Increase to target output frequency For example, the sampling rate can be increased from 1kHz to 10kHz. A multi-order finite impulse response digital filter is used to perform anti-aliasing filtering on the high-frequency suppression command sequence after the sampling rate is increased, eliminating high-frequency image noise interference introduced by interpolation. A high-precision digital-to-analog converter is then used to perform digital-to-analog conversion on the filtered high-frequency suppression command sequence, generating an initial stepped analog signal. The initial analog signal is then input to a low-pass analog filter with a cutoff frequency slightly higher than the control bandwidth to filter out high-frequency harmonic components introduced during the digital-to-analog conversion, obtaining the analog high-frequency compensation signal. It is understandable that wavelet packet transform can finely decompose signal energy into different time-frequency units, thereby effectively separating disturbance components from different physical sources based on frequency characteristics. It is also understandable that sampling rate enhancement and anti-aliasing filtering are key steps in ensuring that digital control commands can be converted into high-quality analog drive signals.

[0112] Optionally, the number of wavelet packet decomposition layers can be adjusted to 4 or more, depending on the highest frequency of the signal and the required analysis accuracy. Optionally, the frequency segmentation threshold... It can be designed as an adaptive parameter, dynamically adjusted according to the vibration spectrum characteristics monitored in real time.

[0113] In some embodiments, the low-frequency compensation trajectory is subjected to cubic spline interpolation to generate a continuous low-frequency compensation signal. The low-frequency compensation trajectory is output by the compensation decision network and is a series of discrete position command points at a lower time resolution. A cubic spline interpolation algorithm is used to fit these discrete points, generating a trajectory curve that is continuous in time and has a continuous second derivative. The target output frequency is then calculated from this curve. Resampling yields a continuous low-frequency compensation signal. In the time domain, the analog high-frequency compensation signal... With continuous low-frequency compensation signals The superposition is performed using an analog adder circuit to obtain a combined analog drive signal. Alternatively, non-uniform rational B-spline interpolation can be used to handle low-frequency compensated trajectory points with non-uniform time intervals.

[0114] See Figure 4In the linearity control method based on the magnetoelectric converter, the spatiotemporal distribution characteristics of the original magnetic induction data stream are the core input for subsequent disturbance analysis and compensation decisions. The figure uses time (0–5s) as the horizontal axis and spatial location (0–1m) as the vertical axis, visually presenting the dynamic distribution of the magnetic induction voltage (2–10V) through pseudo-color mapping: Periodic spatial fringe characteristics: Within the 0–5s observation window, the magnetic induction voltage exhibits a significant periodic spatial fringe structure. The fringe spacing is highly correlated with the spatial sampling interval of the magnetoelectric converter array, reflecting the inherent spatial pattern of the magnetic field distribution on the straight axis. Amplitude modulation over time: The fringe amplitude exhibits significant modulation at different times. For example, at approximately 0.6s, 2.5s, and 4.7s, voltage peaks (close to 10V) appear at local spatial locations (such as 0.5m and 0.8m), while at approximately 1.5s and 3.5s, voltage valleys (close to 2V) are observed. This amplitude modulation directly corresponds to the magnetic field distortion caused by mechanical vibration and thermal deformation during the linearity execution process. Spatiotemporal coupling disturbance characteristics: High-amplitude regions exhibit quasi-periodic repetition on the time axis and localized banded distribution in space. These regions highly coincide with the local disturbance regions subsequently identified through the virtual magnetic streamline model, providing clear spatiotemporal localization basis for multi-scale time-frequency feature extraction.

[0115] In one embodiment of the present invention, in a specific implementation, the analog integrated drive signal generated by the aforementioned merging high-frequency suppression command and low-frequency compensation trajectory is... The signal is input to a pulse width modulation module to synthesize the analog drive signal. Pulse width modulation (PWM) is performed to generate a pulse sequence. PWM uses a triangular wave as the carrier wave to synthesize the continuous analog drive signal. Compared with a high-frequency triangular wave, in the analog synthetic drive signal A high-level signal is output when the instantaneous value is greater than the instantaneous value of the triangular wave, and a low-level signal is output when the instantaneous value is less than the instantaneous value of the triangular wave, thus generating a series of pulse widths and... Rectangular pulse sequence with proportional amplitude .

[0116] In practical implementation, pulse sequence The signal is fed into a power amplifier circuit to drive the electromagnetic actuator in the linearity actuator. The power amplifier circuit uses an H-bridge topology and operates based on the pulse sequence. The high and low levels control the switching on and off of the power switch, thereby converting the low-voltage control signal into a high-voltage, high-current signal capable of driving the electromagnetic actuator coil. The electromagnetic actuator operates according to the pulse sequence. The change in duty cycle generates a corresponding axial displacement output, which acts as a corrective displacement on the straight axis to be controlled, such as a voice coil motor actuator integrated on the platform base. In some embodiments, the updated magnetic induction data stream after the corrective displacement is collected in real time from the feedback of the magneto-electric converter. At the end of each control cycle, the aforementioned process of synchronously acquiring voltage signals through the magneto-electric converter array, performing synchronous sampling and holding and analog-to-digital conversion, and arranging them according to spatial position to form a two-dimensional data matrix is ​​repeated. However, the new voltage signal output by the magneto-electric converter array unit on the platform after the corrective displacement of the current cycle is used to construct a new two-dimensional data matrix as the updated magnetic induction data stream. The updated magnetic induction data stream is compared with the spatial correlation of a preset target straightness magnetic field template, which is a standard matrix of three-dimensional magnetic field intensity distribution under ideal conditions. .

[0117] The process involves comparing the spatial correlation between the updated magnetic induction data stream and the preset target straightness magnetic field template, and dynamically adjusting the parameters based on the correlation calculation results. This includes: calculating the updated three-dimensional magnetic field intensity distribution matrix by using an electromagnetic field spatial mapping model to access the updated magnetic induction data stream. Calculate the updated three-dimensional magnetic field strength distribution matrix. With the preset target straightness magnetic field template Spatial correlation coefficient between The calculation formula is:

[0118]

[0119] Among them: Represents the spatial correlation coefficient. It is the updated three-dimensional magnetic field strength distribution matrix. In position ,time and components The element value on, It is the target straightness magnetic field template The element values ​​at corresponding locations, times, and components are related to the specific magnetic field environment and sensor calibration. and These are the average values ​​of all elements in the two matrices, respectively. This is the total number of spatial points, and its value depends on the number of magnetoelectric converter array units deployed. This is the length of the time series, and its value depends on the control period or the observation duration. If the spatial correlation coefficient... Below the adaptive adjustment threshold Then, based on the updated three-dimensional magnetic field strength distribution matrix The change in the axial gradient vector field updates the parameters of the mapping function in the electromagnetic field spatial mapping model, and simultaneously adjusts the internal weight matrix of the cross-coupled compensation decision network. It can be understood that the spatial correlation coefficient provides a quantitative assessment of the degree of matching between the current system state and the ideal target state. It can also be understood that adaptively adjusting the threshold... The settings determine the triggering conditions for system parameter updates.

[0120] Optional, adaptive threshold adjustment It can be set to a value close to 1, such as 0.95, to achieve strict control precision requirements. Optionally, the gradient update process can employ a stochastic gradient descent algorithm to update the coefficient matrix parameters in the mapping function.

[0121] See Figure 5 In the multi-scale time-frequency feature extraction stage, the core objective is to separate the fast-fluctuation component dominated by mechanical vibration and the slow-drift component dominated by thermal deformation from the original magnetic induction data stream corresponding to the local disturbance region, providing independent control variables for subsequent cross-coupling compensation decisions. As clearly seen in the figure, the fast-fluctuation component (mechanical vibration), represented by the red curve, exhibits high-frequency, small-amplitude oscillation characteristics, with its amplitude constantly changing rapidly around 0, which highly matches the instantaneous and high-frequency characteristics of mechanical vibration. Conversely, the slow-drift component (thermal deformation), represented by the green curve, shows a low-frequency, large-amplitude, slow-changing trend, with its amplitude fluctuating slowly within the range of -0.5 to 1.6, consistent with the cumulative and slowly changing characteristics of thermal deformation. This separation operation relies on wavelet packet transform decomposition of the perturbation data sequence: first, the original magnetic induction data stream segment corresponding to the local perturbation region is extracted to obtain the perturbation data sequence; then, it is decomposed into a series of sub-band signals with different center frequencies through wavelet packet transform; next, based on a preset frequency segmentation threshold, the sub-band signals with frequencies above the threshold are reconstructed into fast fluctuation components, and the sub-band signals with frequencies below the threshold are reconstructed into slow drift components. Through this multi-scale time-frequency feature extraction method, the error sources of two different physical mechanisms mixed in the original signal are effectively decoupled, enabling the subsequent compensation decision network to output precise high-frequency suppression commands and low-frequency compensation trajectories for high-frequency mechanical vibration and low-frequency thermal deformation respectively, thereby improving the overall accuracy and robustness of straightness control.

[0122] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for controlling linearity based on a magnetoelectric converter, characterized in that, include: The raw magnetic induction data stream related to the target straightness is obtained through a magneto-electric converter; The original magnetic induction data stream is input into a preset electromagnetic field space mapping model to calculate a three-dimensional magnetic field intensity distribution matrix. An axial gradient vector field reflecting the axis offset characteristics is extracted from the three-dimensional magnetic field intensity distribution matrix. A virtual magnetic streamline model is constructed based on the axial gradient vector field, and a local disturbance region defined by magnetic field distortion exceeding a preset distortion threshold is identified in the virtual magnetic streamline model. Multi-scale time-frequency feature extraction is performed on the original magnetic induction data stream corresponding to the local disturbance region to separate the fast fluctuation component caused by mechanical vibration and the slow drift component caused by thermal deformation. The fast fluctuation component and the slow drift component are treated as independent variables and input into a compensation decision network based on cross-coupling. The compensation decision network outputs a high-frequency suppression command for the fast fluctuation component and a low-frequency compensation trajectory for the slow drift component, respectively. The high-frequency suppression command is converted from digital to analog to obtain an analog high-frequency compensation signal. The low-frequency compensation trajectory is processed by cubic spline interpolation to generate a continuous low-frequency compensation signal. The analog high-frequency compensation signal and the continuous low-frequency compensation signal are superimposed in the time domain to obtain an analog comprehensive driving signal. The construction of the virtual magnetic streamline model based on the axial gradient vector field includes: Within the spatial range defined by the axial gradient vector field, multiple starting points are selected at preset intervals. Starting from each starting point, integral tracking is performed along the direction of the gradient vector axially at the starting point to generate a continuous trajectory line; The set of all trajectory lines constitutes the virtual magnetic streamline model.

2. The method for linearity control based on a magnetoelectric converter according to claim 1, characterized in that, The acquisition of the raw magnetic induction data stream related to the target straightness via a magneto-electric converter includes: Multiple magnetoelectric converter array units are arranged at equal intervals on the straight axis to be controlled; Simultaneously acquire the analog voltage signal output from each of the magneto-electric converter array units; Each analog voltage signal is simultaneously sampled and held and converted from analog to digital to obtain the corresponding discrete voltage value sequence; The discrete voltage value sequences of all magnetoelectric converter array units are arranged and combined according to the array spatial position to form a two-dimensional data matrix with time as the first dimension and spatial position as the second dimension, which serves as the original magnetic induction data stream.

3. The method for linearity control based on a magnetoelectric converter according to claim 1, characterized in that, The step of inputting the original magnetic induction data stream into a preset electromagnetic field spatial mapping model to calculate the three-dimensional magnetic field intensity distribution matrix includes: Based on the arrangement geometric model and calibration conversion coefficients of the magneto-electric converter, a mapping function from the two-dimensional data matrix to the magnetic field strength at a spatial point is established. The electromagnetic field spatial mapping model is the mapping function. Substitute each discrete voltage value sequence in the two-dimensional data matrix into the mapping function to calculate the three-dimensional magnetic field strength component of the corresponding spatial point at the corresponding time. By integrating the three-dimensional magnetic field intensity components of all spatial points at all times, a three-dimensional matrix is ​​constructed with the spatial point index as the first dimension, the time index as the second dimension, and the magnetic field component direction as the third dimension, which serves as the three-dimensional magnetic field intensity distribution matrix.

4. The method for linearity control based on a magnetoelectric converter according to claim 3, characterized in that, Extracting the axial gradient vector field reflecting the axis offset characteristics from the three-dimensional magnetic field intensity distribution matrix includes: In the three-dimensional magnetic field intensity distribution matrix, a series of spatial points are selected as the axial observation chain along the axial direction of the preset reference line. Calculate the difference in magnetic field intensity vector between adjacent spatial points on the axial observation chain at each time step to obtain the axial gradient vector at each observation location; Arrange the axial gradient vectors at all observation locations in spatial and temporal order to form the axial gradient vector field. The calculation of the magnetic field strength vector difference between adjacent spatial points on the axial observation chain at each time step yields the axial gradient vector at each observation location, including: For each spatial point on the axial observation chain, determine its next adjacent spatial point on the axial direction of the reference line; At each sampling time, extract the three-dimensional magnetic field intensity vector of the current spatial point and the three-dimensional magnetic field intensity vector of the adjacent spatial points; Calculate the vector difference between the three-dimensional magnetic field strength vector of the adjacent spatial point and the three-dimensional magnetic field strength vector of the current spatial point; The vector difference is normalized to eliminate the magnitude difference caused by the spacing between different spatial points; The normalized vector difference is used as the axial gradient vector of the current spatial point at the current time. Store the axial gradient vectors of all spatial points at all times, in order of spatial point order and temporal order.

5. The method for linearity control based on a magnetoelectric converter according to claim 1, characterized in that, In the virtual magnetic streamline model, the local disturbance region defined by magnetic field distortion exceeding a preset distortion threshold is identified, including: Calculate the local curvature of each trajectory line in the virtual magnetic streamline model; The local curvature is compared with a preset curvature threshold, and trajectory segments with local curvature greater than the curvature threshold are selected. The spatial ranges occupied by all the selected trajectory segments are merged to form the local disturbance region; The calculation of the local curvature of each trajectory line in the virtual magnetic streamline model includes: For each trajectory line in the virtual magnetic streamline model, three consecutive sampling points on the trajectory line are selected using a sliding window method; Calculate the forward tangent vector and the backward tangent vector at the intermediate sampling point, wherein the forward tangent vector points from the intermediate sampling point to the next sampling point, and the backward tangent vector points from the previous sampling point to the intermediate sampling point; The forward and backward tangent vectors are averaged to obtain the average tangent vector at the intermediate sampling point; Calculate the cosine of the angle between the forward tangent vector and the backward tangent vector, and determine the curvature angle of the trajectory line at the intermediate sampling point based on the cosine of the angle. By combining the spatial distance between adjacent sampling points, the local curvature value of the trajectory line at the intermediate sampling point is obtained through the curvature calculation formula; Move the sliding window along the trajectory line and repeat the steps to obtain a sequence of local curvature values ​​at all sampling points on the trajectory line.

6. The method for linearity control based on a magnetoelectric converter according to claim 1, characterized in that, Multi-scale time-frequency feature extraction is performed on the original magnetic induction data stream corresponding to the local disturbance region, including: Extract segments of the original magnetic induction data stream corresponding to the local disturbance region to obtain the disturbance data sequence; The perturbation data sequence is decomposed by wavelet packet transform to obtain a set of sub-band signals; Based on the center frequency of the sub-band signal, identify the set of sub-band signals with frequencies higher than the frequency segmentation threshold, and reconstruct them into fast fluctuation components. The set of sub-band signals with frequencies below the frequency segmentation threshold is identified and reconstructed into slowly drifting components.

7. The method for linearity control based on a magnetoelectric converter according to claim 1, characterized in that, The step of performing digital-to-analog conversion on the high-frequency suppression command to obtain an analog high-frequency compensation signal includes: The high-frequency suppression command is subjected to sampling rate boosting processing to increase its sampling frequency to the target output frequency; The high-frequency suppression command after the sampling rate is increased is subjected to anti-aliasing filtering through a multi-stage digital filter to eliminate high-frequency noise interference. A high-precision digital-to-analog converter is invoked to perform digital-to-analog conversion on the filtered high-frequency suppression command to generate an initial analog signal. The initial analog signal is subjected to analog filtering to remove harmonic components introduced during the conversion process, resulting in an analog high-frequency compensation signal.

8. The method for linearity control based on a magnetoelectric converter according to claim 7, characterized in that, The method further includes: The integrated drive signal is converted into a pulse sequence to drive the straightness actuator to generate a corresponding corrective displacement. Real-time acquisition of updated magnetic induction data stream after correction of displacement action fed back by the magnetoelectric converter; The updated magnetic induction data stream is spatially correlated with the preset target straightness magnetic field template, and the parameters of the electromagnetic field spatial mapping model and the cross-coupled compensation decision network are dynamically adjusted based on the correlation calculation results. The step of converting the integrated drive signal into a pulse sequence to drive the straightness actuator to generate a corresponding corrective displacement includes: The simulated composite drive signal is pulse-width modulated to generate a pulse sequence; The electromagnetic actuator in the linearity actuator is driven by a power amplifier circuit. The electromagnetic actuator generates a corresponding axial displacement output based on the change in the duty cycle of the pulse sequence, which acts as a corrective displacement on the straight axis to be controlled.

9. The method for linearity control based on a magnetoelectric converter according to claim 8, characterized in that, The step of comparing the spatial correlation between the updated magnetic induction data stream and the preset target straightness magnetic field template, and dynamically adjusting the parameters of the electromagnetic field spatial mapping model and the cross-coupled compensation decision network based on the correlation calculation results, includes: The updated magnetic induction data stream is used to calculate the updated three-dimensional magnetic field intensity distribution matrix through the electromagnetic field space mapping model; Calculate the spatial correlation coefficient between the updated three-dimensional magnetic field intensity distribution matrix and the preset target straightness magnetic field template; If the spatial correlation coefficient is lower than the adaptive adjustment threshold, the mapping function parameters in the electromagnetic field spatial mapping model are updated by gradient based on the updated three-dimensional magnetic field intensity distribution matrix and the change in the axial gradient vector field, and the internal weight matrix of the cross-coupled compensation decision network is adjusted synchronously.

Images