A voice coil motor adaptive linearization control method and system

Abstract

The application relates to the technical field of servo control, and discloses a voice coil motor adaptive linearization control method and system. The method comprises the following steps: collecting coil current and valve core displacement, extracting a compensation residual symbol value, and arranging the compensation residual symbol sequence in a sliding window; calculating a commutation symbol reversal strength index at a commutation moment to obtain a commutation direction bias diagnosis flag; when the diagnosis flag is triggered, recursively identifying parameters independently in a positive current area and a negative current area to obtain a positive force constant parameter set, a negative force constant parameter set and a hysteresis asymmetry degree index; mapping the hysteresis asymmetry degree index into a commutation disturbance compensation weight, adjusting an extended state observer gain, superimposing an inverse model compensation current and a basic current instruction to output a driving current instruction. The application improves the position tracking accuracy and control robustness of a voice coil motor servo valve at a commutation moment under a large-stroke high-frequency reciprocating working condition.

Description

Technical Field

[0001] This application relates to the field of servo control technology, and in particular to an adaptive linearization control method and system for a voice coil motor. Background Technology

[0002] Voice coil motors (VCOs), as direct-drive actuators, are widely used in the valve core actuation of electro-hydraulic servo valves. Their basic working principle involves a energized coil moving linearly within a static magnetic field generated by a permanent magnet under the influence of Ampere's force. Theoretically, the electromagnetic thrust is proportional to the coil current. To compensate for the inherent nonlinear characteristics of VCOs, current technologies typically employ recursive least squares methods to identify the electromagnetic force-current relationship online, establishing a force constant polynomial inverse model. This inverse model feedforward compensation current is then superimposed onto the position feedback control law output. Combined with an extended state observer, external hydraulic disturbances are estimated and compensated in real-time, forming an adaptive linearized control framework that coordinates three stages: online identification, inverse model feedforward, and disturbance observation.

[0003] However, existing technologies have fundamental flaws in parameter identification. Current recursive least-squares identification schemes use a unified set of force constant parameters for both positive and negative current regions, implicitly assuming that the electromagnetic force-current relationship is centrally symmetric within the positive and negative half-cycles. This assumption is acceptable when the magnetic circuit saturation is low, but under conditions of long stroke and high-frequency reciprocating motion, the hysteresis effect causes the positive and negative currents to produce different actual electromagnetic forces under the same amplitude conditions, resulting in a systematic difference in the polynomial coefficients of the force constants on both sides. When a single parameter set is used to fit data from both sides simultaneously, the identification result can only converge to the weighted average of the true parameter values ​​on both sides. This leads to a directional compensation error in the inverse model at the commutation moment, where the direction of the compensation current is opposite to the actual required compensation direction. The feedback control law cannot eliminate this error by accelerating the parameter update rate because the root cause of the error lies in the structural misuse of the parameter set rather than an insufficient update rate. Summary of the Invention

[0004] This application provides an adaptive linearization control method and system for voice coil motors, which solves the problems in existing adaptive linearization control methods for voice coil motors, such as directional compensation errors at the commutation time caused by the unified identification of a single set of parameters, and the inability of existing technologies to distinguish between two failure modes: insufficient compensation and compensation direction deflection. This improves the position tracking accuracy and control robustness of the voice coil motor servo valve at the commutation time under the long stroke high frequency reciprocating condition.

[0005] In a first aspect, this application provides an adaptive linearization control method for a voice coil motor, the adaptive linearization control method for a voice coil motor comprising: Step S1: During the adaptive control of the voice coil motor position, the coil current and valve core displacement are collected. The symbol value is extracted based on the difference between the valve core displacement and the expected output of the inverse model. The symbol values ​​are arranged in time sequence within the sliding window to obtain the compensation residual symbol sequence. Step S2: At the moment of coil current commutation, the mean forward sign and the mean backward sign before and after commutation are calculated from the compensation residual sign sequence. The product of the two is used as the commutation sign reversal intensity index and compared with a preset threshold to obtain the commutation direction bias diagnostic mark. Step S3: When the commutation direction bias diagnostic flag is triggered, the coil current is divided into a positive current region and a negative current region. In the two regions, the recursive parameter identification is performed independently with the coil current and the compensation residual to obtain the positive force constant parameter set and the negative force constant parameter set. The difference vector norm of the two is used as an index of the degree of hysteresis asymmetry. Step S4: Within the commutation window before and after the coil current commutation moment, the hysteresis asymmetry index is converted into a commutation disturbance compensation weight by piecewise linear mapping. The total disturbance estimate update gain of the extended state observer is adjusted with the commutation disturbance compensation weight. The inverse model compensation current obtained by solving the positive force constant parameter set or the negative force constant parameter set is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is applied to the coil of the voice coil motor to realize position closed-loop control.

[0006] Secondly, this application provides an adaptive linearization control system for a voice coil motor, the adaptive linearization control system for a voice coil motor comprising: The acquisition module is used to acquire coil current and valve core displacement during the adaptive control of voice coil motor position. It extracts symbol values ​​based on the difference between the valve core displacement and the expected output of the inverse model, and arranges them in time sequence within a sliding window to obtain a compensation residual symbol sequence. The comparison module is used to calculate the mean forward sign and the mean backward sign before and after the commutation from the compensation residual sign sequence at the moment of coil current commutation, and use the product of the two as the commutation sign reversal intensity index, compare it with a preset threshold, and obtain the commutation direction bias diagnostic flag. The identification module is used to divide the coil current into a positive current region and a negative current region when the commutation direction bias diagnostic flag is triggered. In the two regions, the recursive parameter identification is performed independently with the coil current and the compensation residual to obtain the positive force constant parameter set and the negative force constant parameter set. The difference vector norm of the two is used as an index of the degree of hysteresis asymmetry. The superposition module is used to convert the hysteresis asymmetry index into a commutation disturbance compensation weight by piecewise linear mapping within the commutation window before and after the commutation time of the coil current. The total disturbance estimate update gain of the extended state observer is adjusted with the commutation disturbance compensation weight. The inverse model compensation current obtained by solving the positive force constant parameter set or the negative force constant parameter set is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is applied to the coil of the voice coil motor to realize position closed-loop control.

[0007] Thirdly, a voice coil motor adaptive linearization control device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the voice coil motor adaptive linearization control device to execute the above-described voice coil motor adaptive linearization control method.

[0008] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the aforementioned adaptive linearization control method for a voice coil motor.

[0009] In the technical solution provided in this application, during the adaptive position control of a voice coil motor, the sign value of the compensation residual is extracted from the inverse model and arranged sequentially within a sliding window to form a compensation residual sign sequence. This independently separates the temporal structure information of the compensation direction from the numerical amplitude information, enabling the control system to directly observe the state of the compensation direction deviation for the first time. Based on this, the product of the average values ​​of the signs before and after the commutation moment is used to construct the commutation sign reversal intensity index, and the commutation direction offset diagnostic flag is obtained through the statistical proportion of multiple consecutive commutation moments. This diagnostic mechanism explicitly distinguishes between two fundamentally different failure modes—"insufficient compensation" and "compensation direction deflection"—from the mixed tracking error signal. This overcomes the fundamental limitation of existing technologies that rely solely on error numerical amplitude feedback and cannot identify directional compensation failures, providing a logically unique triggering condition for subsequent differentiated processing.

[0010] When the commutation direction bias diagnostic flag is triggered, the coil current is divided into a positive current region and a negative current region. In each of the two regions, the recursive parameter identification is performed independently using the coil current and the compensation residual, resulting in a positive force constant parameter set and a negative force constant parameter set. The two parameter sets are completely isolated from the cross-contamination of the identification data through a freezing mechanism, so that each of them independently converges to the true electromagnetic force nonlinearity characteristics of the corresponding current region. This fundamentally eliminates the directional compensation error caused by the positive and negative half-cycle hysteresis asymmetry effect when a single set of parameters is uniformly identified. By using a hysteresis asymmetry index composed of the difference vector norms of two sets of parameters, the hysteresis asymmetry state inside the voice coil motor is transformed into a continuously quantized signal that is updated in real time. This signal is then converted into a commutation disturbance compensation weight through piecewise linear mapping. This weight is used to dynamically adjust the total disturbance estimation update gain of the extended state observer within the commutation window. This allows the observer to respond to directional disturbances with a gain intensity that matches the hysteresis degree during the commutation shock duration and to restore the standard gain outside the commutation window to avoid noise amplification in the steady-state segment. This achieves direct data-driven operation of the observer parameters by the internal structural differences of the compensator, and solves the long-standing irreconcilable contradiction between the disturbance compensation response and steady-state noise suppression at the commutation moment in the fixed-bandwidth extended state observer. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of an embodiment of the adaptive linearization control method for voice coil motors in this application. Figure 2 This is a schematic diagram of the timing structure of the compensation residual symbol sequence in the embodiments of this application. Detailed Implementation

[0013] This application provides an adaptive linearization control method and system for a voice coil motor. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0014] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the adaptive linearization control method for voice coil motors in this application includes: Step S1: During the adaptive control of the voice coil motor position, the coil current and valve core displacement are collected. The sign value is extracted based on the difference between the valve core displacement and the expected output of the inverse model. The sign value is then arranged in time sequence within the sliding window to obtain the compensation residual sign sequence. Specifically, the compensation residual symbol sequence is a sequence of symbols obtained by extracting the positive and negative signs of the difference between the actual electromagnetic force and the expected electromagnetic force after the inverse model compensation effect on a cycle-by-cycle basis and arranging them in time sequence within a sliding window of 20 control cycles. The window length is approximately 4 ms when the control cycle is 0.2 ms, which corresponds to 20 control cycles. This time scale can cover about two commutation cycles under 5 Hz sinusoidal excitation, which is sufficient to capture the concentrated changes in the symbol structure near the commutation. This sequence separates the systematic deviation of the compensation direction from the numerical amplitude information and retains the temporal structure information of the direction deviation separately.

[0015] Step S2: At the moment of coil current commutation, the mean forward sign and the mean backward sign before and after commutation are calculated from the compensation residual sign sequence. The product of the two is used as the commutation sign reversal intensity index and compared with the preset threshold to obtain the commutation direction bias diagnostic mark. Specifically, the reversal sign inversion intensity index is obtained by multiplying the average sign value of the 5 steps before the reversal time with the average sign value of the 5 steps after the reversal time. The value ranges from negative one to positive one. When the average sign values ​​of the signs before and after the reversal are opposite and the absolute values ​​are both large, the product tends to be negative one, indicating that the compensation residual has undergone concentrated reversal before and after the reversal. The preset threshold is set to -0.6 as the single discrimination boundary. When the proportion of direction bias type reversal events reaches two-thirds within three consecutive reversal times, the reversal direction bias diagnosis flag is triggered. The three statistical windows exclude single reversal noise interference. The two-thirds proportion threshold achieves a balance between response timeliness and noise resistance.

[0016] Step S3: When the commutation direction bias diagnostic flag is triggered, the coil current is divided into a positive current region and a negative current region. In the two regions, the recursive parameter identification is performed independently with the coil current and the compensation residual to obtain the positive force constant parameter set and the negative force constant parameter set. The difference vector norm of the two is used as the index of the degree of hysteresis asymmetry. Specifically, the positive force constant parameter set and the negative force constant parameter set each contain three components. For the constant term, first term and second term coefficients of the stress constant polynomial, the two parameter sets independently perform recursive least squares identification within their respective current intervals with a forgetting factor of 0.97. The hysteresis asymmetry index is obtained by summing the squares of the differences between the corresponding components of the two parameter sets into a difference vector, and then taking the square root after weighting by weighting coefficients of 1.0, 0.5 and 0.1. The constant term has the highest weight because it has the most direct influence on the compensation direction at the moment of small current commutation. The dead zone threshold is set to 0.05A to cover the current zero-point jitter range caused by sensor noise.

[0017] Step S4: Within the commutation window before and after the coil current commutation moment, the hysteresis asymmetry index is converted into a commutation disturbance compensation weight by piecewise linear mapping. The total disturbance estimate update gain of the extended state observer is adjusted with the commutation disturbance compensation weight. The inverse model compensation current obtained by solving the positive force constant parameter set or the negative force constant parameter set is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is applied to the coil of the voice coil motor to realize position closed-loop control.

[0018] Specifically, in the piecewise linear mapping of the commutation disturbance compensation weight, the low threshold is set to 0.5 N / A and the high threshold is set to 2.0 N / A. The weight changes linearly between 1.0 and 2.5. The upper limit of 2.5 is set to prevent the excessive gain of the extended state observer from amplifying sensor noise and causing control oscillation. The commutation window range is set to 7 cycles, 3 control cycles before and after the commutation moment. The width of this window covers the duration of hysteresis impact during the commutation transition of the voice coil motor. The gain outside the window is restored to the standard value to ensure that the steady-state control performance is not affected. The inverse model compensation current and the basic current command are superimposed to form a drive current command that acts on the coil to complete the position closed-loop control.

[0019] In one specific embodiment, step S1 includes: During the adaptive control of the voice coil motor position, the coil current and valve core displacement are collected; The velocity is obtained by performing a first-order difference process on the valve core displacement to obtain the acceleration. The mechanical motion equation is then solved based on the acceleration, velocity, mover mass, and viscous damping coefficient to obtain the actual electromagnetic force. The difference between the actual electromagnetic force and the expected output electromagnetic force of the inverse model is calculated, and the difference is processed by a sign function to obtain the sign value. The sign values ​​are arranged sequentially within a sliding window of a preset length to obtain the compensated residual sign sequence.

[0020] Specifically, the valve core displacement is processed by first-order differential to obtain the velocity, and the velocity is then processed by first-order differential to obtain the acceleration. The differential calculation is performed by dividing the displacement difference between adjacent sampling times by the control cycle time interval, which is 0.2 ms. The mechanical motion equation sums the product of acceleration and mover mass, and the product of velocity and viscous damping coefficient, to obtain the actual electromagnetic force on the voice coil motor coil at the current moment. This calculation process converts the displacement sensor signal into force domain information, allowing the compensation effect to be directly quantified in the force domain. The inverse model expects the output electromagnetic force to be the target electromagnetic force required by the control law of the current control cycle. After subtracting the actual electromagnetic force from the expected output electromagnetic force of the inverse model, a sign function is applied to the difference. When the difference is greater than zero, the sign value is +1; when the difference is less than zero, the sign value is -1; and when the difference is equal to zero, the sign value is zero. This sign function operation removes the amplitude information, retaining only the deviation state of the compensation direction.

[0021] The symbol values ​​are arranged sequentially according to the acquisition time within a sliding window of 20 control cycles, forming a compensation residual symbol sequence. The window length is set to 4 ms, corresponding to 20 control cycles. This duration, under a control cycle of 0.2 ms, can cover approximately two complete commutation cycles under a 5Hz sinusoidal excitation. This ensures that the window contains a sufficient number of symbol samples before and after the commutation to support subsequent mean calculations, while avoiding an excessively long window that would overwhelm the symbol structure changes near the current commutation. The sliding window extends to the left from the current time as its right endpoint, sliding one step to the right each control cycle. The latest symbol value within the window replaces the oldest symbol value in real time, ensuring that the compensation residual symbol sequence always reflects the temporal structure of the compensation direction deviation within the most recent 20 control cycles.

[0022] In one specific embodiment, step S2 includes: At the commutation moment of the coil current, the compensation residual symbol sequence is truncated to obtain the forward symbol subsequence before the commutation moment and the backward symbol subsequence after the commutation moment; The mean of the forward symbol subsequence is calculated, and the mean of the backward symbol subsequence is calculated. Multiply the mean of the forward signs by the mean of the backward signs to obtain the reversal strength index of the reversal sign. Within a series of commutation moments, the proportion of commutation events where the commutation sign reversal intensity index is lower than a preset threshold is statistically analyzed. The proportion is then compared with the discrimination threshold to obtain the commutation direction offset diagnostic indicator.

[0023] Specifically, the commutation moment of the coil current is defined as the moment when the sign of the coil current changes between two adjacent control cycles, i.e., when the product of the coil current in the previous cycle and the coil current in the current cycle is negative. At this moment, five symbol values ​​are truncated from the compensation residual symbol sequence before the commutation moment to form a forward symbol subsequence, and five symbol values ​​are truncated from the commutation moment to form a backward symbol subsequence. The truncation length is set to 1 ms, corresponding to five control cycles. This duration is sufficient to reflect the stable deviation of the compensation direction before and after the commutation, while avoiding the inclusion of symbols from the far-end non-commutation interval in the calculation due to an excessively wide truncation range. The arithmetic mean of the five symbol values ​​in the forward symbol subsequence is obtained, and the arithmetic mean of the five symbol values ​​in the backward symbol subsequence is obtained. The values ​​of both means are between -1 and +1. The larger the absolute value of the mean, the more concentrated and consistent the corresponding side's symbols are.

[0024] Multiplying the mean of the forward sign by the mean of the backward sign yields the reversal sign intensity index, which ranges from -1 to +1. When there is a concentrated reversal in the compensation residual direction before and after the reversal, the means on both sides have opposite signs, the product is negative and tends towards -1, and the product is positive when the compensation direction does not reverse before and after the reversal. The preset threshold is -0.6. A value below this indicates that the degree of opposite sign of the mean of the forward and backward signs exceeds 60%, and the reversal event is judged as a direction-biased reversal event. The number of direction-biased reversal events is counted within three consecutive reversal moments and divided by 3 to obtain the percentage. Three reversal moments are used as the statistical window because a single reversal may be affected by sensor noise, which may lead to occasional misjudgment. Three statistical counts strike a balance between timely response and noise resistance. The percentage is compared with the discrimination threshold of two-thirds. When the percentage is not less than two-thirds, the commutation direction bias diagnosis flag is set to the triggered state. When it is less than two-thirds, the diagnosis flag remains in the non-triggered state. The discrimination threshold of two-thirds means that at least two out of three commutations are judged as commutation direction bias type, thus excluding the influence of single occasional abnormalities on the diagnosis results.

[0025] Figure 2 This is a schematic diagram of the timing structure of the compensation residual symbol sequence in the embodiments of this application. Figure 2In the middle layer, the upper curve is the coil current waveform of the voice coil motor under 5Hz sinusoidal excitation, and the gray vertical bars mark the commutation time intervals. The middle curve is the compensation residual formed by the difference between the actual electromagnetic force and the expected electromagnetic force after the inverse model compensation. It can be seen that the compensation residual has a sudden amplitude change near the commutation time. The lower layer is the compensation residual symbol sequence obtained by performing sign function processing on the compensation residual cycle by cycle and arranging it in time sequence within the sliding window. The symbol sequence is randomly and alternately distributed in the non-commutation interval, while the symbol concentration and reversal structure feature appears in the short time window before and after the commutation time. This feature is the observable basis for commutation direction bias diagnosis. The commutation symbol reversal intensity index is obtained by multiplying the average of the symbols in the five steps before and after the commutation. When the product is lower than the preset threshold of -0.6, it is judged as a direction bias type commutation event.

[0026] In one specific embodiment, in step S3, when the commutation direction bias diagnostic flag is triggered, the coil current is divided into a positive current region and a negative current region, including: When the commutation direction bias diagnostic flag is triggered, the operating range of the coil current is divided into a positive current region and a negative current region, with the coil current being zero as the boundary. When the coil current is greater than zero, the coil current is classified as the positive current region; when the coil current is less than zero, the coil current is classified as the negative current region. At the sampling moment when the absolute value of the coil current is lower than the dead zone threshold, the current parameter identification results of the positive current region and the negative current region are processed by linear interpolation to obtain the force constant parameters of the transition region. The force constant parameters of the transition region are substituted into the inverse model equation to obtain the inverse model compensation current of the transition region.

[0027] Specifically, after the commutation direction bias diagnostic flag is triggered, the operating range of the coil current is divided into a positive current region and a negative current region, with the coil current equal to zero as the boundary. The positive current region corresponds to all sampling moments when the coil current is greater than zero, and the negative current region corresponds to all sampling moments when the coil current is less than zero. The division is based on the hysteresis effect in the voice coil motor's magnetic circuit, which causes the positive and negative currents to generate different actual electromagnetic forces at the same amplitude. The force constant polynomial coefficients on both sides have systematic differences. If a unified parameter set is used for identification, the data on both sides will contaminate each other, leading to directional compensation errors at the commutation moment. Therefore, the data on both sides must be strictly isolated. At sampling moments when the coil current is greater than zero, the coil current and compensation residual at that moment are included in the positive current region for identification and updating of the positive force constant parameter set. At sampling moments when the coil current is less than zero, the coil current and compensation residual at that moment are included in the negative current region for identification and updating of the negative force constant parameter set.

[0028] The dead zone threshold is set to 0.05A. This value is based on the measurement noise amplitude of the current sensor. When the absolute value of the coil current is below 0.05A, the sensor output signal is greatly affected by noise, and the current sign judgment is uncertain. Forcing the current into a positive or negative current region within this range will introduce false data into parameter identification. At the sampling moment when the absolute value of the coil current is below the dead zone threshold, the corresponding components of the positive and negative force constant parameter sets are linearly interpolated. The interpolation weight is determined by the relative position of the current coil current between the positive and negative dead zone thresholds. The closer the coil current is to the positive dead zone threshold, the greater the weight of the positive parameter set; the closer it is to the negative dead zone threshold, the greater the weight of the negative parameter set. The interpolation result is the transition zone force constant parameter, which contains three components: the constant term, the first term, and the second term coefficients of the stress constant polynomial. The force constant polynomial describes the nonlinear relationship between the electromagnetic force and the coil current of a voice coil motor. The electromagnetic force is equal to the product of the force constant and the coil current. The force constant itself is a quadratic polynomial of the coil current. These three components together determine the magnitude of the compensation current required under a given desired electromagnetic force. The inverse model solution uses the desired electromagnetic force as a known quantity and the compensation current as an unknown quantity. The three components of the force constant parameter in the transition region are substituted into this polynomial relationship. The compensation current value that satisfies the desired electromagnetic force condition is gradually approximated by Newton's iteration method. The initial value of the iteration is the compensation current value of the previous control cycle. The iteration terminates when the absolute value of the difference between two adjacent iteration results is less than 0.001A. The inverse model compensation current in the transition region is obtained. The outputs of the positive and negative inverse models are smoothly connected in the commutation transition interval, avoiding control jitter introduced by hard parameter switching.

[0029] In one specific embodiment, step S3 involves independently performing recursive parameter identification using the coil current and the compensation residual in the two intervals, including: Within the forward current region, the coil current and its squared terms form a regression vector. The regression vector is then used in conjunction with the recursive least squares identifier with a forgetting factor to update the parameters, resulting in the forward prediction error. Based on the forward prediction error, the gain vector, and the covariance matrix, the forward force constant parameter set is recursively corrected to obtain the updated forward force constant parameter set. Within the negative current region, a regression vector is constructed using the absolute value of the coil current and its squared terms. The regression vector is then used in conjunction with a recursive least squares identifier with a forgetting factor to update the parameters, resulting in the negative prediction error. Based on the negative prediction error, the gain vector, and the covariance matrix, the negative force constant parameter set is recursively corrected to obtain the updated negative force constant parameter set. When the coil current is in the positive current region, the negative force constant parameter set remains frozen and is not updated; when the coil current is in the negative current region, the positive force constant parameter set remains frozen and is not updated, so that the identification data of the two parameter sets do not interfere with each other.

[0030] Specifically, within the forward current region, a regression vector is constructed using three components: the coil current value at the current sampling moment, the square of the coil current value, and a constant term. The inner product of this regression vector and the three components in the forward force constant parameter set is the predicted value of the actual electromagnetic force by the inverse model at the current moment. The difference between this predicted value and the compensation residual at the current moment yields the forward prediction error, which reflects the fitting deviation of the current forward force constant parameter set to the nonlinear characteristics of the electromagnetic force in the forward current region. The gain vector is obtained by dividing the product of the current covariance matrix and the regression vector by the sum of the forgetting factor, the product of the regression vector, and the covariance matrix. The physical meaning of the gain vector is the correction weight of the newly arrived data on the parameter estimation; the greater the difference between the new data and the historical data, the larger the magnitude of the gain vector and the stronger the correction. The three components of the forward force constant parameter set are recursively corrected for the current period based on the product of the gain vector and the forward prediction error, resulting in the updated forward force constant parameter set. The covariance matrix is ​​scaled by a forgetting factor after the parameters are updated. The forgetting factor is 0.97. This value causes the weight of historical data from about 33 control cycles ago to be reduced to about one-third of the initial weight. This achieves a balance between parameter tracking speed and identification stability. If the forgetting factor is too large, the response to gradual changes in operating conditions will be sluggish. If it is too small, the identification results will be affected by single-cycle noise and fluctuate drastically.

[0031] Within the negative current region, a regression vector is constructed using the absolute value of the coil current, its squared term, and the constant term. The reason for using the absolute value of the coil current instead of the original negative value is that the force constant polynomial in the negative current region describes the electromagnetic force characteristics with the current amplitude as the independent variable. If a negative value is directly substituted, the signs of the polynomial terms will become confused, leading to distorted identification results. The recursive calculation logic for the negative current region is completely symmetrical to that for the positive current region. The negative prediction error, negative gain vector, and negative covariance matrix are independently recursively derived according to the same rules, resulting in the updated set of negative force constant parameters. The freezing mechanism operates as follows: when the coil current is in the positive current region, the three components of the negative force constant parameter set and their corresponding covariance matrix remain unchanged from the previous time step during the control cycle. When the coil current is in the negative current region, the positive force constant parameter set and its covariance matrix are also frozen. The freezing mechanism completely cuts off the cross-contamination between the two parameter sets at the data level, ensuring that the positive force constant parameter set is driven only by the measured data in the positive current region and the negative force constant parameter set is driven only by the measured data in the negative current region. The two parameter sets converge independently to the true electromagnetic force nonlinearity characteristics of the corresponding current region.

[0032] In one specific embodiment, step S3 uses the vector norm of the difference between the positive force constant parameter set and the negative force constant parameter set as an index of the degree of hysteresis asymmetry, including: The difference vector is obtained by subtracting the corresponding parameter components of the positive force constant parameter set and the negative force constant parameter set; The components in the difference vector are weighted and summed according to preset weight coefficients to obtain a weighted sum of squares. The constant term component has the largest weight, and the quadratic term component has the smallest weight. The square root of the weighted sum of squares is then taken to obtain the hysteresis asymmetry index. The hysteresis asymmetry index is updated in real time during each control cycle. The updated hysteresis asymmetry index is then input into a piecewise linear mapping for transformation to obtain the commutation disturbance compensation weight.

[0033] Specifically, both the positive and negative force constant parameter sets contain three components. For the constant term coefficient, first-order coefficient, and quadratic coefficient of the stress constant polynomial, the three components corresponding to the positions in the positive and negative parameter sets are subtracted one by one, resulting in a difference vector containing three components. The absolute value of each component in the difference vector reflects the degree of parameter deviation in the positive and negative current regions at that order. These three components collectively describe the asymmetrical distribution of the hysteresis effect in each order of the force constant polynomial under the current operating condition of the voice coil motor. The three components of the difference vector are squared and then summed according to preset weighting coefficients. The constant term component has a weight of 1.0, the first-order component has a weight of 0.5, and the quadratic component has a weight of 0.1. The reason for the decreasing weights is that the error in the constant term coefficient has the most direct impact on the direction of electromagnetic force compensation at low current commutation, the error in the first-order coefficient has a moderate impact, and the error in the quadratic coefficient is only significant under high current conditions. The weight ratio of the three reflects the relative contribution of the parameter deviations of each order to the commutation compensation accuracy. The hysteresis asymmetry index is obtained by taking the square root of the weighted sum of squares. The square root process keeps the dimensions of the index consistent with those of the components of the difference vector, which facilitates numerical comparison with the low and high thresholds of the subsequent piecewise linear mapping.

[0034] The hysteresis asymmetry index is recalculated in real time during each control cycle as the positive force constant parameter set and the negative force constant parameter set are updated. The index value increases when the difference between the two parameter sets widens, and the index value approaches zero when the two parameter sets become consistent. The real-time updated hysteresis asymmetry index is input into a piecewise linear mapping. This mapping divides the index value into three intervals with a low threshold of 0.5 N / A and a high threshold of 2.0 N / A as the dividing points. When the index value is below 0.5 N / A, the lower limit of the mapping output is 1.0, and when the index value is above 2.0 N / A, the upper limit of the mapping output is 2.5. When the index value is between 0.5 N / A and 2.0 N / A, the mapping output is calculated by linear proportional interpolation. The low threshold of 0.5 N / A corresponds to the typical difference amplitude between the two parameter sets under slight hysteresis conditions, and the high threshold of 2.0 N / A corresponds to the amplitude boundary where the parameter difference reaches saturation under significant hysteresis conditions. The upper limit of 2.5 is set to prevent the extended state observer gain from being too large, which would amplify the sensor noise and cause control oscillations. The mapping output is the commutation disturbance compensation weight. This weight is updated in real time in each control cycle and is used to adjust the total disturbance estimation update gain of the extended state observer within the commutation window.

[0035] In one specific embodiment, step S4 includes: Within the commutation window before and after the coil current commutation moment, the hysteresis asymmetry index is compared with the preset low threshold and high threshold. When the hysteresis asymmetry index is lower than the low threshold, the commutation disturbance compensation weight is reset to the lower limit value. When the hysteresis asymmetry index is higher than the high threshold, the commutation disturbance compensation weight is reset to the upper limit value. When the hysteresis asymmetry index is between the low threshold and the high threshold, the hysteresis asymmetry index is linearly interpolated to obtain the commutation disturbance compensation weight. The commutation disturbance compensation weight is multiplied by the total disturbance estimate update gain of the extended state observer to obtain the adjusted total disturbance estimate update gain. Outside the commutation window, the total disturbance estimate update gain is restored to the standard value. When the coil current is in the positive current region, the components of each parameter in the positive force constant parameter set and the desired electromagnetic force constitute a nonlinear equation system. The nonlinear equation system is solved iteratively. When the coil current is in the negative current region, the components of each parameter in the negative force constant parameter set and the desired electromagnetic force constitute a nonlinear equation system. The nonlinear equation system is solved iteratively to obtain the inverse model compensation current. The inverse model compensation current is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is then applied to the coil of the voice coil motor to complete the position closed-loop control.

[0036] Specifically, the commutation window is defined as a time interval of 7 control cycles, centered on the moment of coil current commutation and extending forward and backward by 3 control cycles. The window width is 1.4 ms, corresponding to 7 control cycles. This duration covers the duration of hysteresis shock disturbance during the commutation transition phase of the voice coil motor. If the window is too narrow, the commutation shock will not be fully covered; if the window is too wide, the steady-state control performance in the non-commutation range will be disturbed. Within the commutation window, the hysteresis asymmetry index is compared with a low threshold of 0.5 N / A and a high threshold of 2.0 N / A. When it is below 0.5 N / A, the commutation disturbance compensation weight is set to a lower limit of 1.0; when it is above 2.0 N / A, the upper limit of 2.5 is set; when it is between the two, the commutation disturbance compensation weight is calculated by linear interpolation based on the relative position of the current hysteresis asymmetry index within the range of 0.5 N / A to 2.0 N / A. The total disturbance estimate update gain of the extended state observer is the third gain component in the third-order gain parameter of the observer responsible for disturbance state correction. The standard value is configured as the cube of the observer bandwidth according to the observer bandwidth method. The adjusted total disturbance estimate update gain is obtained by multiplying the commutation disturbance compensation weight with this standard value. When the weight is greater than 1.0, the gain is amplified so that the observer tracks the commutation shock disturbance with a stronger correction force within the commutation window. Outside the commutation window, the gain is restored to the standard value to avoid amplification of steady-state noise.

[0037] The force constant polynomial describes the relationship between electromagnetic force and the product of the force constant and coil current. The force constant itself is a quadratic polynomial of the coil current amplitude, containing three components: a constant term, a first-order term, and a quadratic term. Given the desired electromagnetic force, the three components of the positive or negative force constant parameter set are substituted into this polynomial relationship. A cubic algebraic equation about the compensation current is constructed with the compensation current as the unknown. The Newton-Raphson iteration method is used to numerically solve this cubic algebraic equation. The initial value of the iteration is the compensation current value of the previous control cycle. Each iteration updates the iteration value by dividing the function value at the current iteration value by the derivative value as the correction step size. The iteration terminates when the absolute value of the difference between two adjacent iteration results is less than 0.001A. The maximum number of iterations is 5 to ensure that the calculation is completed within 0.2ms of a single control cycle, thus obtaining the inverse model compensation current. The base current command is calculated from the position tracking error using a proportional-derivative feedback control law, and then the total disturbance estimate converted current output by the extended state observer is subtracted. This reflects the base drive current required by the control law under the combined action of the current position deviation and disturbance compensation. The inverse model compensation current is summed with the basic current command to obtain the drive current command. After being amplified by power, the drive current command acts on the voice coil motor coil to generate actual electromagnetic force, which pushes the valve core to move towards the target position. The above calculation is executed cyclically within a 0.2ms control cycle to complete the position closed-loop control.

[0038] The adaptive linearization control method for voice coil motors in the embodiments of this application has been described above. The adaptive linearization control system for voice coil motors in the embodiments of this application is described below. One embodiment of the adaptive linearization control system for voice coil motors in the embodiments of this application includes: The acquisition module is used to acquire coil current and valve core displacement during the adaptive control of voice coil motor position. It extracts symbol values ​​based on the difference between the valve core displacement and the expected output of the inverse model, and arranges them in time sequence within a sliding window to obtain a compensation residual symbol sequence. The comparison module is used to calculate the mean forward sign and the mean backward sign before and after the commutation from the compensation residual sign sequence at the moment of coil current commutation, and use the product of the two as the commutation sign reversal intensity index, compare it with a preset threshold, and obtain the commutation direction bias diagnostic flag. The identification module is used to divide the coil current into a positive current region and a negative current region when the commutation direction bias diagnostic flag is triggered. In the two regions, the recursive parameter identification is performed independently with the coil current and the compensation residual to obtain the positive force constant parameter set and the negative force constant parameter set. The difference vector norm of the two is used as an index of the degree of hysteresis asymmetry. The superposition module is used to convert the hysteresis asymmetry index into a commutation disturbance compensation weight by piecewise linear mapping within the commutation window before and after the commutation time of the coil current. The total disturbance estimate update gain of the extended state observer is adjusted with the commutation disturbance compensation weight. The inverse model compensation current obtained by solving the positive force constant parameter set or the negative force constant parameter set is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is applied to the coil of the voice coil motor to realize position closed-loop control.

[0039] This invention also provides a voice coil motor adaptive linearization control device, which can be a server. The device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, designed as a computer, provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0040] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the voice coil motor adaptive linearization control method.

[0041] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0042] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a voice coil motor adaptive linearization control device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0043] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive linearization control method for a voice coil motor, characterized in that, The method includes: Step S1: During the adaptive control of the voice coil motor position, the coil current and valve core displacement are collected. The symbol value is extracted based on the difference between the valve core displacement and the expected output of the inverse model. The symbol values ​​are arranged in time sequence within the sliding window to obtain the compensation residual symbol sequence. Step S2: At the moment of coil current commutation, the mean forward sign and the mean backward sign before and after commutation are calculated from the compensation residual sign sequence. The product of the two is used as the commutation sign reversal intensity index and compared with a preset threshold to obtain the commutation direction bias diagnostic mark. Step S3: When the commutation direction bias diagnostic flag is triggered, the coil current is divided into a positive current region and a negative current region. In the two regions, the recursive parameter identification is performed independently with the coil current and the compensation residual to obtain the positive force constant parameter set and the negative force constant parameter set. The difference vector norm of the two is used as an index of the degree of hysteresis asymmetry. Step S4: Within the commutation window before and after the coil current commutation moment, the hysteresis asymmetry index is converted into a commutation disturbance compensation weight by piecewise linear mapping. The total disturbance estimate update gain of the extended state observer is adjusted with the commutation disturbance compensation weight. The inverse model compensation current obtained by solving the positive force constant parameter set or the negative force constant parameter set is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is applied to the coil of the voice coil motor to realize position closed-loop control.

2. The adaptive linearization control method for a voice coil motor according to claim 1, characterized in that, Step S1 includes: During the adaptive control of the voice coil motor position, the coil current and valve core displacement are collected; The velocity is obtained by performing a first-order difference process on the valve core displacement to obtain the acceleration. The mechanical motion equation is then solved based on the acceleration, the velocity, the mover mass, and the viscous damping coefficient to obtain the actual electromagnetic force. The difference between the actual electromagnetic force and the expected output electromagnetic force of the inverse model is calculated, and the difference is processed by a sign function to obtain a sign value. The symbol values ​​are arranged sequentially within a sliding window of a preset length to obtain a compensated residual symbol sequence.

3. The adaptive linearization control method for a voice coil motor according to claim 1, characterized in that, Step S2 includes: At the commutation moment of the coil current, the compensation residual symbol sequence is truncated to obtain the forward symbol subsequence before the commutation moment and the backward symbol subsequence after the commutation moment; The mean of the forward symbol subsequence is calculated to obtain the mean of the forward symbols, and the mean of the backward symbol subsequence is calculated to obtain the mean of the backward symbols. The average value of the forward symbols is multiplied by the average value of the backward symbols to obtain the reversal strength index of the reversal symbol. Within a series of commutation moments, the proportion of commutation events where the commutation symbol reversal intensity index is lower than a preset threshold is statistically analyzed. The proportion is then compared with a discrimination threshold to obtain a commutation direction offset diagnostic flag.

4. The adaptive linearization control method for a voice coil motor according to claim 1, characterized in that, In step S3, when the commutation direction bias diagnostic flag is triggered, the coil current is divided into a positive current region and a negative current region, including: When the commutation direction bias diagnostic flag is triggered, the operating range of the coil current is divided into a positive current region and a negative current region, with the coil current being zero as the boundary. At the sampling time when the coil current is greater than zero, the coil current is classified into the positive current region; at the sampling time when the coil current is less than zero, the coil current is classified into the negative current region. At the sampling moment when the absolute value of the coil current is lower than the dead zone threshold, the current parameter identification results of the positive current region and the negative current region are processed by linear interpolation to obtain the transition zone force constant parameter. The transition zone force constant parameter is substituted into the inverse model equation to obtain the transition zone inverse model compensation current.

5. The adaptive linearization control method for a voice coil motor according to claim 4, characterized in that, In step S3, recursive parameter identification is performed independently in the two intervals using the coil current and the compensation residual, including: Within the positive current region, a regression vector is constructed using the coil current and its squared terms. The regression vector is then used in conjunction with the recursive least squares identifier with a forgetting factor to update the parameters, resulting in a positive prediction error. Based on the positive prediction error, the gain vector, and the covariance matrix, the positive force constant parameter set is recursively corrected to obtain the updated positive force constant parameter set. Within the negative current region, a regression vector is constructed using the absolute value of the coil current and its squared term. The regression vector is then used in conjunction with the recursive least squares identifier with a forgetting factor to update the parameters, resulting in a negative prediction error. Based on the negative prediction error, the gain vector, and the covariance matrix, the negative force constant parameter set is recursively corrected to obtain the updated negative force constant parameter set. When the coil current is in the positive current region, the negative force constant parameter set remains frozen and is not updated; when the coil current is in the negative current region, the positive force constant parameter set remains frozen and is not updated, so that the identification data of the two sets of parameters do not interfere with each other.

6. The adaptive linearization control method for a voice coil motor according to claim 5, characterized in that, In step S3, the difference vector norm between the positive force constant parameter set and the negative force constant parameter set is used as an index of the degree of hysteresis asymmetry, including: The difference vector is obtained by subtracting the corresponding parameter components of the positive force constant parameter set and the negative force constant parameter set; The components in the difference vector are weighted and summed according to preset weight coefficients to obtain a weighted sum of squares, where the constant term component has the largest weight and the quadratic term component has the smallest weight. The square root of the weighted sum of squares is then performed to obtain the hysteresis asymmetry index. The hysteresis asymmetry index is updated in real time during each control cycle. The updated hysteresis asymmetry index is then input into a piecewise linear mapping for transformation to obtain the commutation disturbance compensation weight.

7. The adaptive linearization control method for a voice coil motor according to claim 1, characterized in that, Step S4 includes: Within the commutation window before and after the coil current commutation moment, the hysteresis asymmetry index is compared with a preset low threshold and a high threshold. When the hysteresis asymmetry index is lower than the low threshold, the commutation disturbance compensation weight is reset to the lower limit value. When the hysteresis asymmetry index is higher than the high threshold, the commutation disturbance compensation weight is reset to the upper limit value. When the hysteresis asymmetry index is between the low threshold and the high threshold, the hysteresis asymmetry index is linearly interpolated to obtain the commutation disturbance compensation weight. The commutation disturbance compensation weight is multiplied by the total disturbance estimation update gain of the extended state observer to obtain the adjusted total disturbance estimation update gain. The total disturbance estimation update gain is then restored to the standard value outside the commutation window. When the coil current is in the positive current region, a nonlinear equation system is formed by the parameter components of the positive force constant parameter set and the desired electromagnetic force. The nonlinear equation system is then iteratively solved. When the coil current is in the negative current region, a nonlinear equation system is formed by the parameter components of the negative force constant parameter set and the desired electromagnetic force. The nonlinear equation system is then iteratively solved to obtain the inverse model compensation current. The inverse model compensation current is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is then applied to the coil of the voice coil motor to complete the position closed-loop control.

8. An adaptive linearization control system for a voice coil motor, characterized in that, For implementing the adaptive linearization control method for a voice coil motor as described in any one of claims 1-7, the adaptive linearization control system for the voice coil motor includes: The acquisition module is used to acquire coil current and valve core displacement during the adaptive control of voice coil motor position. It extracts symbol values ​​based on the difference between the valve core displacement and the expected output of the inverse model, and arranges them in time sequence within a sliding window to obtain a compensation residual symbol sequence. The comparison module is used to calculate the mean forward sign and the mean backward sign before and after the commutation from the compensation residual sign sequence at the moment of coil current commutation, and use the product of the two as the commutation sign reversal intensity index, compare it with a preset threshold, and obtain the commutation direction bias diagnostic flag. The identification module is used to divide the coil current into a positive current region and a negative current region when the commutation direction bias diagnostic flag is triggered. In the two regions, the recursive parameter identification is performed independently with the coil current and the compensation residual to obtain the positive force constant parameter set and the negative force constant parameter set. The difference vector norm of the two is used as an index of the degree of hysteresis asymmetry. The superposition module is used to convert the hysteresis asymmetry index into a commutation disturbance compensation weight by piecewise linear mapping within the commutation window before and after the commutation time of the coil current. The total disturbance estimate update gain of the extended state observer is adjusted with the commutation disturbance compensation weight. The inverse model compensation current obtained by solving the positive force constant parameter set or the negative force constant parameter set is superimposed with the basic current command output by the feedback control law to obtain the drive current command. The drive current command is applied to the coil of the voice coil motor to realize position closed-loop control.

9. A voice coil motor adaptive linearization control device, characterized in that, The system includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the voice coil motor adaptive linearization control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the voice coil motor adaptive linearization control method as described in any one of claims 1 to 7.

Images