A rudder zero drift online compensation system

Abstract

The present application relates to the technical field of steering engine compensation, and particularly relates to a steering engine zero position drift online compensation system, comprising: a calculation module, which is used for calculating an initial zero position drift amount through a preset multi-factor coupling drift model according to a current environment temperature, a current power supply voltage and an accumulated running time length of the steering engine; an updating module, which is used for updating model parameters of the preset multi-factor coupling drift model based on the initial zero position drift amount and a current feedback position of the steering engine, to obtain an updated multi-factor coupling drift model; and the calculation module is further used for calculating a corrected zero position drift amount according to the updated multi-factor coupling drift model, correcting a control instruction of the steering engine according to the corrected zero position drift amount to drive the steering engine to move, so as to solve the problem that the compensation of the zero position drift of the steering engine depends on single temperature compensation, cannot comprehensively process voltage fluctuation and long-term wear, and the compensation precision is insufficient under complex working conditions.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of rudder compensation, in particular to a rudder zero drift online compensation system. BACKGROUND

[0002] As an actuator, the rudder is widely used in the fields of robots, aerospace and precision instruments, and the control precision of the rudder directly affects the overall performance of the system. The zero drift, which is caused by temperature changes, power voltage fluctuations and long-term running wear and tear, is a common problem of the rudder. Effective compensation of the zero drift is the key to ensuring that the rudder operates stably in a high-precision state.

[0003] At present, a common compensation method is to integrate a temperature sensor in the rudder system, to form a lookup table by pre-calibrating the zero drift at different temperatures, and to obtain a compensation value by looking up the table according to the real-time temperature during system operation for correction. Although this method can alleviate the influence of temperature drift, it only considers a single temperature variable and does not take into account the power voltage fluctuations, especially the voltage drop during dynamic operation of the rudder, and the mechanical wear and tear accumulated over time due to gear wear and lubricant aging, which are not included in the compensation step. As a result, under complex actual working conditions, especially after long-term operation, the compensation accuracy is significantly reduced, and it is difficult to meet the stability requirements of high-precision systems. SUMMARY

[0004] In order to solve the technical problems that the compensation of the rudder zero drift relies on single temperature compensation, cannot comprehensively handle voltage fluctuations and long-term wear and tear, and the compensation accuracy is insufficient under complex working conditions, the application provides a rudder zero drift online compensation system.

[0005] The rudder zero drift online compensation system provided by the application adopts the following technical scheme: A rudder zero drift online compensation system, comprising: A calculation module configured to calculate an initial zero drift amount by a preset multi-factor coupled drift model according to a current environmental temperature, a current power voltage and a cumulative running time of the rudder; An updating module configured to update model parameters of the preset multi-factor coupled drift model based on the initial zero drift amount and a current feedback position of the rudder, to obtain an updated multi-factor coupled drift model; The calculation module is further configured to calculate a corrected zero drift amount according to the updated multi-factor coupled drift model, and correct a control instruction of the rudder according to the corrected zero drift amount to drive the rudder to move.

[0006] Further, the step of calculating the initial zero drift amount by the preset multi-factor coupled drift model according to the current environmental temperature, the current power voltage and the cumulative running time of the rudder includes the following steps: After obtaining the feedback position information of the rudder under the zero position control instruction in the calibration conditions of a plurality of historical temperature-power supply voltage combinations, the zero position drift data in each calibration condition is calculated based on the difference between the feedback position information and the zero position control instruction; The model fitting parameters for describing the relationship among the zero position drift, temperature and power supply voltage are obtained through fitting processing according to the historical temperature-power supply voltage combinations and the zero position drift data in each calibration condition; On the basis of the model fitting parameters, the historical cumulative running time length at the same time stamp as each historical temperature-power supply voltage combination is introduced to construct a multi-factor coupling compensation function, and the multi-factor coupling compensation function is taken as the preset multi-factor coupling drift model.

[0007] Further, the step of obtaining the feedback position information of the rudder under the zero position control instruction in the calibration conditions of a plurality of historical temperature-power supply voltage combinations comprises: The analog feedback signal returned by the rudder under the zero position control instruction in each calibration condition is received, and the analog feedback signal is subjected to analog-digital conversion sampling to obtain a sampling feedback value; The sampling feedback value is subjected to statistical filtering processing to obtain a steady-state feedback value for representing the position of the rudder in the corresponding calibration condition; The steady-state feedback value is associated with the historical temperature-power supply voltage combination in the corresponding calibration condition to generate the feedback position information.

[0008] Further, the step of obtaining the model fitting parameters for describing the relationship among the zero position drift, temperature and power supply voltage through fitting processing according to the historical temperature-power supply voltage combinations and the zero position drift data in each calibration condition comprises: The historical temperature-power supply voltage combinations and the zero position drift data in each calibration condition are subjected to data separation to obtain a temperature influence data set and a voltage influence data set; The first group of model coefficients for describing the relationship between the zero position drift and the temperature are obtained through one-dimensional nonlinear regression fitting based on the temperature influence data set, and the second group of model coefficients for describing the relationship between the zero position drift and the voltage are obtained through one-dimensional nonlinear regression fitting based on the voltage influence data set; The predicted drift values corresponding to each historical temperature-power supply voltage combination are calculated according to the first group of model coefficients and the second group of model coefficients, and the residual data set is calculated according to the difference between each predicted drift value and the zero position drift data corresponding to each predicted drift value; The third group of model coefficients for describing the interaction are obtained through multivariate regression fitting based on the residual data set and each historical temperature-power supply voltage combination; The first group of model coefficients, the second group of model coefficients and the third group of model coefficients are merged to obtain the model fitting parameters.

[0009] Further, based on the model fitting parameters, the historical cumulative running time at the same timestamp as each historical temperature-power supply voltage combination is introduced, and the step of constructing the multi-factor coupling compensation function includes: The first group of model coefficients, the second group of model coefficients, and the third group of model coefficients are combined with each historical temperature-power supply voltage combination to calculate the first predicted drift component related to temperature and voltage in the zero drift data; The power transformation is performed on each historical cumulative running time to obtain a plurality of running time characteristic items of different powers; According to the zero drift data, the first predicted drift component, and the plurality of running time characteristic items, a multi-element regression fitting is performed to obtain a fourth group of model coefficients for describing the nonlinear relationship between the zero drift and the running time; Based on the model fitting parameters and the fourth group of model coefficients, a multi-factor coupling compensation function is constructed, including a temperature first-order term, a voltage first-order term, a running time power term, a temperature-voltage interaction term, a temperature-time interaction term, and a voltage-running time interaction term.

[0010] Further, the preset multi-factor coupling drift model includes the first group of model coefficients, the second group of model coefficients, the third group of model coefficients, and the fourth group of model coefficients, and according to the current environmental temperature, the current power supply voltage, and the cumulative running time of the rudder, the step of calculating the initial zero drift amount by the preset multi-factor coupling drift model includes: According to the current environmental temperature and the first group of model coefficients, a temperature drift component is calculated, according to the current power supply voltage and the second group of model coefficients, a voltage drift component is calculated, and according to the cumulative running time and the third group of model coefficients, a time drift component is calculated; The current environmental temperature, the current power supply voltage, the cumulative running time, and the fourth group of model coefficients are calculated to obtain a coupling drift component; The temperature drift component, the voltage drift component, the time drift component, and the coupling drift component are summed to obtain the initial zero drift amount.

[0011] Further, based on the initial zero drift amount and the current feedback position of the rudder, the model parameter updating of the preset multi-factor coupling drift model is performed to obtain the updated multi-factor coupling drift model, and the step includes: After calculating the current zero drift amount according to the current control instruction and the current feedback position of the rudder, the difference between the current zero drift amount and the initial zero drift amount is calculated to obtain a drift prediction error; Based on the drift prediction error, the current environmental temperature, the current power supply voltage, the cumulative running time, and the model parameters of the preset multi-factor coupling drift model, recursive estimation is performed to obtain the current update increment of the model parameters; Summing up the model parameters and the current update increment to update the model parameters, to obtain an updated multi-factor coupled drift model.

[0012] The beneficial effects achieved are: The application provides an online compensation system for zero drift of a steering engine, comprising: a calculation module, configured to calculate an initial zero drift amount according to a current ambient temperature, a current power supply voltage and a cumulative running time length of the steering engine through a preset multi-factor coupled drift model; an update module, configured to update model parameters of the preset multi-factor coupled drift model based on the initial zero drift amount and a current feedback position of the steering engine, to obtain an updated multi-factor coupled drift model; and the calculation module is further configured to correct a control instruction of the steering engine according to the corrected zero drift amount after calculating the corrected zero drift amount according to the updated multi-factor coupled drift model, to drive the steering engine to move.

[0013] That is, in the application, when calculating the initial zero drift amount, the current ambient temperature, the current power supply voltage and the cumulative running time length are simultaneously taken as input variables through the preset multi-factor coupled drift model, which fundamentally changes the compensation logic of the traditional method that only relies on a single temperature variable, so that the instantaneous deviation caused by voltage fluctuation and the progressive deviation caused by the gradual accumulation of mechanical wear over time are jointly included in the calculation framework of real-time compensation, thereby directly solving the technical defect that single temperature compensation cannot cover voltage and wear factors; then, an online model parameter updating mechanism based on the initial zero drift amount and the current feedback position is introduced, so that the multi-factor coupled drift model becomes a dynamic self-learning system, which can continuously correct the model parameters according to the current feedback data in actual operation, and real-time approximate and track the drift rule changes caused by complex working condition changes such as voltage instability and wear aggravation, thereby effectively overcoming the fundamental problems of poor adaptability and insufficient long-term precision of fixed parameter models in complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is a module schematic diagram of the online compensation system for zero drift of the steering engine of the application; Figure 2 is a construction execution flowchart of the preset multi-factor coupled drift model of the application; Figure 3 is a step execution flowchart for calculating the initial zero drift amount based on the preset multi-factor coupled drift model of the application; Figure 4 is a step execution flowchart for updating the preset multi-factor coupled drift model of the application.

[0015] BRIEF DESCRIPTION OF DRAWINGS 10, calculation module; 20, update module. DETAILED DESCRIPTION

[0016] The following description is made in connection with Figures 1 to 4 The application is further described in detail.

[0017] The application discloses a rudder zero drift online compensation system.

[0018] Please refer to Figure 1 The rudder zero drift online compensation system comprises: The calculation module is configured to calculate an initial zero drift amount according to the current environmental temperature, the current power supply voltage and the cumulative running time of the rudder by using a preset multi-factor coupling drift model; the updating module is configured to update the model parameters of the preset multi-factor coupling drift model based on the initial zero drift amount and the current feedback position of the rudder, so as to obtain an updated multi-factor coupling drift model; and the calculation module is further configured to correct the control command of the rudder according to the corrected zero drift amount after the corrected zero drift amount is calculated according to the updated multi-factor coupling drift model, so as to drive the rudder to move.

[0019] By constructing a system that can simultaneously respond to multiple environmental and aging factors and has self-correction capability, the limitations of traditional fixed or single-factor compensation methods are overcome. The combination of the calculation module and the updating module realizes a closed-loop process of “prediction-execution-feedback-optimization”. The calculation module first calculates the initial zero drift amount under the joint action of the current environmental temperature, the current power supply voltage and the cumulative running time according to the preset multi-factor coupling drift model, so that the compensation reference is no longer dependent on temperature alone, but covers several key physical variables that affect the zero accuracy, providing a comprehensive initial basis for compensation.

[0020] The updating module compares the initial zero drift amount calculated by the preset multi-factor coupling drift model with the real deviation revealed by the current feedback position of the rudder, and updates the model parameters of the preset multi-factor coupling drift model online by using the difference, so as to give the whole system the self-learning and self-adaptive capability, and enable the multi-factor coupling drift model to dynamically evolve with the changes of the actual operation conditions of the rudder (such as changes in voltage fluctuation mode and wear entering a new stage), so as to establish an intelligent compensation mechanism that can not only consider the combined effects of temperature, voltage and wear for accurate compensation at any time, but also track the system characteristic drift through continuous small parameter updates in long-term operation, thereby continuously maintaining high compensation accuracy in complex and variable application scenarios.

[0021] In a feasible implementation manner, referring to Figure 2 The specific construction manner of the preset multi-factor coupling drift model comprises steps S01-S03: Step S01, after obtaining the feedback position information of the rudder under the zero position control instruction under the calibration conditions of a plurality of historical temperature-power supply voltage combinations, the zero position drift data under each calibration condition is calculated based on the difference between the feedback position information and the zero position control instruction.

[0022] By collecting the historical zero position deviation data of the rudder under the influence of different temperature and power supply voltage combinations in a controlled laboratory environment, a plurality of calibration conditions defined by temperature and power supply voltage are created and covered to obtain observation samples that can clearly reflect the influence of single variable and interaction between variables on zero position drift, so as to controllably decompose and observe the drift phenomenon mixed together in the field, obtain the corresponding zero position drift data, and thus avoid isolated data points with unknown interference in the field, but generate a series of historical temperature-power supply voltage combinations with clear calibration condition labels, which can be used for quantitative analysis of causal relationship, provide the possibility for subsequent separation of independent influence and coupling influence of temperature and voltage, and ensure that the multi-factor coupling drift model is established based on systematic experimental observation rather than empirical assumption, thereby improving the objectivity and reliability of the preset multi-factor coupling drift model at the root.

[0023] Among them, the feedback position information obtained is subtracted from the theoretical position information corresponding to the zero position control instruction received by the rudder under the corresponding calibration condition by algebraic operation. It should be noted that the zero position control instruction usually corresponds to a known theoretical position reference value (for example, the midpoint of the PWM duty cycle or the digital quantity corresponding to the theoretical zero voltage value of the potentiometer feedback voltage), and the feedback position information is the digital quantity corresponding to the position actually reached and maintained by the rudder under the zero position control instruction and the current calibration condition. The difference between the two is directly quantified as the zero position drift data under the corresponding historical temperature-power supply voltage combination, which can be angle, minimum driving unit or analog-to-digital conversion count, providing a unique reliable and unambiguous quantitative data basis for subsequent separation and quantification of the independent influence of temperature and power supply voltage and the coupling influence between them, ensuring that the input of the entire multi-factor coupling drift model construction process is accurate and objective.

[0024] Further, step S01 further includes steps S011-S013: Step S011, receiving the analog feedback signal returned by the rudder under the zero position control instruction under each calibration condition, and performing analog-to-digital conversion sampling on the analog feedback signal to obtain a sampled feedback value.

[0025] In each calibration condition constituted by a specific ambient temperature and a specific power supply voltage, the control system sends a pulse width modulation zero position control command representing the theoretical zero position to the steering engine, and the internal potentiometer of the steering engine outputs a voltage signal proportional to the actual angle of the steering wheel under the zero position control command, which is the analog feedback signal.

[0026] Then, the analog feedback signal is connected to the analog-to-digital converter input pin of the microcontroller. After the steering engine moves according to the zero position control command and reaches a steady state, a pre-defined time window is started. Within the time window, the analog-to-digital converter periodically triggers the conversion process at a pre-configured fixed sampling frequency. Each time the conversion process is triggered, the sampling and holding circuit inside the analog-to-digital converter instantaneously captures and holds the instantaneous voltage value of the analog feedback signal. Then, the instantaneous voltage value is compared with a reference voltage value, and through successive approximation, it is converted into a digital code proportional to the input voltage. The digital code is an integer, and its value is calculated by the formula wherein, is the instantaneous voltage value, is the reference voltage, is the resolution bit number of the analog-to-digital converter. Each conversion obtains a digital quantity representing the instantaneous voltage value at the corresponding sampling time. The analog-to-digital converter continuously performs the above capture, hold and conversion operations multiple times within the fixed time window according to the set sampling frequency, thereby obtaining a sequence of sampling feedback values arranged in time sequence. The physical position information of the steering engine under the specific condition is converted from the continuous and easily disturbed analog domain to the digital domain value sequence that can be directly stored and processed by the system, providing a premise for subsequent extraction of reliable data that can truly reflect the steady state position of the steering engine under the corresponding calibration condition through digital signal processing technology, thereby ensuring the accuracy and repeatability of data acquisition.

[0027] In step S012, statistical filtering is performed on the sampling feedback values to obtain a steady state feedback value representing the position of the steering engine under the corresponding calibration condition.

[0028] After the analog-digital conversion is completed, the microprocessor reads and stores all the sampling feedback values obtained in the time window, sums all the sampling feedback values and divides the sum by the total number of sampling points, and the quotient is the steady-state feedback value for representing the position of the rudder under the corresponding calibration condition. By using the inherent random noise suppression characteristics of the arithmetic average algorithm, random errors introduced by electromagnetic interference, circuit thermal noise, etc. in the sampling process are effectively smoothed and offset, so that from the sampling feedback values containing fluctuations and interference, a single digital quantity that best represents the actual position of the rudder under the corresponding stable condition is extracted, which is stable and reliable, and provides high-quality input for subsequent accurate calculation of zero drift data, and fundamentally avoids distortion of the model calibration data caused by accidental errors or transient interference in a single sampling.

[0029] Step S013, associating the steady-state feedback value with the historical temperature-power supply voltage combination under the corresponding calibration condition to generate feedback position information.

[0030] After the data acquisition and processing of one calibration condition are completed, a data record is created, which at least includes the steady-state feedback value calculated under the corresponding calibration condition, so as to bind the steady-state feedback value representing the position of the rudder with the two experimental condition parameters (i.e. historical temperature and historical power supply voltage) that lead to the result, form a complete and ordered feedback position information with clear meaning, and provide structured data input related to the associated condition label for the subsequent model fitting step, ensuring that each position data point corresponds to the specific environmental conditions under which it is generated when analyzing the zero drift rule, so that it is possible to separate the independent effects of temperature and voltage and their interaction from the data, thereby laying a data foundation for constructing a high-precision multi-factor coupled drift model.

[0031] Step S02, fitting processing according to the historical temperature-power supply voltage combination and the zero drift data under each calibration condition to obtain model fitting parameters for describing the relationship between the zero drift, temperature and power supply voltage.

[0032] Based on the zero drift data obtained in the previous step, the influence law of temperature and power voltage on zero drift is refined and quantified from the experimental data. The core role is to convert the discrete and specific calibration condition observation data into a general mathematical relationship that can continuously describe the zero drift prediction value under any given historical temperature-power voltage combination, thereby providing a core algorithm basis for real-time calculation of the drift amount caused by environmental factors. Through the fitting process, not only the linear or nonlinear coefficients of temperature and voltage independent influence on drift are obtained, but more importantly, the coupling interaction effect between temperature and voltage is separated and captured, so that the subsequent multi-factor coupling drift model can more truly reflect the comprehensive influence under complex physical field, fundamentally overcoming the defects of existing technology that only considers single temperature variable and ignores voltage fluctuation and its interaction, laying a key parameter foundation for constructing a precise compensation model that can respond to changes in environmental temperature and power voltage.

[0033] Further, step S02 further includes steps S021-S025: Step S021, the historical temperature-power voltage combination and zero drift data under each calibration condition are separated to obtain a temperature influence data set and a voltage influence data set.

[0034] From all the historical temperature-power voltage combinations of all calibration conditions, multiple data sets with the same power voltage value but different temperature values are selected, the corresponding relationship between the zero drift data and the temperature change in these data sets is extracted, and the temperature influence data set is formed by summarizing the corresponding relationship. The data set reflects the law of zero drift change with temperature under the condition of constant voltage as a single variable; secondly, from the same historical temperature-power voltage combination of all calibration conditions, multiple data sets with the same temperature value but different power voltage values are selected, the corresponding relationship between the zero drift data and the voltage change in these data sets is extracted, and the voltage influence data set is formed by summarizing the corresponding relationship. The data set reflects the law of zero drift change with voltage under the condition of constant temperature as a single variable. In this way, the complex influence of temperature and voltage is preliminarily decoupled by the idea of control variable method, and two relatively pure data subsets with temperature and voltage as the dominant influencing factors are obtained, which provides a clear structure and strong pertinence input data basis for subsequent one-dimensional nonlinear regression fitting to accurately quantify the independent influence of temperature and voltage.

[0035] Step S022, one-dimensional nonlinear regression fitting is performed based on the temperature influence data set to obtain a first group of model coefficients for describing the relationship between zero drift and temperature, and one-dimensional nonlinear regression fitting is performed based on the voltage influence data set to obtain a second group of model coefficients for describing the relationship between zero drift and voltage.

[0036] The temperature values ​​in the temperature-affected dataset are used as independent variables, and their corresponding zero-point drift data are used as dependent variables. These are then substituted into a pre-defined univariate nonlinear equation. Specifically, this refers to a series of temperature values ​​recorded in a dataset that affect temperature. Substituting each value into the preset univariate nonlinear equation, the corresponding predicted drift values ​​are calculated. After generating the predicted drift value sequence, the zero-point drift data corresponding to this predicted drift value sequence in the temperature influence dataset are calculated. The difference between them, i.e., the difference Based on the principle of least squares, a method is constructed with parameters [a, b, c] as the optimization objective and the sum of squared differences of all data points as the optimization criterion. For the optimization problem of the loss function, an iterative optimization algorithm (such as the Gauss-Newton method or the Levenberg-Marquardt algorithm) is used to minimize the loss function. Starting from an initial parameter estimate, the gradient or Jacobian matrix of the loss function with respect to the parameters [a, b, c] is calculated in each iteration, and the parameter estimate is updated in the direction that makes the loss function decrease. This process is repeated until the change in parameters is less than a preset threshold or the maximum number of iterations is reached. The set of parameters [a, b, c] that minimizes the sum of squared differences is the first set of model coefficients.

[0037] Furthermore, the power supply voltage values ​​in the voltage influence dataset are used as independent variables, and their corresponding zero-point drift data are used as dependent variables, which are then substituted into another pre-defined univariate nonlinear equation. Similarly, the least squares method is used to calculate a set of parameters [d, e, f]. These parameters are the second set of model coefficients describing the relationship between zero drift and voltage. In this way, the nonlinear influence of ambient temperature on zero drift and the nonlinear influence of power supply voltage on zero drift under ideal single variable conditions are mathematically quantified, thus providing a clear and quantitative independent benchmark for separating and calculating the coupled interaction between temperature and voltage in subsequent steps.

[0038] It should be noted that the calculation logic of the second set of model coefficients is the same as that of the first set of model coefficients, so it will not be repeated.

[0039] Step S023: Based on the first set of model coefficients and the second set of model coefficients, calculate the predicted drift values ​​corresponding to each historical temperature-power supply voltage combination, and calculate the residual dataset based on the difference between each predicted drift value and the zero drift data corresponding to each predicted drift value.

[0040] For each historical temperature-power supply voltage combination (T, V), first, the temperature value T is substituted into the monomial nonlinear equation determined by the first set of model coefficients obtained in step S022 to calculate the drift component contributed by temperature independently; at the same time, the voltage value V is substituted into another monomial nonlinear equation determined by the second set of model coefficients obtained in step S022 to calculate the drift component contributed by voltage independently; then, the two drift components are added together, and the sum is the predicted drift value corresponding to the calibration condition.

[0041] After obtaining the predicted drift values of all calibration conditions, for each historical temperature-power supply voltage combination, the corresponding predicted drift value is subtracted from the corresponding zero-point drift data, and the difference is the corresponding residual. By traversing each historical temperature-power supply voltage combination and collecting all residuals, a set of residual data is formed. By subtracting the quantified temperature and voltage independent influence components (i.e., the predicted drift value) from the measured total drift data (i.e., the zero-point drift data), the remaining part that cannot be explained by the independent effects of temperature and voltage is effectively stripped and separated from the original data. This residual data mainly contains the influence of the nonlinear coupling interaction between temperature and voltage on the zero-point drift, providing pure and targeted data input for subsequent modeling of the interaction effect.

[0042] Step S024, based on the residual data set and each historical temperature-power supply voltage combination, a multiple regression fitting is performed to obtain a third set of model coefficients for describing the interaction effect.

[0043] The temperature values and power supply voltage values in each historical temperature-power supply voltage combination are taken as two basic independent variables, and combined features including the interaction terms between the two independent variables are constructed, for example , , As new independent variables, these combined features and the residuals in the corresponding residual data set are substituted into a preset multiple linear equation containing multiple interaction terms , and again using the calculation steps shown in step S022, a set of parameters [g, h, i] that minimize the fitting error of the equation to the residuals is calculated by the least squares method. This set of parameters is the third set of model coefficients for describing the coupling interaction between temperature and power supply voltage. The main effect achieved by doing this is to quantitatively separate and capture the additional nonlinear effects produced by the combined action of temperature and voltage changes beyond the simple superposition of their independent effects, thereby enabling the final drift model to more realistically and completely reflect the comprehensive effects under the combined action of complex environmental factors, significantly improving the prediction accuracy of the model under varying conditions.

[0044] wherein is the predicted residual value.

[0045] Step S025, merging the first group of model coefficients, the second group of model coefficients and the third group of model coefficients to obtain the model fitting parameters.

[0046] The first group of model coefficients, the second group of model coefficients and the third group of model coefficients respectively describing the temperature independent influence, the voltage independent influence and the temperature-voltage interaction influence are integrated into a polynomial mathematical expression with temperature and power supply voltage as independent variables according to the mathematical terms they correspond to. Specifically, the unified expression form is:

[0047] Therefore, the model fitting parameters obtained by merging are {a, b, c, d, e, f, g, h, i}, that is, the model fitting parameters. The temperature influence coefficient, the voltage influence coefficient and their coupled interaction influence coefficient identified by decoupling in the sub-steps are integrated into a complete expression. The expression can simultaneously output the zero drift prediction value caused by temperature, voltage and their complex interaction, thereby laying a directly applicable parameter foundation for subsequent construction of a comprehensive compensation model covering all core environmental factors.

[0048] Step S03, on the basis of the model fitting parameters, introducing the historical cumulative running time length at the same time stamp as each historical temperature-power supply voltage combination to construct a multi-factor coupled compensation function, and taking the multi-factor coupled compensation function as a preset multi-factor coupled drift model.

[0049] On the basis of the expression established by step S02, the historical cumulative running time length representing the cumulative effect of rudder mechanical wear is further introduced as a third dimension key variable, thereby constructing a complete compensation function capable of simultaneously depicting the joint action of environmental factors and time aging factors. In this way, the original static two-dimensional model describing only instantaneous environmental influence is extended to a dynamic three-dimensional model capable of describing long-term performance degradation. By integrating the cumulative running time length in the form of power term and interaction term with other variables, the multi-factor coupled compensation function can simultaneously respond to the instantaneous changes of temperature and voltage and the gradual mechanical deviation accumulated over time due to gear wear, lubricant performance degradation, etc. This fundamentally solves the key defect that the prior art cannot compensate for the zero drift caused by long-term running wear, so that the finally constructed preset multi-factor coupled drift model has the ability to integrally and continuously compensate for environmental changes and self performance degradation in the whole life cycle of the rudder, significantly improving the long-term adaptability and reliability of the compensation system.

[0050] Further, step S03 further includes steps S031-S034: Step S031, combining the first group of model coefficients, the second group of model coefficients and the third group of model coefficients with each historical temperature-power supply voltage combination, to obtain a first predicted drift component related to temperature and voltage in the zero drift data.

[0051] For each historical temperature-power supply voltage combination corresponding to a calibration condition, the temperature value and the power supply voltage value in each calibration condition are substituted into the expression obtained by combining the expressions in step S025, and the calculation result is obtained by executing the polynomial calculation including the temperature term, the voltage term and the interaction term. The first predicted drift component related to temperature and voltage calculated for the corresponding calibration condition is separated and extracted from the total zero drift data measured in the experiment by using the established expression, so as to provide a data basis after removing the environmental variable interference for the subsequent step of separately analyzing and fitting the influence of factors such as mechanical wear represented by the cumulative running time.

[0052] Step S032, performing power transformation on each historical cumulative running time to obtain a plurality of running time characteristic terms of different powers.

[0053] For each recorded historical cumulative running time t, t itself is taken as a first-order term (i.e. ), and the square value of t is taken as a second-order term (i.e. ), and the cube value of t is taken as a third-order term (i.e. ), and so on, so as to generate a group of running time characteristic terms of different powers for each historical cumulative running time t, i.e. , , , so as to expand the single running time variable into a series of nonlinear basis features with different mathematical powers, thereby providing a mathematical expression capability for describing the possible complex nonlinear relationship between the long-term effects such as mechanical wear and material fatigue and the zero drift, so that the final multi-factor coupled drift model can more flexibly and accurately fit the possible nonlinear change trend of the drift with the running time, such as acceleration, deceleration or saturation, instead of being limited to a simple linear assumption.

[0054] Step S033, performing multiple regression fitting according to the zero drift data, the first predicted drift component and the plurality of running time characteristic terms to obtain a fourth group of model coefficients for describing the nonlinear relationship between the zero drift and the running time.

[0055] The zero drift data corresponding to each calibration condition is , and the first predicted drift component of the same calibration condition calculated in step S031 is , to obtain the residual error which mainly contains the drift component contributed by the accumulated running time and random noise, and then, the multiple running time characteristic terms generated in step S032 are taken as independent variables, and the calculated residual error is taken as dependent variable, to substitute into the preset multiple linear equation about running time , in which each running time characteristic term is taken as a row element of design matrix X, and all corresponding residual errors are combined into observation vector Y, so as to construct the linear equation group , in which, is the parameter vector to be solved , by solving the equation , the least square estimation value of the parameter vector to be solved is calculated, that is , so as to obtain a set of parameters [j, k, l], so that the calculation result of the multiple linear equation about running time is the minimum overall fitting error of the residual error , and the parameters [j, k, l] are the fourth set of model coefficients for describing the nonlinear relationship between the zero drift and the running time.

[0056] By extracting and quantifying the complex nonlinear law of the change of the zero drift data with the accumulation of running time after stripping the environmental influences of temperature and voltage, the drift characteristics caused by the time accumulation effects such as mechanical wear and material fatigue are described, which provides indispensable key parameters for finally constructing a multi-factor coupling compensation model covering both environmental factors and time aging factors.

[0057] In step S034, based on the model fitting parameters and the fourth set of model coefficients, a multi-factor coupling compensation function including the temperature first-order term, the voltage first-order term, the running time power term, the temperature-voltage interaction term, the temperature-time interaction term and the voltage-running time interaction term is constructed.

[0058] All mathematical terms corresponding to the first set of model coefficients [a, b, c], the second set of model coefficients [d, e, f] and the third set of model coefficients [g, h, i] in the model fitting parameters (i.e. the temperature first-order term and its high-order terms, the voltage first-order term and its high-order terms, and the temperature-voltage interaction term), the mathematical term corresponding to the fourth set of model coefficients [j, k, l] (i.e. the running time power term), and the newly introduced interaction terms (such as the product term mTt of temperature and running time, the product term nVt of voltage and running time, etc., whose coefficients m and n are obtained by fitting together with the fourth set of model coefficients or as an extension part thereof) are combined and added in algebra, so as to form the multi-factor coupling compensation function as shown below:

[0059] The multi-factor coupling compensation function enables the real-time input of temperature, voltage and running time to output a zero drift amount that comprehensively reflects the combined effect of environmental instantaneous changes and mechanical long-term wear in a single calculation, thereby providing a complete calculation core that is comprehensive, accurate and has clear physical meaning for subsequent online compensation.

[0060] wherein, is the predicted output value of the multi-factor coupling drift model.

[0061] In a feasible implementation, referring to Figure 3 , the specific implementation of the calculation module includes steps S11-S13: Step S11, calculating a temperature drift component according to the current environmental temperature and the first set of model coefficients, calculating a voltage drift component according to the current power supply voltage and the second set of model coefficients, and calculating a time length drift component according to the cumulative running time and the third set of model coefficients.

[0062] Substitute the real-time collected current environmental temperature T into the polynomial expression determined by the first set of model coefficients, i.e., calculate wherein a, b and c are specific parameters in the first set of model coefficients, thereby obtaining the temperature drift component quantified by the current environmental temperature alone.

[0063] Substitute the real-time collected current power supply voltage V into the polynomial expression determined by the second set of model coefficients, i.e., calculate wherein d, e and f are specific parameters in the second set of model coefficients, thereby obtaining the voltage drift component quantified by the current power supply voltage alone.

[0064] Substitute the real-time collected cumulative running time t into the polynomial expression determined by the third set of model coefficients, i.e., calculate wherein j, k and l are specific parameters in the third set of model coefficients, thereby obtaining the time length drift component quantified by the mechanical wear and other factors represented by the cumulative running time alone.

[0065] By decomposing the complex multi-factor coupling calculation into multiple simple single calculations that can be executed in parallel or in quick sequence, the complexity of online real-time calculation and the instantaneous demand for microprocessor operation resources are reduced, enabling efficient and reliable completion of the rapid evaluation of the drift caused by each independent factor in the system, thereby providing immediate and accurate component input for subsequent comprehensive summation and real-time compensation.

[0066] Step S12, calculating the coupling drift component by performing interaction term calculation on the current environmental temperature, the current power supply voltage, the cumulative running time and the fourth set of model coefficients.

[0067] The real-time collected ambient temperature, current power supply voltage, and cumulative running time are substituted into the mathematical expressions of the interaction terms corresponding to the coefficients of the fourth set of models and calculated, i.e., the temperature-voltage interaction term. Temperature-duration interaction item and voltage-duration interaction terms Where g, h, i, m, and n are specific parameters in the fourth group of model coefficients. By algebraically summing the calculation results of all interaction terms, the coupled drift component is obtained. This achieves a specific targeting and quantification of the nonlinear additional drift effect generated by the mutual coupling and influence between the three core factors of temperature, voltage, and operating time. This effect cannot be explained by the independent action of any single factor. By accurately calculating this coupled component and superimposing it on each independent component, the final multi-factor coupled drift model can more realistically and completely reflect the comprehensive drift behavior under actual complex working conditions in principle, thereby fundamentally improving the comprehensive prediction accuracy of the system when facing intertwined changes in multiple variables.

[0068] Step S13: Summate the temperature drift component, voltage drift component, duration drift component, and coupling drift component to obtain the initial zero-position drift amount.

[0069] The calculated temperature drift component Voltage drift component Duration drift component The initial zero-point drift is obtained by adding the calculated coupled drift component to the initial zero-point drift. This allows the independent and coupled factors, which are evaluated separately, to be integrated into a total drift prediction value with clear physical meaning through a linear superposition. This completes the key transformation from multi-factor analysis to single compensation value output, providing a basis for deviation in subsequent real-time compensation control and ensuring the accuracy and consistency of compensation operations.

[0070] In one feasible implementation, refer to Figure 4 As shown, the specific implementation of the update module includes steps S21 to S23: Step S21: Based on the current control command and current feedback position of the servo motor, calculate the current zero-position drift amount, and then calculate the difference between the current zero-position drift amount and the initial zero-position drift amount to obtain the drift prediction error.

[0071] The difference between the set value corresponding to the current control instruction representing the theoretical zero position currently received by the steering engine and the value of the current feedback position representing the actual position obtained by feedback of the internal potentiometer of the steering engine and analog-digital conversion and filtering processing is the current zero position drift. Then, the difference between the calculated current zero position drift and the initial zero position drift is the drift prediction error, which generates information reflecting the difference between the prediction accuracy of the direct quantization model and the actual system deviation, and reflects the inaccuracy of the model prediction under the current specific working condition, providing feedback input for subsequent steps of adjusting and optimizing the model parameters.

[0072] In step S22, based on the drift prediction error, the current environment temperature, the current power supply voltage, the cumulative running time, and the model parameters of the preset multi-factor coupled drift model, recursive estimation is performed to obtain the current update increment of the model parameters.

[0073] The current environment temperature T, the current power supply voltage V, and the cumulative running time t are taken as input vectors, and the partial derivative of the predicted output value of the preset multi-factor coupled drift model with respect to each model parameter (such as coefficients a, b, d, e, etc.) in the mathematical expression of the preset multi-factor coupled drift model is calculated under the current input vector. These partial derivatives collectively form a row vector, which is the sensitivity vector of the preset multi-factor coupled drift model to parameter changes at the current working point.

[0074] Specifically, for:

[0075] The partial derivative with respect to a is The partial derivative with respect to b is The partial derivative with respect to d is The partial derivative with respect to g is and so on. Then, the current environment temperature T, the current power supply voltage V, and the cumulative running time t collected at the current time are substituted into each of the derived partial derivative functions to calculate the specific value of each partial derivative at the current working condition. Finally, the series of calculated specific values are arranged in accordance with the consistent order of the model parameters, i.e.

[0076] combined into a row vector, which is the sensitivity vector of the preset multi-factor coupled drift model under the current input vector [T, V, t].

[0077] Subsequently, a gain matrix is dynamically maintained by the recursive least square method, which is updated in the algorithm iteration to reflect the uncertainty of the model parameter estimation, the sensitivity vector is multiplied by the gain matrix, that is, the gain matrix is multiplied by the transpose of the sensitivity vector, and then divided by a normalization factor determined by the forgetting factor, the sensitivity vector and the gain matrix, an intermediate gain vector is calculated, and then the intermediate gain vector is multiplied by the drift prediction error, that is, each element of the intermediate gain vector is multiplied by the drift prediction error, to obtain a new sensitivity vector, each element in the new sensitivity vector corresponds to the current update increment of a model parameter in turn, so that the drift prediction error at the current time is distributed and converted into the correction of each model parameter in the multi-factor coupled drift model according to the specific working condition in which the drift prediction error occurs, thereby realizing the optimization of the model parameters based on the latest single observation data without re-performing global complex fitting, and providing a core update mechanism for the model to continuously track the real dynamic changes of the system.

[0078] It should be noted that the gain matrix is a key matrix dynamically maintained in the recursive least square method, and its mathematical essence is a normalized parameter estimation error covariance matrix, which is dynamically adjusted according to historical observation data and is used to quantify the degree of trust of the system to the current new input data. The gain matrix is updated by algorithm iteration, and the update rule is to calculate the gain matrix at the last time, the model sensitivity vector under the current input vector and a forgetting factor, so that the gain matrix can be adaptively adjusted. When the prediction ability of the preset multi-factor coupled drift model for the new working condition is uncertain, the gain matrix will increase to make the model parameters update by a large amplitude, and when the prediction tends to be stable, the gain matrix will decrease to avoid excessive adjustment of the model parameters. In the parameter update step, the core role of the gain matrix is to weight and distribute the current drift prediction error according to the direction indicated by the model sensitivity vector, and finally convert it into a set of current update increments. The forgetting factor is a preset constant between 0 and 1, and its core role is to realize the exponential weighted attenuation of the influence of historical data, thereby giving the algorithm the ability to track time-varying parameters.

[0079] In step S23, the model parameters and the current update increments are summed to update the model parameters, and an updated multi-factor coupled drift model is obtained.

[0080] Add all the model parameters constituting the preset multi-factor coupled drift model stored in the current system and the current update increment vector with the same dimension and sequence calculated element by element, thereby generating a set of updated model parameters. Replace the original model parameters with the updated model parameters, that is, complete the update of the preset multi-factor coupled drift model, obtain the updated multi-factor coupled drift model, and accumulate and fuse the parameter adjustment amount calculated based on the single observation error into the original model, thereby enabling the mathematical expression of the model to continuously and smoothly approximate and track the real slow time-varying process of the drift characteristics of the rudder system caused by long-term wear, component aging or unmodeled factors, and ultimately ensure that the compensation system maintains high precision throughout the life cycle.

[0081] It should be noted that the current real-time collected current environment temperature, current power voltage and cumulative running time are used as inputs and substituted into the updated multi-factor coupled drift model for calculation. The calculation process is consistent with the method described in steps S11 to S13, but the updated model parameters are used, thereby outputting a corrected zero drift quantity that is more consistent with the real state of the current system. According to the corrected zero drift quantity, the control command of the rudder is corrected, and the specific operation is as follows: superimpose a compensation value equal in size and opposite in direction to the corrected zero drift quantity on the original planned control command value (i.e. the theoretical position command expected to be reached by the rudder), thereby generating a new control command that has been pre-compensated, and converting the corrected control command into a corresponding PWM signal output to drive the rudder to move, thereby actively offsetting the zero deviation caused by environmental factors, long-term wear and their coupling effects before the rudder actually responds, and ultimately enabling the actual running position of the rudder to continuously and stably track the expected theoretical position command accurately.

[0082] The above are preferred embodiments of the present application, and do not limit the protection scope of the present application, therefore: any equivalent changes made in structure, shape, principle, etc. according to the present application should be covered within the protection scope of the present application.

Claims

1. A servo motor zero-position drift online compensation system, characterized in that, include: The calculation module is used to calculate the initial zero-position drift amount based on the current ambient temperature, current power supply voltage and cumulative running time of the servo motor through a preset multi-factor coupled drift model. The update module is used to update the model parameters of the preset multi-factor coupled drift model based on the initial zero-position drift amount and the current feedback position of the servo motor, so as to obtain the updated multi-factor coupled drift model. The calculation module is further configured to calculate the corrected zero-point drift amount based on the updated multi-factor coupled drift model, and then correct the control command of the servo motor according to the corrected zero-point drift amount to drive the servo motor to move.

2. The servo motor zero-position drift online compensation system according to claim 1, characterized in that, Before the step of calculating the initial zero-position drift amount based on the current ambient temperature, current power supply voltage, and cumulative operating time of the servo motor using a preset multi-factor coupled drift model, the following steps are included: Under various historical temperature-power supply voltage combinations, the feedback position information of the servo motor under the zero-position control command is obtained. Based on the difference between the feedback position information and the zero-position control command, the zero-position drift data under each calibration condition is calculated. The historical temperature-power supply voltage combination and zero drift data under each calibration condition are fitted to obtain model fitting parameters that describe the relationship between zero drift, temperature and power supply voltage. Based on the model fitting parameters, the historical cumulative running time at the same timestamp as each of the historical temperature-power supply voltage combinations is introduced to construct a multi-factor coupling compensation function, and the multi-factor coupling compensation function is used as the preset multi-factor coupling drift model.

3. The servo motor zero-position drift online compensation system according to claim 2, characterized in that, The step of obtaining the feedback position information of the servo motor under the zero-position control command under various historical temperature-power supply voltage combinations during calibration includes: Receive the analog feedback signal returned by the servo under each of the calibration conditions according to the zero-position control command, perform analog-to-digital conversion sampling on the analog feedback signal, and obtain the sampled feedback value; The sampled feedback values ​​are subjected to statistical filtering to obtain steady-state feedback values ​​that characterize the position of the servo motor under the corresponding calibration conditions. The steady-state feedback value is correlated with the historical temperature-power supply voltage combination under the corresponding calibration conditions to generate the feedback location information.

4. The servo motor zero-position drift online compensation system according to claim 3, characterized in that, The step of performing fitting processing based on the historical temperature-power supply voltage combination and zero-point drift data under each of the calibration conditions to obtain model fitting parameters for describing the relationship between zero-point drift, temperature, and power supply voltage includes: Data separation was performed on the historical temperature-power supply voltage combination and zero drift data under each of the calibration conditions to obtain temperature influence dataset and voltage influence dataset. Based on the temperature influence dataset, a univariate nonlinear regression fitting is performed to obtain a first set of model coefficients describing the relationship between zero drift and temperature; and based on the voltage influence dataset, a univariate nonlinear regression fitting is performed to obtain a second set of model coefficients describing the relationship between zero drift and voltage. Based on the first set of model coefficients and the second set of model coefficients, the predicted drift value corresponding to each of the historical temperature-power supply voltage combinations is calculated, and the residual dataset is calculated based on the difference between each of the predicted drift values ​​and the zero drift data corresponding to each of the predicted drift values. Based on the residual dataset and each of the historical temperature-power supply voltage combinations, a multivariate regression fitting is performed to obtain a third set of model coefficients to describe the interactive effects. The model coefficients of the first group, the second group, and the third group are combined to obtain the model fitting parameters.

5. The servo motor zero-position drift online compensation system according to claim 4, characterized in that, The step of constructing a multi-factor coupled compensation function by introducing the historical cumulative runtime at the same timestamp as each of the historical temperature-power supply voltage combinations, based on the model fitting parameters, includes: By combining the first set of model coefficients, the second set of model coefficients, and the third set of model coefficients with each of the historical temperature-power supply voltage combinations, the first predicted drift component related to temperature and voltage in the zero-point drift data is calculated. The historical cumulative runtime is transformed by a power factor to obtain multiple runtime feature terms with different powers. Based on the zero-position drift data, the first predicted drift component, and multiple runtime feature terms, a multivariate regression fitting is performed to obtain a fourth set of model coefficients to describe the nonlinear relationship between zero-position drift and runtime. Based on the model fitting parameters and the fourth set of model coefficients, a multi-factor coupling compensation function is constructed, including a first-order temperature term, a first-order voltage term, a power term for runtime, a temperature-voltage interaction term, a temperature-duration interaction term, and a voltage-runtime interaction term.

6. The servo motor zero-position drift online compensation system according to claim 1, characterized in that, The preset multi-factor coupled drift model includes a first set of model coefficients, a second set of model coefficients, a third set of model coefficients, and a fourth set of model coefficients. The step of calculating the initial zero-position drift amount based on the current ambient temperature, current power supply voltage, and cumulative operating time of the servo motor through the preset multi-factor coupled drift model includes: The temperature drift component is calculated based on the current ambient temperature and the first set of model coefficients; the voltage drift component is calculated based on the current power supply voltage and the second set of model coefficients; and the duration drift component is calculated based on the cumulative running time and the third set of model coefficients. The coupling drift component is obtained by performing an interaction term calculation on the current ambient temperature, the current power supply voltage, the cumulative running time, and the fourth set of model coefficients. The initial zero-point drift is obtained by summing the temperature drift component, the voltage drift component, the duration drift component, and the coupling drift component.

7. The servo motor zero-position drift online compensation system according to claim 1, characterized in that, The step of updating the model parameters of the preset multi-factor coupled drift model based on the initial zero-position drift amount and the current feedback position of the servo motor to obtain the updated multi-factor coupled drift model includes: Based on the current control command of the servo and the current feedback position, the current zero-position drift is calculated, and the drift prediction error is obtained by calculating the difference between the current zero-position drift and the initial zero-position drift. Based on the drift prediction error, the current ambient temperature, the current power supply voltage, the cumulative running time, and the model parameters of the preset multi-factor coupled drift model, a recursive estimation is performed to obtain the current update increment of the model parameters; The model parameters and the current update increment are summed to update the model parameters, resulting in the updated multi-factor coupled drift model.

Images