Electric steering engine feedback anomaly detection and suppression method based on adaptive confidence coefficient

Abstract

The invention provides an electric steering engine feedback anomaly detection and suppression method based on adaptive confidence, and the method is characterized in that the method comprises the following steps: building an electric steering engine feedback prediction model, and predicting a feedback sampling value at the next moment; determining model parameters through a steering engine position instruction frequency sweep test method; in each control period, calculating an instantaneous error; updating an error weight adjustment factor; constructing a confidence evaluation mechanism, and evaluating the reliability of the sampling signal according to the confidence; generating a system feedback true value according to the confidence coefficient of the feedback sampling value at the current moment; and calculating the current error between the feedback truth value and the feedback sampling value of the system, stabilizing the current error in a reasonable range, and updating the recursive weighted average error. The method gets rid of dependence on a fixed discrimination threshold, and has higher adaptive capacity; the continuous suppression of abnormal jump is realized, and the completeness and robustness of system feedback are enhanced; the method has a good fault-tolerant capability and shows a strong general adaptive capability.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to a method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence. Background Technology

[0002] In an electric servo feedback control system, an angle sensor detects the deflection angle of the control surface in real time and converts it into an electrical signal, providing position feedback information to the control system. This drives the actuator to dynamically adjust based on the deviation between the target command and the measured angle, achieving position tracking and stable control of the servo system. However, angle sensors are susceptible to electromagnetic noise interference and temperature drift effects in signal processing circuits, making them prone to feedback anomalies. Feedback anomalies can cause deviations between the actual deflection angle of the control surface and the control command, leading to problems such as loss of flight attitude control and flight path deviation.

[0003] Existing feedback anomaly detection methods are mainly based on discriminant inequalities. These methods determine whether the current sampled value exceeds the "normal range" by setting certain threshold conditions, and then discard or correct the feedback signal. Common discriminant inequalities include feedback rate constraints, steady-state error tolerance judgments, and amplitude range limitations. While discriminant inequality-based feedback anomaly detection methods are effective in handling isolated and sporadic anomalies, they have significant shortcomings in practical engineering applications, including the following:

[0004] On the one hand, the detection accuracy of the discrimination inequality method depends on the threshold setting. However, the selection of the threshold requires a trade-off between the "anomaly detection rate" and the "false alarm rate." If the threshold is set too loosely, outliers are easily missed; conversely, if it is set too strictly, normal fluctuations may be misjudged as anomalies, resulting in unnecessary data rejection or control disturbances. In addition, actual feedback noise often exhibits non-constant variation, and a fixed threshold is difficult to adapt to the dynamic characteristics under different operating conditions, leading to unstable detection accuracy.

[0005] On the other hand, traditional methods often employ a "hard-determination" strategy, in which a sampled signal is completely discarded once it is determined to be an abnormal feedback signal. This approach lacks a smooth transition and is prone to losing the original signal trend during continuous discarding, making it difficult to recover the system's true value information and further affecting the control loop closure. Especially during high-frequency feedback change phases, excessive discarding can lead to a severe misalignment between servo control commands and the actual state, reducing system response speed and accuracy. Summary of the Invention

[0006] The technical problem to be solved by this invention is that traditional inequality discrimination methods have insufficient ability to detect continuous anomalies and lack robustness to dynamic changes in the system.

[0007] To address the aforementioned technical problems, this invention discloses a method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence, characterized by comprising the following steps:

[0008] Step 1: Based on the dynamic model, establish an electric servo motor feedback prediction model; using this model, the feedback sample value for the next moment can be predicted based on the current control input and the state at the previous moment.

[0009] Step 2: Determine the relevant parameters of the model by using the servo position command frequency sweep test method, and set the initial recursive weighted average error value, error weight adjustment factor, and upper and lower bounds of error distribution, etc.

[0010] Step 3: In each control cycle, the system feedback sampling signal, i.e., the feedback sampling value, is obtained through the position sensor. Based on the feedback prediction model, the system predicted feedback value, i.e., the predicted value, is calculated, and the instantaneous error is calculated.

[0011] Step 4: Update the error weight adjustment factor based on the ratio of the difference between the system feedback sample value, the predicted value, and the system input command;

[0012] Step 5: Construct a confidence evaluation mechanism. Based on the cumulative density function (CDF) of the standard normal distribution, calculate the confidence of the current feedback sampling signal, and evaluate the reliability of the sampling signal based on the confidence.

[0013] Step 6: Based on the confidence level of the current feedback sample value, the model prediction value and the actual sample value are weighted and fused to generate the system feedback true value, which then replaces the system sample value and enters the closed-loop control of the system.

[0014] Step 7: Calculate the current error between the system feedback true value and the feedback sample value, and stabilize the current error within a reasonable range by limiting the error amplitude; and use a dynamic weighting mechanism to update the recursive weighted average error by combining the current error with historical error information.

[0015] Furthermore, the implementation steps for establishing the electric servo motor feedback prediction model in step 1 include:

[0016] Step 1.1: Establish the voltage balance equation, which is as follows:

[0017]

[0018] Where u(t) is the armature voltage, R is the total armature resistance, L is the total armature inductance, i(t) is the armature current, and E A It is the back electromotive force;

[0019] Step 1.2: Based on Step 1.1, establish the dynamic model of the electric servo motor. The dynamic model is as follows:

[0020]

[0021] Among them, K B θ is the motor torque coefficient, J is the moment of inertia, θ(t) is the actual output angle of the servo motor, and B is the linear damping coefficient.

[0022] Step 1.3: Calculate and correct the angular acceleration using the discretized dynamic model. The angular acceleration calculation method is as follows:

[0023] acc model (k)=(K B i(k)-B·ω(k)) / J

[0024] acc(k) = acc model (k)+Δa(k)

[0025] Δa(k)=δ(k)+Δa(k-1)

[0026] δ(k)=λ·δ(k-1)+(1-λ)[K D ·e(k) / T]

[0027] Among them, acc model (k) is the theoretical value of the servo motor's angular acceleration, acc(k) is the corrected servo motor angular acceleration, Δa(k) is the angular acceleration correction amount, T is the control cycle of the servo motor system, and K D To correct for the gain, λ is the scaling factor;

[0028] Step 1.4: Update the angular velocity and perform smoothing. The angular velocity is calculated as follows:

[0029] ω(k)=ω(k-1)+T·acc(k)

[0030] ω fliter (k)=α·ω fliter (k-1)+(1-α)·ω(k)

[0031] Where ω(k) is the theoretical value of the servo angular velocity, ω(k-1) is the servo angular velocity of the previous cycle, and ω fliter (k) represents the smoothed servo angular velocity, ω fliter (k-1) is the servo angular velocity after smoothing in the previous cycle, and α is the weighting coefficient;

[0032] Step 1.5: Establish a feedback prediction model. The feedback prediction value is calculated as follows:

[0033]

[0034] Among them, y predict (k) represents the feedback prediction value, ytrue (k-1) represents the true value of the feedback at the previous time step, ω fliter (k) is the smoothed servo angular velocity in step 1.4, and acc(k) is the corrected servo angular acceleration in step 1.3.

[0035] Furthermore, the implementation steps of the frequency sweep test method for determining model parameters in step 2 include:

[0036] Step 2.1: Within the servo motor's stable control range, input a series of small-amplitude sine or step command signals of different frequencies;

[0037] Step 2.2: Record the corresponding servo position feedback response;

[0038] Step 2.3: Extract parameters using curve fitting methods based on frequency domain or time domain data.

[0039] Furthermore, in step 3, the feedback sampling value y of the current control cycle sample (k) and predicted value y predict (k), calculate the instantaneous error err(k), the instantaneous error err(k) is:

[0040] err(k)=y predict (k)-y sample (k).

[0041] Furthermore, the formula for updating the error weight adjustment factor in step 4 is as follows:

[0042]

[0043] Among them, e weaken e is the error weight adjustment factor. norm y is the ratio of the difference between the feedback sample value, the predicted value, and the input command position. sample (k) is the feedback sample value in step 3, y predict (k) is the predicted value in step 3, and u(k) is the system input command position.

[0044] Furthermore, the confidence level calculation formula in step 5 is as follows:

[0045] β = 1 - 2|cdf - 0.5|

[0046]

[0047] Where β is the confidence level, cdf is the cumulative density of the current error, σ is the recursive weighted average error described in step (7), F(x; μ, σ) is the cumulative distribution function of the standard normal distribution, used to predict the probability of feedback error; μ represents the mean parameter, and in this prediction model, the mean error is 0; φ represents the standard deviation parameter, which defines the degree of dispersion of the prediction error distribution; erf(z) is the error function, z represents the input variable, and t is the dummy variable in the definite integral operation, with its integration range from 0 to z.

[0048] Furthermore, the formula for calculating the feedback truth value in step 6 is as follows:

[0049] y true (k)=β·y sample (k)+(1-β)·y predict (k)

[0050] Among them, y true (k) represents the true feedback value for the current period, β represents the confidence level in step 5, and y sample (k) represents the feedback sample value for the current period, y predict (k) represents the feedback prediction value for the current period.

[0051] Furthermore, the formula for calculating the recursive weighted average error in step 7 is as follows:

[0052] σ(k)=μ·σ(k-1)+(1-μ)(y predict (k)-y sample (k))

[0053]

[0054] Where σ(k) is the recursive weighted average error, μ is the weighting factor, and σ max σ is the upper limit of the error. min This is the lower limit of the error.

[0055] Compared with the prior art, the present invention has at least the following beneficial effects:

[0056] First, a second-order linear electric servo motor model was established to predict feedback values ​​online. An error distribution model was constructed by combining historical sampling data, and the degree of anomaly of the current feedback sampling value was estimated based on the normality assumption, thereby getting rid of the dependence on a fixed discrimination threshold and having stronger adaptive capabilities.

[0057] Second, a confidence-weighted mechanism was designed to fuse model predictions with current sampled values ​​to generate more reliable true feedback values. When anomalies occur, this invention can dynamically reduce the weight of anomalous data, achieving continuous suppression of anomalous jumps, thereby enhancing the integrity and robustness of the system feedback.

[0058] Third, this invention has low dependence on model accuracy and good fault tolerance. Furthermore, it can effectively identify and mitigate abnormal feedback even when the model is dynamically changing or the system experiences nonlinear disturbances, demonstrating strong general adaptability. Attached Figure Description

[0059] Figure 1 The flowchart shows the method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence.

[0060] Figure 2 This is a schematic diagram of an electric servo motor architecture;

[0061] Figure 3 The implementation process of servo motor feedback prediction. Detailed Implementation

[0062] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings.

[0063] This invention relates to the field of automation control. To address the problems of insufficient continuous anomaly detection capability and lack of robustness to dynamic changes in the system by traditional inequality discrimination methods, this invention proposes an electric servo motor feedback anomaly detection and suppression method based on adaptive confidence, which is applicable to feedback signal anomaly detection and suppression in electric servo motor feedback control systems.

[0064] This method establishes a feedback prediction model for an electric servo system based on a dynamic model, enabling real-time prediction of position feedback. A confidence evaluation framework is constructed to assess the reliability of feedback sample values ​​in each control cycle using confidence levels. Based on a confidence-weighted fusion strategy, the true feedback value is calculated and used to replace the sampled feedback value before entering the electric servo system, completing closed-loop control. An error weight adjustment factor is designed; when the model prediction deviates from the command, increasing the error weight adjustment factor reduces the weight of the feedback prediction value, achieving adaptive adjustment of the feedback prediction model and ensuring the accuracy of the feedback prediction value. Compared with existing technologies, this invention has the advantages and effects of improving the identification and suppression capabilities of abnormal system signals without relying on prior feedback anomaly features. Furthermore, by designing a confidence evaluation framework and a confidence-weighted fusion strategy, the weight of abnormal data is dynamically reduced, achieving continuous suppression of abnormal jumps and enhancing the integrity and robustness of system feedback.

[0065] Figure 1 This is a flowchart of a method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence. Figure 1 As shown, the specific flow of the electric servo motor feedback anomaly detection and suppression method based on adaptive confidence degree of the present invention is as follows:

[0066] Step 1: Based on the dynamic model, establish an electric servo motor feedback prediction model; using this model, the feedback sample value at the next moment can be predicted based on the current control input and the state at the previous moment.

[0067] In one implementation method, the specific approach to establishing the electric servo motor feedback prediction model is as follows:

[0068] (1) Establish the voltage balance equation

[0069] The voltage balance equation is established based on the architecture of the electric servo motor. Figure 2 This is a schematic diagram of an electric servo motor architecture. Figure 2 As shown, the electric servo system mainly consists of a drive module, a reduction gear mechanism, and a position feedback unit. In the electric servo system, the voltage balance equation reveals the energy conversion mechanism of the armature circuit. This equation, by relating armature voltage, current, and back electromotive force, quantitatively describes the dynamic equilibrium relationship between electrical energy and mechanical energy. The voltage balance equation of the electric servo system is as follows:

[0070]

[0071] In the formula, u(t) is the armature voltage, R is the total armature resistance, L is the total armature inductance, i(t) is the armature current, and E... A It is the back electromotive force.

[0072] (2) Establish the dynamic model of the electric servo motor.

[0073] Assuming the dynamics are mainly affected by rotational inertia and linear damping, the dynamic equations of the electric servo motor control system are established based on Newton's second law. The dynamic equations are as follows:

[0074]

[0075] In the formula, K B Let J be the motor torque coefficient, θ(t) be the moment of inertia, θ(t) be the actual output angle of the servo motor, i(t) be the armature current, and B be the linear damping coefficient. For the servo motor angular velocity, This represents the angular acceleration of the servo motor.

[0076] (3) Angular acceleration update and correction

[0077] Discretizing equation (2) yields the theoretical value of the servo motor's angular acceleration, acc. model (k) is:

[0078] acc model (k)=(K B i(k)-B·ω(k)) / J (3) accmodel (k) is the theoretical value of the angular acceleration of the servo motor, and ω(k) is the angular velocity of the servo motor.

[0079] Since the actual value of the moment of inertia J often varies with load changes, causing it to deviate from the theoretical value, it is necessary to correct the angular acceleration to ensure the accuracy of the prediction model. The corrected servo angular acceleration acc(k) is:

[0080] acc(k) = acc model (k)+Δa(k) (4)

[0081] In the formula, Δa(k) is the angular acceleration correction.

[0082] Within each sampling period, an angular acceleration correction is designed using the real-time position error. The angular acceleration correction is as follows:

[0083] Δa(k)=δ(k)+Δa(k-1) (5)

[0084] δ(k)=λ·δ(k-1)+(1-λ)[K D ·e(k) / T] (6)

[0085] In the formula, T is the control cycle of the servo system, e(k) is the error at the current moment, and K... D To correct the gain, which is used to balance the system's response speed and stability; λ is the proportional coefficient, which can suppress noise and sudden disturbances through exponential weighted averaging. δ(k) is the instantaneous change of error in the current feedback cycle, δ(k-1) is the instantaneous change of error in the previous cycle, and Δa(k-1) represents the angular acceleration correction in the previous cycle.

[0086] (4) Angular velocity update and smoothing

[0087] Calculate the angular velocity of the electric servo motor based on its angular acceleration. The angular velocity is:

[0088] ω(k)=ω(k-1)+T·acc(k) (7)

[0089] In the formula, ω(k) is the theoretical value of the servo angular velocity, ω(k-1) is the servo angular velocity of the previous cycle, and T is...

[0090] To prevent abrupt changes in angular acceleration due to disturbances, the angular velocity needs to be smoothed. The smoothed servo angular velocity is as follows:

[0091] ω fliter (k)=α·ω fliter (k-1)+(1-α)·ω(k) (8)

[0092] In the formula, ω fliter(k) represents the smoothed servo angular velocity, ω fliter (k-1) represents the servo angular velocity after smoothing in the previous cycle, and α is a weighting coefficient used to measure the relative importance of the current predicted angular velocity and the historical angular velocity.

[0093] (5) Establish a feedback prediction model

[0094] Based on the previous position, estimated angular acceleration, and estimated angular velocity of the electric servo motor, a feedback prediction model is established to predict the current position of the servo motor. The feedback prediction model is as follows:

[0095]

[0096] Among them, y predict (k) represents the feedback prediction value, y true (k-1) represents the true value of the feedback at the previous time step, ω fliter (k) is the smoothed servo angular velocity in (4), and acc(k) is the corrected servo angular acceleration in (3).

[0097] Step 2: Determine the model parameters by using the servo position command frequency sweep test method, and set parameters such as the initial recursive weighted average error value, error weight adjustment factor, and upper and lower bounds of error distribution.

[0098] In one implementation, the specific implementation of the servo position command sweep frequency test method for determining the parameters of the feedback prediction model is as follows: First, within the stable control range of the servo, a series of small-amplitude sinusoidal or step command signals of different frequencies are input; then, the corresponding servo position feedback response is recorded; finally, the parameters are extracted using a curve fitting method with frequency domain or time domain data.

[0099] Step 3: In each control cycle, acquire the system feedback sampling signal through the position sensor; calculate the system predicted feedback value based on the feedback prediction model, and calculate the instantaneous error.

[0100] Figure 3 This describes the implementation process of servo motor feedback prediction. In one implementation, based on the electric servo motor feedback prediction model, the predicted servo motor feedback value is obtained. The specific implementation process is as follows: Figure 3 As shown. First, the servo angular acceleration at the current moment is determined based on the input command and the angular acceleration at the previous moment; then, the angular velocity of the servo is calculated according to equations (7) and (8) and smoothed; finally, the predicted value of the servo position at the current moment is calculated according to equation (9).

[0101] In one implementation, based on the feedback sample value y of the current control cycle sample (k) and predicted value y predict(k), calculate the instantaneous error err(k), the instantaneous error err(k) is:

[0102] err(k)=y predict (k)-y sample (k) (10)

[0103] Step 4: Update the error weight adjustment factor based on the ratio of the difference between the system feedback sample value, the predicted value and the system input command position.

[0104] In one implementation, the error weight adjustment factor is updated based on the accuracy of the feedback prediction. Typically, the error weight adjustment factor e... weaken If the system output sample value is closer to the system input command than the predicted value, then the feedback predicted value is considered to have a large deviation, and e should be increased. weaken To increase the weight of the feedback sample value, the formula for determining the error weight adjustment factor is:

[0105]

[0106] In the formula, e norm e is the ratio of the difference between the feedback sample value, the feedback predicted value, and the system input command position. It is used to measure the closeness of the sampled value, the predicted value, and the system input command. norm for:

[0107]

[0108] y sample (k) is the feedback sample value in step 3, y predict (k) is the predicted value in step 3, and u(k) is the system input command position.

[0109] Step 5: Construct a confidence evaluation mechanism. Calculate the confidence level of the current feedback sampling signal based on the cumulative density function (CDF) of the standard normal distribution. Evaluate the reliability of the sampling signal based on the confidence level.

[0110] In one implementation, confidence is assessed using the cumulative density function (CDF) of the standard normal distribution, based on the recursive weighted average error and the instantaneous error.

[0111] The CDF function is:

[0112]

[0113] Wherein, cdf is the cumulative density of the current error, σ is the recursive weighted average error described in step (7); F(x; μ,σ) is the cumulative distribution function of the standard normal distribution, used to predict the probability of feedback error; μ represents the mean parameter, in this prediction model, the mean error is 0; φ represents the standard deviation parameter, which defines the degree of dispersion of the prediction error distribution.

[0114] The erf(z) function is the error function, and its output range is (-1, 1). In low-performance computing platforms (such as the DSP F2812), it can be calculated using a lookup table method or an approximation function method. In one implementation, it is calculated using the following formula:

[0115]

[0116] Where z represents the input variable, and t is a dummy variable in the definite integral operation, with its integration range from 0 to z.

[0117] In this step, the confidence level of the feedback sampling signal is calculated based on the CFD function value of the instantaneous error. If the instantaneous error is far from the recursive weighted average error, i.e., the instantaneous error is located at the tail end of the distribution, the current feedback value is considered to have low confidence and is identified as abnormal feedback.

[0118] The confidence level β mentioned in this step is:

[0119] β=1-2|cdf-0.5| (16)

[0120] Step 6: Based on the confidence level of the current feedback sample value, perform weighted fusion of the model prediction value and the actual sample value to generate the system feedback true value, which then replaces the system sample value and enters the closed-loop control of the system.

[0121] In one implementation, to overcome the problem of signal loss and closed-loop control interruption that may result from directly removing abnormal data, a confidence-weighted fusion strategy is adopted. This strategy combines feedback sampled values ​​and predicted values ​​to calculate the true feedback value, which then replaces the feedback sampled values ​​and enters the closed-loop control of the servo control system. The core idea of ​​the confidence-weighted fusion strategy is to determine the weight of the feedback sampled values ​​based on the confidence level. When the confidence level is high, it indicates that the feedback sampled values ​​and predicted values ​​match well, and the sampled values ​​are tended to be used as the feedback signal. When the confidence level is low, the weight of the feedback sampled values ​​is reduced, and the true feedback value relies more on the model's predicted values, thereby smoothing out abnormal fluctuations. The formula for calculating the true feedback value is:

[0122] y true (k)=β·y sample (k)+(1-β)·y predict (k) (17)

[0123] Among them, y true(k) represents the true feedback value, β represents the confidence level in step 5, and y sample (k) represents the feedback sample value for the current period, y predict (k) represents the feedback prediction value for the current period.

[0124] Step 7: Calculate the current error between the system feedback true value and the feedback sample value, and stabilize the current error within a reasonable range by limiting the error amplitude; and use a dynamic weighting mechanism to update the recursive weighted average error by combining the current error with historical error information.

[0125] In one implementation, the recursive weighted average error is used as an indicator for error evaluation. It integrates current and historical error information through a dynamic weighting mechanism. The recursive weighted average error σ(k) is:

[0126] σ(k)=μ·σ(k-1)+(1-μ)(y predict (k)-y sample (k)) (18)

[0127] Where σ(k) is the recursive weighted average error and μ is the weighting factor.

[0128] In one implementation, error limiting ensures the effectiveness of the current error. Setting a lower error limit ensures that the method has a greater tolerance for model prediction errors when the actual error is small; setting an upper error limit prevents excessive contamination of error statistics when continuous, high-density, and large outliers occur. Even if anomalies persist for a long time, the method can still ensure that the system has a certain degree of suppression of anomalies. The error limiting formula is:

[0129]

[0130] Where, σ max σ is the upper limit of the error. min This is the lower limit of the error.

[0131] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

[0132] It should be noted that the above description is merely illustrative and explanatory of the present invention. Those skilled in the art should understand that any modifications and substitutions to the present invention fall within the scope of protection of the present invention.

Claims

1. A method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence, characterized in that, Includes the following steps: Step 1: Based on the dynamic model, establish an electric servo motor feedback prediction model; using this model, the feedback sample value for the next moment can be predicted based on the current control input and the state at the previous moment. Step 2: Determine the relevant parameters of the model by using the servo position command frequency sweep test method, and set the initial recursive weighted average error value, error weight adjustment factor, and upper and lower bounds of error distribution, etc. Step 3: In each control cycle, the system feedback sampling signal, i.e., the feedback sampling value, is obtained through the position sensor. Based on the feedback prediction model, the system predicted feedback value, i.e., the predicted value, is calculated, and the instantaneous error is calculated. Step 4: Update the error weight adjustment factor based on the ratio of the difference between the system feedback sample value, the predicted value, and the system input command; Step 5: Construct a confidence evaluation mechanism. Based on the cumulative density function (CDF) of the standard normal distribution, calculate the confidence of the current feedback sampling signal, and evaluate the reliability of the sampling signal based on the confidence. Step 6: Based on the confidence level of the current feedback sample value, the model prediction value and the actual sample value are weighted and fused to generate the system feedback true value, which then replaces the system sample value and enters the closed-loop control of the system. Step 7: Calculate the current error between the system feedback true value and the feedback sample value, and stabilize the current error within a reasonable range by limiting the error amplitude; and use a dynamic weighting mechanism to update the recursive weighted average error by combining the current error with historical error information.

2. The method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence as described in claim 1, characterized in that, The steps for establishing the electric servo motor feedback prediction model in step 1 include: Step 1.1: Establish the voltage balance equation, which is as follows: Where u(t) is the armature voltage, R is the total armature resistance, L is the total armature inductance, i(t) is the armature current, and E A It is the back electromotive force; Step 1.2: Based on Step 1.1, establish the dynamic model of the electric servo motor. The dynamic model is as follows: Among them, K B θ is the motor torque coefficient, J is the moment of inertia, θ(t) is the actual output angle of the servo motor, and B is the linear damping coefficient. Step 1.3: Calculate and correct the angular acceleration using the discretized dynamic model. The angular acceleration calculation method is as follows: acc model (k)=(K B i(k)-B ω(k)) / J acc(k)=acc model (k)+Δa(k) Δa(k)=δ(k)+Δa(k-1) δ(k)=λ·δ(k-1)+(1-λ)[K D ·e(k) / T] Among them, acc model (k) is the theoretical value of the servo motor's angular acceleration, acc(k) is the corrected servo motor angular acceleration, Δa(k) is the angular acceleration correction amount, T is the control cycle of the servo motor system, and K D To correct for the gain, λ is the scaling factor; Step 1.4: Update the angular velocity and perform smoothing. The angular velocity is calculated as follows: ω(k)=ω(k-1)+T·acc(k) oh fliter (k)=a·ω fliter (k-1)+(1-α)·ω(k) Where ω(k) is the theoretical value of the servo angular velocity, ω(k-1) is the servo angular velocity of the previous cycle, and ω fliter (k) represents the smoothed servo angular velocity, ω fliter (k-1) is the servo angular velocity after smoothing in the previous cycle, and α is the weighting coefficient; Step 1.5: Establish a feedback prediction model. The feedback prediction value is calculated as follows: Among them, y predict (k) represents the feedback prediction value, y true (k-1) represents the true value of the feedback at the previous time step, ω fliter (k) is the smoothed servo angular velocity in step 1.4, and acc(k) is the corrected servo angular acceleration in step 1.

3.

3. The method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence as described in claim 1, characterized in that, The implementation steps of the frequency sweep test method for determining model parameters in step 2 include: Step 2.1: Within the servo motor's stable control range, input a series of small-amplitude sine or step command signals of different frequencies; Step 2.2: Record the corresponding servo position feedback response; Step 2.3: Extract parameters using curve fitting methods based on frequency domain or time domain data.

4. The method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence as described in claim 1, characterized in that, In step 3, the feedback sampling value y of the current control cycle sample (k) and predicted value y predict (k), calculate the instantaneous error err(k), the instantaneous error err(k) is: err(k)=y predict (k)-y sample (k)。 5. The method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence as described in claim 1, characterized in that, The formula for updating the error weight adjustment factor in step 4 is as follows: Among them, e weaken e is the error weight adjustment factor. norm y is the ratio of the difference between the feedback sample value, the predicted value, and the input command position. sample (k) is the feedback sample value in step 3, y predict (k) is the predicted value in step 3, and u(k) is the system input command position.

6. The method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence as described in claim 1, characterized in that, The confidence level calculation formula in step 5 is as follows: β = 1 - 2|cdf - 0.5| Where β is the confidence level, cdf is the cumulative density of the current error, σ is the recursive weighted average error described in step (7), F(x; μ, σ) is the cumulative distribution function of the standard normal distribution, used to predict the probability of feedback error; μ represents the mean parameter, and in this prediction model, the mean error is 0; φ represents the standard deviation parameter, which defines the degree of dispersion of the prediction error distribution; erf(z) is the error function, z represents the input variable, and t is the dummy variable in the definite integral operation, with its integration range from 0 to z.

7. The method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence as described in claim 1, characterized in that, The formula for calculating the feedback truth value in step 6 is: and true (k)=β·y sample (k)+(1-β)·y predict (k) Among them, y true (k) is the true value of the feedback for the current period, β is the confidence level in step 5, is the sampled value of the feedback for the current period, and is the predicted value of the feedback for the current period.

8. The method for detecting and suppressing feedback anomalies in electric servo motors based on adaptive confidence as described in claim 1, characterized in that, The formula for calculating the recursive weighted average error in step 7 is as follows: σ(k)=μ·σ(k-1)+(1-μ)(y predict (k)-y sample (k)) Where σ(k) is the recursive weighted average error, μ is the weighting factor, and σ max σ is the upper limit of the error. min This is the lower limit of the error.

Images