Servo cylinder piston member offset correction method and system

Abstract

The application discloses a servo cylinder piston piece offset correction method and system, comprising: generating a training set through a differential operation method; continuously training the training set by using a random forest method to obtain a trained prediction model, and then forming a radial offset prediction model of the servo cylinder piston piece; predicting radial offset data of the servo cylinder piston piece according to real-time motion data of the servo cylinder piston piece; judging whether the radial offset data of the servo cylinder piston piece exceeds a safety range; if yes, making an offset correction unit move towards the servo cylinder piston piece to pre-correct the piston piece of the servo cylinder to the safety range, and if not, reacquiring the real-time motion data of the servo cylinder piston piece. The servo cylinder piston piece offset correction method and system can predict the radial offset of the piston piece, pre-correct the radial offset of the piston piece, improve the correction efficiency of the piston piece, and avoid affecting the machining efficiency.

Description

Technical Field

[0001] This invention relates to the field of linear reciprocating motion equipment technology, specifically to a method and system for correcting piston offset in a servo electric cylinder. Background Technology

[0002] Servo electric cylinders are characterized by high precision, high repeatability, fast response, and programmability, and are widely used in automated systems requiring precise positioning and control, such as machine tools, printing equipment, packaging machinery, and robots. The working principle of existing servo electric cylinders is as follows: the electric motor, through a reducer and lead screw, converts the rotational motion of the servo motor into linear motion. Simultaneously, a position feedback system monitors the slider position in real time and sends feedback signals to the control system. The control system, based on the set target position and parameters, applies appropriate current and voltage to the motor through the controller and driver to achieve precise control of the piston position. However, during use, existing servo electric cylinders are prone to radial offset due to changes in the dynamic characteristics of the piston (such as changes in position data, motion data, and acceleration data). This radial offset can affect machining accuracy in precision applications. Currently, existing servo electric cylinders are typically monitored periodically by operators or automated robots using measuring instruments to detect radial offset data. When the radial offset exceeds a set value, the servo electric cylinder is disassembled for repair. Therefore, this method suffers from drawbacks such as inaccurate radial offset detection, high correction costs, and reduced machining efficiency. Therefore, there is an urgent need for a servo electric cylinder piston offset correction method that can predict the radial offset of the piston each time it works, and then perform radial offset correction on the piston in advance to improve the correction efficiency of the piston and avoid affecting the processing efficiency. Summary of the Invention

[0003] To overcome the above shortcomings, the purpose of this invention is to provide a servo electric cylinder piston offset correction method and system, which is flexible in application, can predict the radial offset distance of the piston based on the dynamic characteristics of the piston, and provides pre-correction for pistons that are predicted to exceed the safe range of radial offset distance, thereby improving the real-time performance of radial offset detection of servo electric cylinder pistons, reducing correction costs, and avoiding affecting the processing efficiency of servo electric cylinders.

[0004] One aspect of the present invention provides a servo electric cylinder piston offset correction system, comprising: a servo electric cylinder; a state detection unit installed at the output end of the servo electric cylinder for acquiring piston state data and radial offset data of the servo electric cylinder; wherein the state detection unit comprises: an eddy current sensor for acquiring radial offset data of the servo electric cylinder piston; a velocity sensor for acquiring motion velocity data of the servo electric cylinder piston; an acceleration sensor for acquiring motion acceleration data of the servo electric cylinder piston; a displacement sensor for acquiring motion position data of the servo electric cylinder piston; and a vibration detection sensor for collecting vibration signal data of the servo electric cylinder piston during its movement. The offset correction unit includes a mounting ring, a correction drive, an abutment, and a rolling element. The mounting ring is mounted on the output end of the servo cylinder and has adjustment holes, with several adjustment holes arranged in an array along the circumference of the mounting ring. The correction drive is mounted on the mounting ring and is at least partially inserted into the adjustment holes. The abutment is connected to the output end of the correction drive and at least partially passes through the adjustment holes, and the abutment has a rolling groove. The rolling element is mounted in the rolling groove and abuts against the piston of the servo cylinder to be corrected. The control unit includes: a model training module, used to generate a training set through differential operations, continuously train the training set using a random forest method to obtain a trained prediction model, and then embed the prediction model into the operating system of the servo cylinder to form a radial offset prediction model for the servo cylinder piston; an offset prediction module, used to predict the radial offset data of the servo cylinder piston based on the real-time motion data of the servo cylinder piston; and an offset correction module, used to determine whether the radial offset data of the servo cylinder piston exceeds the safe range. If so, the offset correction unit moves towards the servo cylinder piston to pre-correct the piston to the safe range. If not, the real-time motion data of the servo cylinder piston is reacquired.

[0005] Furthermore, one aspect of the present invention also provides a method for correcting the offset of a servo electric cylinder piston, comprising the following steps: Step S1: Generating a training set by a differential operation method.

[0006] Step S1 also includes the following training set generation process: Step S10: Collect motion data of the piston component of the servo electric cylinder to obtain dynamic data of the piston component; Step S11: Perform preset processing on the dynamic data; Step S12: Perform a first difference operation on the preset processed dynamic data: Δx(t) = x(t) - x(t-1); where x(t) is the original motion data, and Δx(t) is the dynamic data after the first difference operation; Step S13: Perform a second difference operation on the dynamic data after the first difference operation: Δ 2 x(t)=Δx(t)-Δx(t-1); where, Δ 2x(t) is the dynamic data after the second difference operation; Step S14: Extract the first preset feature from the dynamic data after the second difference operation; Step S15: Use the extracted first preset feature as the input feature of the training set, and use the radial offset of the piston of the servo electric cylinder as the output label to construct the training set.

[0007] Step S2: The training set is continuously trained using a random forest method to obtain a trained prediction model. This prediction model is then embedded into the operating system of the servo cylinder to form a radial offset prediction model for the servo cylinder piston. Step S2 also includes the following training process for the prediction model: Step S20: Preset motion data of the servo cylinder piston is collected, along with radial offset values ​​matching the preset motion data; Step S21: Random forest training parameters are set; Step S22: The training set is assigned to the root node; Step S23: For the current node, each feature is divided into a left subset and a right subset based on preset segmentation features and thresholds; Step S24: The minimum squared difference of the target variables of the divided left and right subsets is calculated.

[0008]

[0009] Where, x i,j y is the value of the i-th sample on feature j. t s is the radial offset of the corresponding servo electric cylinder piston; s is the split point; Step S25: Create two child nodes, corresponding to the left subset and the right subset respectively, and assign training samples to the child nodes; Step S26: Recursively execute steps S22-S25 for each child node until the preset termination condition is met; Step S27: Then generate a prediction model, and use the average value of the samples in the leaf nodes as the prediction result of the decision tree to predict the radial offset data of the servo electric cylinder piston.

[0010] Step S2 also includes the following optimization process for the prediction model: Step S200: Extract preset data from the piston motion data of the servo electric cylinder, and then use the preset data as a test set; Step S201: Use the prediction model trained above to predict the samples in the test set and obtain the prediction results; Step S202: Compare the prediction results with the actual radial offset data in the test set and calculate the evaluation index: Where n is the number of samples in the test set, Y i It is the actual radial offset data of the i-th sample. It is the predicted value; Step S203: Optimize the prediction model according to the results of the evaluation index; Step S204: Repeat steps S201-S203 until the preset model performance is achieved.

[0011] Step S3: Based on the real-time motion data of the servo electric cylinder piston, predict the radial offset data of the servo electric cylinder piston.

[0012] Step S3 also includes the following correction process based on historical motion data: Step S30: Predict the radial offset data of the servo cylinder piston based on the historical motion data of the servo cylinder piston; Step S31: Determine whether the radial offset data of the servo cylinder piston exceeds the safe range; Step S32: If yes, when the servo cylinder piston moves to a preset distance, move the offset correction unit toward the servo cylinder piston to pre-correct the servo cylinder piston to the safe range; if no, obtain the real-time motion data of the servo cylinder piston.

[0013] Step S4: Determine whether the radial offset data of the servo electric cylinder piston exceeds the safe range.

[0014] Step S5: If yes, move the offset correction unit toward the piston of the servo cylinder to pre-correct the piston of the servo cylinder to a safe range; if no, return to step S3 and reacquire the real-time motion data of the piston of the servo cylinder.

[0015] Furthermore, one aspect of the present invention also provides another method for correcting the offset of a servo electric cylinder piston component, the method further comprising the following steps: Step S100: Collect vibration data of the piston component of the servo electric cylinder and generate a training set containing n samples: X={x1,x2,…,x n}, where each sample contains m features: x i =[x i1 ,x i2 ,…,x im Step S101: Collect the radial offset data of the servo electric cylinder piston component matched with the vibration data, and generate a label set: Q = {q1,q2,…,q} n Step S102: Train the training set using a regression-based support vector machine method to generate a radial offset prediction model based on vibration analysis, ensuring that the predicted values ​​of the radial offset data are close to the actual values.

[0016]

[0017]

[0018] Where f(X) represents the prediction function; C is the regularization parameter, which controls the complexity; ξ i and ∈ is a slack variable, and ∈ is the tolerance for prediction accuracy; Step S103: Based on the real-time motion data of the servo cylinder piston, predict the radial offset data of the servo cylinder piston using a radial offset prediction model based on vibration analysis; Step S104: Determine whether the radial offset data of the servo cylinder piston exceeds the safe range; Step S105: If yes, move the offset correction unit toward the servo cylinder piston to correct the piston to the safe range; if no, return to step S103 and reacquire the real-time motion data of the servo cylinder piston.

[0019] Furthermore, one aspect of the present invention also provides another method for correcting the offset of a servo electric cylinder piston component, the method further comprising the following steps: Step S1000: Acquiring time-series data related to the radial offset of the servo electric cylinder piston component, including: the measured value of the radial offset of the piston rod and the matching timestamp; Step S1001: Assuming a model: Where Δ is the difference operation, This is the original time series data, where α is the constant term, β is the time trend term, γ is the coefficient, p is the lag order, and θ is the time series data. i It is the lag coefficient, ε t It is the error term; Step S1002: Calculate the unit root test statistic: in, This is the estimated value of γ. It is the standard deviation of γ; Step S1003: Compare the unit root test statistic with the critical value: if it is less than, reject the null hypothesis and consider the time series data to be stationary; if it is greater than, accept the null hypothesis and consider the time series data to be non-stationary; Step S1004: Based on the stationary time series data, set the time series prediction model and its order.

[0020] The method further includes the following steps when performing unit root tests on the time series data: Step S1100: Hypothetical model: in, This is the original time series data, where μ is the mean and ε is the value of the time series data. t This is the error term, p is the lag order, and θ is the error term. i It is the lag coefficient; Step S1101: Calculate the unit root test statistic: in, It is the residual of the sequence, σ 2It is the variance of the residuals; Step S1102: Compare the unit root test statistic with the critical value: if it is less than, accept the null hypothesis and consider the time series data to be stationary; if it is greater than, reject the null hypothesis and consider the time series data to be non-stationary; Step S1103: Based on the stationary time series data, set the time series prediction model and its order.

[0021] The method further includes the following steps after comparing the unit root test statistic with the critical value: Step S1200: If the time series data does not meet the stationarity requirement, then perform a first-order differencing operation on the time series data: ΔR t =R t -R t-1 Among them, R t It is the t-th observation, R t-1 This is the (t-1)th observation; Step S1201: Determine whether the time series data after the first-order differencing operation meets the stationarity requirement. If yes, then set the time series prediction model and its order based on the stationary time series data; otherwise, perform the next-order differencing operation: Δ 2 R t =ΔR t -ΔR t-1 ; where ΔR t It is the t-th observation of the time series data after first-order differencing, ΔR t-1 It is the (t-1)th observation of the time series data after first-order differencing; Step S1202: Repeat steps S1200-S1201 until the time series data after differencing meets the stationarity requirement, and then set the time series prediction model and its order based on the stationarized time series data.

[0022] Furthermore, one aspect of the present invention also provides another method for correcting the offset of a servo electric cylinder piston. This method dynamically tracks and corrects the piston by real-time tracking of its radial offset data. The method includes the following steps: Step S50: Acquire real-time radial offset data of the pre-corrected servo electric cylinder piston; Step S51: Based on the real-time radial offset data, determine whether the real-time radial offset distance of the piston still exceeds the safe range. If not, return to Step S50 and reacquire the real-time radial offset data of the piston. If yes, generate radial adjustment data for the piston based on the real-time radial offset data; Step S52: Based on the radial adjustment data, move the offset correction unit towards the piston, bringing the output end of the offset correction unit into contact with the piston, thereby increasing the pushing force of the offset correction unit and adjusting the radial offset distance of the piston to within the safe range.

[0023] The technical solution of the present invention has the following advantages compared with the prior art:

[0024] 1. Based on the dynamic characteristics of the piston, the random forest algorithm can be used to predict the radial offset distance of the piston, and pre-correction can be provided for pistons that exceed the safe range of radial offset distance. This improves the real-time performance of radial offset detection of servo electric cylinder pistons, reduces correction costs, and avoids affecting the processing efficiency of servo electric cylinders.

[0025] 2. Based on the dynamic characteristics of the piston component, the radial offset distance of the piston component can be predicted using vibration analysis methods. The spectral characteristics, time domain characteristics, or time-frequency characteristics of the vibration signal can better reflect the actual situation of the piston rod movement, thereby helping to improve the accuracy of the prediction.

[0026] 3. It can predict the radial offset distance of the piston component based on the radial offset data and its related time series data using a time series prediction method. It can better capture the pattern and trend of the radial offset data changing over time; it can handle nonlinear relationships, thus better capturing the complex relationship between the radial offset data of the piston component and time; it can also provide accurate prediction results and uncertainty estimates for the control unit, thereby helping the control unit to make accurate decisions. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0028] Figure 1 This is a first cross-sectional schematic diagram of the servo electric cylinder of the present invention.

[0029] Figure 2 This is a first perspective view of the servo electric cylinder of the present invention.

[0030] Figure 3 This is a second cross-sectional schematic diagram of the servo electric cylinder of the present invention.

[0031] Figure 4 This is a second perspective view of the servo electric cylinder of the present invention.

[0032] Figure 5 This is the first connection diagram of the servo electric cylinder piston offset correction system of the present invention.

[0033] Figure 6 This is the second connection diagram of the servo electric cylinder piston offset correction system of the present invention.

[0034] Figure 7 This is a flowchart of the servo electric cylinder piston offset correction method of the present invention.

[0035] Figure 8 This is a flowchart of the training set construction method of the present invention.

[0036] Figure 9 This is a flowchart of the training set training method of the present invention.

[0037] Figure 10 This is a flowchart of the prediction model optimization method of the present invention.

[0038] Figure 11 This is a flowchart of the radial offset prediction method based on vibration analysis of the present invention.

[0039] Figure 12 This is a flowchart of the radial offset prediction method based on time series analysis of the present invention.

[0040] Figure 13 This is a flowchart of the KPSS test method of the present invention.

[0041] Figure 14 This is a flowchart of the stable operation method of the present invention.

[0042] Figure 15 This is a flowchart of the dynamic tracking and correction method for radial offset of piston components according to the present invention. Detailed Implementation

[0043] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0044] One aspect of this invention provides a servo electric cylinder piston offset correction system. The servo electric cylinder described herein is preferably a servo electric cylinder using a lead screw as its transmission mechanism. Servo electric cylinders using a lead screw as their transmission mechanism include, but are not limited to, direct-drive servo electric cylinders and folding servo electric cylinders. A direct-drive servo electric cylinder mainly includes a servo motor, a reducer, a lead screw, a lead screw nut and coupling, and a piston rod. The servo motor drives the lead screw to rotate, and the lead screw nut converts this into linear motion, thereby pushing the piston rod to achieve linear reciprocating motion. A folding servo electric cylinder mainly includes a servo motor, a reducer, a synchronous pulley, a synchronous belt, a lead screw, a lead screw nut, coupling, and a piston rod. The servo motor, in conjunction with the synchronous pulley, drives the lead screw to rotate, and the lead screw nut converts this into linear motion, thereby pushing the piston rod to achieve linear reciprocating motion. The folding installation method shortens the overall length of the servo electric cylinder, meeting the application space requirements of limited installation occasions. The synchronous pulley allows for adjustment of the reduction ratio within a small range. In summary, the basic principle of the servo electric cylinder involved is: using a servo motor to directly or indirectly drive the lead screw to rotate, thereby converting the lead screw nut into linear motion, which in turn pushes the piston rod to achieve linear reciprocating motion.

[0045] refer to Figures 1-6 As shown, in some embodiments, the servo electric cylinder piston component 13 offset correction system involved consists of at least a servo electric cylinder, a status detection unit 2, an offset correction unit 3, and a control unit 4.

[0046] Servo Electric Cylinder

[0047] The servo electric cylinder includes a cylinder body 10, an electric cylinder drive component 11, a transmission component 12, and a piston component 13. The electric cylinder drive component 11 is mounted on the cylinder body 10. The two ends of the transmission component 12 are connected to the output end of the electric cylinder drive component 11 and the input end of the piston component 13, respectively. The electric cylinder drive component 11 drives the transmission component 12 to move, and then the transmission component 12 drives the piston component 13 to move synchronously.

[0048] Among them, reference Figure 1 , Figure 3 As shown, the electric cylinder drive unit 11 includes a reducer 110 and a servo motor 111. The reducer 110 is installed at the input end of the cylinder body 10, and the servo motor 111 is connected to the reducer 110. The output end of the servo motor 111 passes through the reducer 110 and is at least partially inserted into the input port of the cylinder body 10.

[0049] Among them, reference Figure 1 , Figure 3As shown, the transmission component 12 includes a coupling 120, a lead screw 121, and a lead screw 121 nut. The coupling 120 is connected to the output end of the servo motor 111. One end of the lead screw 121 is connected to the coupling 120, and the other end of the lead screw 121 is connected to the piston component 13. The lead screw 121 nut is installed on the lead screw 121. The transmission component 12 also includes a buffer pad. The buffer pad can be any material and style of buffer pad as long as it can provide cushioning for the lead screw nut, including but not limited to pure rubber pads and sandwich rubber buffer pads. The specific design is determined by the operator based on actual processing requirements and cost.

[0050] It should be noted that the first synchronization element 14 and the second synchronization element 15 can also be set according to actual needs.

[0051] Among them, reference Figure 3 , Figure 4 As shown, the first synchronizer 14 is mounted on and connected to the output end of the servo motor 111; the second synchronizer 15 is mounted on and fixedly connected to the coupling 120. The second synchronizer 15 is connected to the first synchronizer 14 via a synchronous belt 16, and the first synchronizer 14 drives the second synchronizer 15 to move synchronously. The arrangement of the first synchronizer 14 and the second synchronizer 15 forms a foldable servo electric cylinder, thereby shortening the overall length of the servo electric cylinder and meeting the application space requirements of limited installation environments.

[0052] Status detection unit 2

[0053] The status detection unit 2 is installed at the output end of the cylinder body 10 of the servo electric cylinder, and the status detection sensors include an eddy current sensor 20, a velocity sensor 21, an acceleration sensor 2221, a displacement sensor 23, and a vibration detection sensor 24. The eddy current sensor 20 is used to acquire radial offset data of the piston 13 of the servo electric cylinder; the velocity sensor 21 is used to acquire motion velocity data of the piston 13 of the servo electric cylinder; the acceleration sensor 2221 is used to acquire motion acceleration data of the piston 13 of the servo electric cylinder; the displacement sensor 23 is used to acquire motion position data of the piston 13 of the servo electric cylinder; and the vibration detection sensor 24 is used to collect vibration signal data during the motion of the piston 13 of the servo electric cylinder.

[0054] Offset correction unit 3

[0055] The offset correction unit 3 includes an adjusting ring 30, a connecting member 31, a correction drive 32, a contact member 33, and a ball bearing 34. The adjusting ring 30 is sleeved on the output end of the piston rod, and the adjusting ring 30 has adjusting holes 300. Several adjusting holes 300 are arranged in a circular array along the circumference of the adjusting ring 30, and in this embodiment, they are arranged at equal intervals around the circumference. At least one connecting member 31 is provided, and one end of the connecting member 31 is fixedly connected to the adjusting ring 30, and the other end of the connecting member 31 is fixedly connected to the cylinder body 10. The correction drive 32 is mounted on the adjusting ring 30, and the output end of the correction drive 32 is connected to the contact member 33. The correction drive 32 drives the contact member 33 to move, thus correcting the piston rod. The moving member 32 can be any type of drive mechanism that can drive the abutment member 33 to move linearly, including but not limited to hydraulic devices, pneumatic devices, servo motors 111 (drive screw, cam type), etc., which provide reciprocating linear motion power. One end of the abutment member 33 is connected to the correction drive member 32, and the other end of the abutment member 33 is at least partially inserted into the adjustment hole 300 and has a ball bearing 34 groove. The ball bearing 34 is installed in the ball bearing 34 groove. The adjustment member drives the ball bearing 34 to move towards the piston member 13 to be adjusted, thereby radially adjusting the piston member 13. Then, the ball bearing 34 abuts against the piston member 13 to assist the movement of the piston member 13 to be adjusted.

[0056] Control Unit 4

[0057] The control unit 4 includes a model training module 40, an offset prediction module 41, and an offset correction module 42.

[0058] By adopting the above technical solution, the state detection unit 2 can detect the radial offset data, motion speed data, motion acceleration data, motion position data and vibration signal data of the piston rod in real time when the piston rod moves, thereby identifying the radial offset of the piston rod in real time, so as to take timely measures to adjust and correct the piston rod radially. By setting the offset correction unit 3, in this embodiment, the piston part 13 can be corrected for the piston part 13 that exceeds the safe range of radial offset distance according to the dynamic characteristics of the piston part 13, reducing the correction cost and avoiding affecting the processing efficiency of the servo electric cylinder.

[0059] Therefore, one aspect of this invention proposes a method for correcting the offset of a servo cylinder piston 13. This method uses a random forest model to predict the radial offset distance of the piston 13, and then provides radial offset correction for pistons whose predicted radial offset exceeds the safe range. Simultaneously, it reduces the impact on the linear motion of the piston 13, improves the real-time performance of radial offset detection, reduces correction costs, and avoids affecting the machining efficiency of the servo cylinder. Preferably, in actual implementation, the method for correcting the offset of a servo cylinder piston 13 is executed by the aforementioned servo cylinder piston 13 offset correction system. For example, this method can be designed into a computer-understandable language format (e.g., a computer program) and stored in a storage medium, so that it can be invoked and executed by the control unit 4 of the servo cylinder piston 13 offset correction system during the operation of the servo cylinder.

[0060] The present invention provides a method for offset correction of piston component 13 in a servo electric cylinder, which provides radial offset prediction for piston component 13 of the servo electric cylinder, thereby providing automated radial offset adjustment. Specifically, refer to... Figures 5-10 In some embodiments, the offset correction method for the piston component 13 of the servo electric cylinder of the present invention consists of steps S1-S5.

[0061] Step S1: Generate a training set using the difference operation method.

[0062] Among them, reference Figure 8 As shown, the method for generating a training set using the difference operation includes the following steps:

[0063] Step S10: Collect motion data of the piston component 13 of the servo electric cylinder and obtain dynamic data of the piston component 13; wherein, the motion data includes, but is not limited to, radial offset data, motion speed data, motion acceleration data, motion position data and vibration signal data of the piston component 13, and then generate real-time dynamic data of the piston component 13.

[0064] Step S11: Perform preset processing on the dynamic data, including noise removal, filtering, and sampling rate adjustment to ensure the quality of the dynamic data; wherein the noise removal process uses a moving average filter to process the dynamic data of piston 13.

[0065]

[0066] in, This is the data after noise removal, and N is the size of the sliding window. For example, assuming that dynamic data of piston part 13 at 100 time points are collected, and a 3-point moving average filter is used for noise removal, the filtered value for the data at the 4th time point is the average value of the data at the previous 3 time points.

[0067] The filtering process employs a low-pass filter to process the dynamic data of piston component 13.

[0068]

[0069] in, This refers to the filtered data, where 'a' is the attenuation coefficient of the filter. For example, assuming a first-order low-pass filter is used to filter the dynamic data of piston 13, and the attenuation coefficient 'a' is set to 0.2, then the filtered data... It is 0.8 times the data from the previous time point plus 0.2 times the data from the current time point.

[0070] The sampling rate adjustment process uses linear interpolation to process the dynamic data of piston component 13.

[0071]

[0072] in, This is the data after sampling rate adjustment, f x It is the sampling rate of dynamic data, f t It is the target sampling rate; for example, assuming the sampling rate of the original piston part 13 dynamic data is 100Hz, we need to adjust it to 50Hz. Using the linear interpolation method, for each time point under the target sampling rate, the interpolation result can be obtained by taking the average of two adjacent time points in the original data.

[0073] In cases where the sampling rate is reduced, a decimation algorithm is used to lower the sampling rate:

[0074]

[0075] in, It is a floor function.

[0076] Step S12: Perform the first difference operation on the pre-processed dynamic data:

[0077] Δx(t) = x(t) - x(t-1)

[0078] Where x(t) is the original motion data, and Δx(t) is the dynamic data after the first difference.

[0079] Step S13: Perform a second difference operation on the dynamic data after the first difference operation:

[0080] Δ 2 x(t) = Δx(t) - Δx(t-1)

[0081] Where, Δ 2 x(t) represents the dynamic data after the second difference.

[0082] Step S14: Extract the first preset feature from the dynamic data after the second difference operation;

[0083] The first preset feature includes, but is not limited to, statistical features (such as mean, variance, peak value, etc.) or frequency domain features (such as spectrum analysis, power spectral density, etc.) to represent the motion characteristics of the piston 13.

[0084] Step S15: Use the extracted first preset features as the input features of the training set, and use the radial offset of the piston part 13 of the servo electric cylinder as the output label to construct the training set.

[0085] Step S2: The training set is continuously trained using the random forest method to obtain the trained prediction model, and then the prediction model is embedded into the operating system of the servo electric cylinder to form the radial offset prediction model of the piston component 13 of the servo electric cylinder.

[0086] Among them, reference Figure 9 As shown, the method of continuously training the training set using random forest includes the following steps:

[0087] Step S20: Collect preset motion data of the servo electric cylinder piston 13, and simultaneously collect radial offset values ​​that match the preset motion data. The preset motion data includes, but is not limited to, motion speed data, motion acceleration data, motion position data, and vibration signal data. The preset motion data and the matching radial offset values ​​are integrated into a dataset, and the dataset is then divided into a training set and a test set. 70% of the data in the dataset is used as the training set, and 30% of the data is used as the test set to ensure the generalization ability of the evaluation model.

[0088] Step S21: Set the training parameters for the random forest, including: the number of decision trees, the maximum depth of each decision tree, and the range of feature subsets.

[0089] Step S22: Assign the training set to the root node.

[0090] Step S23: For the current node, divide each feature into a left subset and a right subset according to the preset segmentation features and thresholds, where the preset segmentation features and thresholds refer to the features and thresholds that minimize the variance of the target variable.

[0091] Step S24: Calculate the minimum squared difference of the objective variable between the left and right subsets after partitioning:

[0092]

[0093] Where, x i,j y is the value of the i-th sample on feature j. ts is the radial offset of the corresponding servo electric cylinder piston component 13; s is the dividing point.

[0094] Step S25: Create two child nodes, corresponding to the left subset and the right subset respectively, and assign the training samples to the child nodes.

[0095] Step S26: Recursively execute steps S22-S25 for each child node until a preset termination condition is met. The preset termination condition includes, but is not limited to, reaching the set maximum depth, the number of subset samples being less than a set value, and the subset being unable to be further divided (i.e., all target variables in the subset are equal).

[0096] Step S27: Then generate a prediction model and use the average value of the samples in the leaf nodes as the prediction result of the decision tree to predict the radial offset data of the servo electric cylinder piston 13.

[0097] Step S3: Based on the real-time motion data of the servo cylinder piston 13, predict the radial offset data of the servo cylinder piston 13.

[0098] Among them, reference Figure 7 As shown, before predicting the radial offset data of the servo cylinder piston 13 based on the real-time motion data of the servo cylinder piston 13, the method further includes the following steps:

[0099] Step S30: Based on the historical motion data of the servo cylinder piston 13, predict the radial offset data of the servo cylinder piston 13. The historical motion data includes, but is not limited to, the historical motion speed data, historical motion acceleration data, historical motion position data, historical vibration signal data, and the radial offset value matched by the above data of the servo cylinder piston 13.

[0100] Step S31: Determine whether the radial offset data of the servo electric cylinder piston 13 exceeds the safe range.

[0101] The safety range is set by the operators based on actual needs and costs.

[0102] Step S32: If yes, when the servo cylinder piston 13 moves to the preset distance, the offset correction unit 3 moves in the direction of the servo cylinder piston 13 to pre-correct the servo cylinder piston 13 to a safe range; if no, then execute step S3 to obtain the real-time motion data of the servo cylinder piston 13.

[0103] The preset distance is set by the operator according to actual needs and cost, and the preset distance is at least greater than the distance between the piston of the servo electric cylinder 13 and the offset correction unit 3.

[0104] Step S4: Determine whether the radial offset data of the servo electric cylinder piston 13 exceeds the safe range.

[0105] Step S5: If yes, move the offset correction unit 3 toward the servo cylinder piston 13 to pre-correct the servo cylinder piston 13 to a safe range; if no, return to step S3 and reacquire the real-time motion data of the servo cylinder piston 13.

[0106] refer to Figure 10 As shown, in one example executed in the order of operation, the method for offset correction of piston component 13 of a servo electric cylinder according to the present invention further includes the following steps when embedding the prediction model into the operating system of the servo electric cylinder:

[0107] Step S200: Extract preset data from the motion data of the piston component 13 of the servo electric cylinder, and then use the preset data as a test set.

[0108] The preset data refers to a dataset containing motion speed data, motion acceleration data, motion position data, vibration signal data, and radial offset values ​​that match the preset data of the servo electric cylinder piston 13. 30% of the data in the dataset is used as a test set to ensure the generalization ability of the evaluation model.

[0109] Step S201: Use the trained prediction model to predict the samples in the test set and obtain the prediction results. Step S202: Compare the prediction results with the actual radial offset data in the test set, and calculate the evaluation index using the root mean square error.

[0110]

[0111] Where n is the number of samples in the test set, Y i It is the actual radial offset data of the i-th sample. This is a predicted value.

[0112] Step S203: Based on the results of the evaluation metrics, optimize the prediction model, including adjusting the model's hyperparameters, increasing the amount of training data, and improving features, and then retrain the prediction model and evaluate it again on the test set.

[0113] Step S204: Repeat steps S201-S203 until the preset model performance is achieved.

[0114] It is important to note that overfitting should be avoided as much as possible when evaluating and optimizing models. Overfitting may cause the model to perform well on the training set but poorly on new, unseen data. Existing cross-validation techniques can be used to help evaluate the model's generalization ability.

[0115] refer to Figures 1-6 , Figure 11 As shown, in one example executed in a working sequence, the servo electric cylinder piston component 13 offset correction method of the present invention further includes the following steps:

[0116] Step S100: Collect vibration data of the piston component 13 of the servo electric cylinder and generate a training set containing n samples: X = {x1, x2, ..., x n}, where each sample contains m features: x i =[x i1 ,x i2 ,…,x im The vibration data includes, but is not limited to, at least one of vibration frequency, vibration amplitude, peak factor, peak-valley factor, and impulse index; the generated training set is normalized using existing features and noise is removed to improve the training effect of the model.

[0117] Step S101: Collect radial offset data of the servo electric cylinder piston 13 for vibration data matching, and generate a label set: Q = {q1,q2,…,q} n}

[0118] Step S102: Train the training set using a regression-based support vector machine (SVM) method to generate a radial offset prediction model based on vibration analysis, ensuring that the predicted values ​​of the radial offset data are close to the actual values. The objective function of the regression-based SVM can be expressed as:

[0119]

[0120]

[0121] Where b is the bias term, f is the weight vector, f(X) is the prediction function; C is the regularization parameter, which controls the complexity; ξ i and ∈ is a slack variable, and ∈ is the tolerance for prediction accuracy.

[0122] Step S103: Based on the real-time motion data of the servo electric cylinder piston 13, predict the radial offset data of the servo electric cylinder piston 13 using a radial offset prediction model based on vibration analysis.

[0123] Step S104: Determine whether the radial offset data of the servo electric cylinder piston 13 exceeds the safe range.

[0124] Step S105: If yes, move the offset correction unit 3 toward the servo cylinder piston 13 to correct the servo cylinder piston 13 to a safe range; if no, return to step S103 and reacquire the real-time motion data of the servo cylinder piston 13.

[0125] For example: Suppose a sample x in the training set i Includes the following characteristic value: vibration frequency x i1 =100Hz, vibration amplitude x i2 =0.5mm, peak factor x i3 =1.2, peak-valley factor x i4 =0.8, pulse index x i5 =0.6, corresponding to the label value Q i =0.4mm.

[0126] By training a support vector machine regression model, a prediction function f(X) can be obtained. Assuming that a new vibration sample is collected during the real-time prediction phase, its feature value is: vibration frequency x... new1 =120Hz, vibration amplitude x new2 =0.6mm, peak factor x new3 =1.5, peak-valley factor x new4 =0.9, impulse index x new5 =0.7; Set the new sample x newi The input is fed into a support vector machine regression model, which then predicts the radial offset of the piston rod, O. new =0.45mm.

[0127] refer to Figures 1-6 , Figure 12 As shown, in one example executed in a working sequence, the servo electric cylinder piston component 13 offset correction method of the present invention further includes the following steps:

[0128] Step S1000: Collect time series data related to the radial offset of the servo electric cylinder piston 13; wherein, the time series data refers to the data related to the radial offset of the servo electric cylinder piston rod, including but not limited to: the measured value of the radial offset of the piston rod and the matching timestamp.

[0129] Step S1001: Assumption Model:

[0130]

[0131] Where Δ is the difference operation, This is the original time series data, where α is the constant term, β is the time trend term, γ is the coefficient, p is the lag order, and θ is the time series data. i It is the lag coefficient, ε t This is the error term.

[0132] Step S1002: Calculate the unit root test statistic:

[0133]

[0134] in, This is the estimated value of γ. It is the standard deviation of γ; the ADF (Augmented Dickey-Fuller Test) test is used to test whether time series data has a unit root (non-stationarity). The null hypothesis of the ADF test is that the series has a unit root (non-stationarity), and the alternative hypothesis is that the series does not have a unit root (stationarity).

[0135] Step S1003: Compare the unit root test statistic with the critical value: If the ADF is less than the critical value, reject the null hypothesis and consider the time series data to be stationary; if the ADF is greater than the critical value, accept the null hypothesis and consider the time series data to be non-stationary.

[0136] For example: Suppose we have a time series dataset [10,15,12,18,16,14,20], and we perform an ADF test to determine the stationarity of the series:

[0137] With a lag order p = 1 and a significance level of 0.05, the ADF statistic is calculated to be -1.8614 based on the null and alternative hypotheses of the ADF test. For the series of the seven observations, the critical value is -3.617. Since the ADF statistic (-1.8614) is greater than the critical value (-3.617), the null hypothesis is accepted, and the series is considered non-stationary. Therefore, further time series analysis and forecasting require differencing to transform the series into a stationary one. Step S1004: Based on the stationary time series data, define the time series forecasting model and its order.

[0138] refer to Figures 1-6 , Figure 13 As shown, in one example executed in the order of operation, the method for offset correction of the piston component 13 of a servo electric cylinder according to the present invention further includes the following steps when performing unit root tests on time series data:

[0139] Step S1100: Assumption Model:

[0140]

[0141] in, This is the original time series data, where μ is the mean and ε is the value of the time series data. t This is the error term, p is the lag order, and θ is the error term. i It is the lag coefficient.

[0142] Step S1101: Calculate the unit root test statistic:

[0143]

[0144] in, It is the residual of the sequence, σ 2 This is the variance of the residuals. The KPSS test is used to test whether time series data has a unit root (non-stationarity). The null hypothesis of the KPSS test is that the series is stationary, and the alternative hypothesis is that the series is non-stationary. Step S1102: Compare the unit root test statistic with the critical value: if the KPSS test result is less than the critical value, the null hypothesis is accepted, and the time series data is considered stationary; if the KPSS test result is greater than the critical value, the null hypothesis is rejected, and the time series data is considered non-stationary.

[0145] For example: Suppose we have a time series dataset [10,15,12,18,16,14,20], and we perform the KPSS test to determine the stationarity of the series:

[0146] With a lag order p = 1 and a significance level of 0.05, the KPSS statistic is calculated to be 0.0815 based on the null and alternative hypotheses of the KPSS test. For the series of the above 7 observations, the critical value is 0.739. Since the ADF statistic (0.0815) is less than the critical value (0.739), the null hypothesis is accepted, and the series is considered stationary. Step S1103: Based on the stationary time series data, define the time series forecasting model and its order.

[0147] In one example where the operation is performed sequentially, the method for correcting the offset of the servo cylinder piston 13 according to the present invention further includes the following steps after acquiring time-series data related to the radial offset of the servo cylinder piston 13:

[0148] Exploratory analysis was conducted on the collected time series data:

[0149] Calculate the average value of time series data:

[0150]

[0151] To calculate the median of time series data: Sort the data in ascending order and find the middle value. If the number of data points is odd, the median is the middle value after sorting; if the number of data points is even, the median is the average of the two middle values. To calculate the variance of time series data:

[0152]

[0153] Calculate the standard deviation of time series data:

[0154]

[0155] The central tendency and dispersion of the data are understood through the calculation of the mean, median, variance, and standard deviation mentioned above. (Reference) Figures 1-6 , Figure 14 As shown, in one example executed in the order of operation, the method for offset correction of the piston component 13 of a servo electric cylinder according to the present invention further includes the following steps after comparing the unit root test statistic with the critical value:

[0156] Step S1200: If the time series data does not meet the stationarity requirement, then perform a first-order differencing operation on the time series data:

[0157] ΔR t =R t -R t-1

[0158] Among them, R t It is the t-th observation, R t-1 It is the (t-1)th observation; where at least a first-order differencing operation is performed on the time series data. If the first-order differencing operation still does not meet the stationarity requirement, a second-order, third-order, or higher-order differencing operation can be performed until a stationary series is obtained.

[0159] Step S1201: Determine whether the time series data after the first-order differencing operation meets the stationarity requirement. If yes, proceed to step S1202: Based on the stationarized time series data, set the time series prediction model and its order. If not, proceed to step S12010: Perform the next-order differencing operation.

[0160] Δ 2 R t =ΔR t -ΔR t-1

[0161] Where, ΔR t It is the t-th observation of the time series data after first-order differencing, ΔR t-1 It is the (t-1)th observation of the time series data after first-order differencing.

[0162] After each difference, a stationarity test must be performed using the ADF test or the KPSS test to ensure that the differenced data meets the stationarity requirements.

[0163] Step S1202: Repeat steps S1200-S1201 until the differenced time series data meets the stationarity requirement. Then, based on the stationarized time series data, set the time series prediction model and its order.

[0164] refer to Figures 1-6 , Figure 15 As shown, in some embodiments, the servo electric cylinder piston component 13 offset correction method provided by the present invention further includes the following steps after pre-correcting the servo electric cylinder piston component 13 to a safe range:

[0165] Step S50: Collect real-time radial offset data of the pre-calibrated servo electric cylinder piston 13;

[0166] In this process, after pre-calibration, the real-time radial offset value of the servo electric cylinder piston 13 is further tracked and detected to avoid calibration misalignment problems during pre-calibration.

[0167] Step S51: Based on the real-time radial offset data, determine whether the real-time radial offset distance of the piston 13 still exceeds the safe range. If not, return to step S50 and reacquire the real-time radial offset data of the piston 13. If yes, generate the radial adjustment data of the piston 13 based on the real-time radial offset data.

[0168] Specifically, when generating radial adjustment data for piston 13, the offset correction unit 3 that requires pushing force is set as the first correction unit, and the offset correction unit 3 that does not require pushing force is set as the second correction unit. The first correction unit is bound to the value of the matching pushing distance that needs to be corrected.

[0169] Step S52: Based on the radial adjustment data, the first correction unit moves towards the piston 13, bringing the output end of the first correction unit into contact with the piston 13, thereby increasing the pushing force of the first correction unit and adjusting the radial offset distance of the piston 13 to a safe range; after the first correction unit corrects the piston 13, the remaining offset correction unit 3, i.e. the second correction unit, moves towards the adjusted piston 13, bringing the ball 34 of the second correction unit into contact with the piston 13, forming a radial adjustment range for the linear movement of the piston 13 through the first correction unit and the second correction unit, thereby reducing the radial offset of the piston 13 during movement.

[0170] Specifically, when generating radial adjustment data for piston 13, the contact 33 of offset correction unit 3 that requires pushing force is designated as the first contact 33, and the contact 33 of offset correction unit 3 that does not require pushing force is designated as the second contact 33.

[0171] Specifically, based on the radial adjustment data, the first abutment 33 is driven to move towards the piston 13 by the correction drive 32, so that the ball 34 of the first abutment 33 abuts against the piston 13. Then, the driving force is increased by the correction drive 32, which in turn increases the pushing force of the first abutment 33, and adjusts the radial offset distance of the piston 13 to a safe range.

[0172] The second abutment 33 is driven by the correction drive 32 to move in the direction of the adjusted piston 13, so that the ball 34 of the second abutment 33 abuts against the piston 13, and the radial adjustment range of the piston 13 is formed by the first abutment 33 and the second abutment 33.

[0173] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0174] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

[0175] Explanation of reference numerals in the attached figures:

[0176] 1. Servo electric cylinder 14. First synchronizer

[0177] 2. Status detection unit 15. Second synchronization element

[0178] 3. Offset correction unit 16. Synchronous belt

[0179] 4. Control Unit 20. Eddy Current Sensor

[0180] 10. Cylinder block 21. Speed ​​sensor

[0181] 11. Electric cylinder drive component 22. Accelerometer sensor

[0182] 13. Piston component 23. Displacement sensor

[0183] 24. Vibration detection sensor; 110. Gearbox

[0184] 30. Adjustment ring; 111. Servo motor

[0185] 31. Connecting parts 120. Couplings

[0186] 32. Correction drive component 121. Lead screw

[0187] 33. Abutting component 122. Lead screw nut

[0188] 34. Ball bearing 130. Gearbox

[0189] 40. Model Training Module 131. Servo Motor

[0190] 41. Offset Prediction Module 300. Adjustment

[0191] 42. Offset Correction Module

Claims

1. A method for correcting piston offset of a servo electric cylinder, comprising a servo electric cylinder, a status detection unit and an offset correction unit, wherein the status detection unit and the offset correction unit are installed at the output end of the servo electric cylinder, the status detection unit is used to acquire piston status data and radial offset data of the servo electric cylinder, and the offset correction unit is used to correct the radial offset of the piston of the servo electric cylinder. Its features are, The method includes the following steps: Step S1: Generate a training set using the difference operation method; Step S2: The training set is continuously trained using the random forest method to obtain the trained prediction model, and then the prediction model is embedded into the operating system of the servo electric cylinder to form a radial offset prediction model of the piston of the servo electric cylinder. Step S3: Based on the real-time motion data of the servo electric cylinder piston, predict the radial offset data of the servo electric cylinder piston. Step S4: Determine whether the radial offset data of the servo electric cylinder piston exceeds the safe range; Step S5: If yes, move the offset correction unit toward the piston of the servo cylinder to pre-correct the piston of the servo cylinder to a safe range; if no, return to step S3 and reacquire the real-time motion data of the piston of the servo cylinder. The method of continuously training the training set using random forest includes the following steps: Step S20: Collect preset motion data of the servo electric cylinder piston component, and simultaneously collect radial offset values ​​that match the preset motion data; Step S21: Set the training parameters for the random forest; Step S22: Assign the training set to the root node; Step S23: For the current node, divide each feature into a left subset and a right subset according to the preset segmentation features and thresholds; Step S24: Calculate the minimum squared difference of the objective variable between the left and right subsets after partitioning: ， in, It is the first Each sample in features The value on, This corresponds to the radial offset of the piston component in the servo electric cylinder. , These are the predicted values ​​for the left and right subsets. It is the dividing point; Step S25: Create two child nodes, corresponding to the left subset and the right subset respectively, and assign the training samples to the child nodes; Step S26: Recursively execute steps S22-S25 for each child node until a preset termination condition is met; Step S27: Then generate a prediction model and use the average value of the samples in the leaf nodes as the prediction result of the decision tree to predict the radial offset data of the servo electric cylinder piston.

2. The servo electric cylinder piston offset correction method according to claim 1, characterized in that, The method for generating a training set using difference operations includes the following steps: Step S10: Collect motion data of the piston component of the servo electric cylinder to obtain dynamic data of the piston component; Step S11: Perform preset processing on the dynamic data; Step S12: Perform the first difference operation on the pre-processed dynamic data: ， in, For the raw motion data, The dynamic data after the first difference; Step S13: Perform a second difference operation on the dynamic data after the first difference operation: ， in, The dynamic data after the second difference; Step S14: Extract the first preset feature from the dynamic data after the second difference operation; Step S15: Use the extracted first preset features as the input features of the training set, and use the radial offset of the piston of the servo electric cylinder as the output label to construct the training set.

3. The servo electric cylinder piston offset correction method according to claim 1, characterized in that, When embedding the prediction model into the operating system of the servo electric cylinder, the method further includes the following steps: Step S200: Extract preset data from the piston motion data of the servo electric cylinder, and then use the preset data as a test set; Step S201: Use the prediction model trained above to predict the samples in the test set and obtain the prediction results; Step S202: Compare the predicted results with the actual radial offset data in the test set, and calculate the evaluation index: ， in, It is the number of samples in the test set. It is the first Actual radial offset data for each sample It is a predicted value; Step S203: Optimize the prediction model based on the results of the evaluation indicators; Step S204: Repeat steps S201-S203 until the preset model performance is achieved.

4. The servo electric cylinder piston offset correction method according to claim 1, characterized in that, Before predicting the radial offset data of the servo electric cylinder piston, the method further includes the following steps: Step S30: Based on the historical motion data of the servo electric cylinder piston, predict the radial offset data of the servo electric cylinder piston; Step S31: Determine whether the radial offset data of the servo electric cylinder piston exceeds the safe range; Step S32: If yes, when the piston of the servo cylinder moves to a preset distance, the offset correction unit moves in the direction of the piston of the servo cylinder to pre-correct the piston of the servo cylinder to a safe range; if no, the real-time motion data of the piston of the servo cylinder is acquired.

5. The servo electric cylinder piston offset correction method according to claim 1, characterized in that, The method further includes the following steps: Step S100: Collect vibration data of the piston component of the servo electric cylinder and generate a data set containing... Training set of 10 samples: Each sample contains One characteristic: ; Step S101: Collect the radial offset data of the servo electric cylinder piston component that matches the vibration data, and generate a tag set: ,in For tag sets; Step S102: Train the training set using a regression-based support vector machine method to generate a radial offset prediction model based on vibration analysis, ensuring that the predicted values ​​of the radial offset data are close to the actual values. ， ， in, It is a weight vector; It is a bias term; Represents the prediction function; It is a regularization parameter that controls complexity; It is the total number of samples; and It is a slack variable. It is the tolerance for prediction accuracy; Step S103: Based on the real-time motion data of the servo electric cylinder piston, predict the radial offset data of the servo electric cylinder piston using a radial offset prediction model based on vibration analysis. Step S104: Determine whether the radial offset data of the servo electric cylinder piston exceeds the safe range; Step S105: If yes, move the offset correction unit toward the piston of the servo cylinder to correct the piston of the servo cylinder to a safe range; if no, return to step S103 and reacquire the real-time motion data of the piston of the servo cylinder.

6. The servo electric cylinder piston offset correction method according to claim 1, characterized in that, The method further includes the following steps: Step S1000: Collect time series data related to the radial offset of the servo electric cylinder piston, including: the measured value of the radial offset of the piston rod and the matching timestamp; Step S1001: Assumption Model: ， in, It's a difference operation. It is the original time series data. It is a constant term. It is a time trend item. It is a coefficient. It is the lag order. It is the lag coefficient. It is an error term; Step S1002: Calculate the unit root test statistic: ， in, It is an estimate. value, yes Standard deviation; Step S1003: Compare the unit root test statistic with the critical value: if it is less than the critical value, reject the null hypothesis and consider the time series data to be stationary; if it is greater than the critical value, accept the null hypothesis and consider the time series data to be non-stationary. Step S1004: Based on the stationary time series data, set the time series prediction model and its order.

7. The servo electric cylinder piston offset correction method according to claim 6, characterized in that, When performing a unit root test on the time series data, the method further includes the following steps: Step S1100: Assumption Model: ， in, It is the original time series data. It is the mean. It is an error term. It is the lag order. It is the lag coefficient; Step S1101: Calculate the unit root test statistic: ， in, It is the residual of the sequence. It is the variance of the residuals; Step S1102: Compare the unit root test statistic with the critical value: if it is less than the critical value, accept the null hypothesis and consider the time series data to be stationary; if it is greater than the critical value, reject the null hypothesis and consider the time series data to be non-stationary. Step S1103: Based on the stationary time series data, set the time series prediction model and its order.

8. The servo electric cylinder piston offset correction method according to claim 6 or 7, characterized in that, After comparing the unit root test statistic with the critical value, the method further includes the following steps: Step S1200: If the time series data does not meet the stationarity requirement, then perform a first-order differencing operation on the time series data: ， in, It is the first One observation value, It is the first One observation value; Step S1201: Determine whether the time series data after the first-order differencing operation meets the stationarity requirement. If yes, then set the time series prediction model and its order based on the stationary time series data. If not, then perform the next-order differencing operation. ， in, It is the first difference of the time series data. One observation value, It is the first difference of the time series data. One observation value; Step S1202: Repeat steps S1200-S1202 until the differenced time series data meets the stationarity requirement. Then, based on the stationarized time series data, set the time series prediction model and its order.

9. The servo electric cylinder piston offset correction method according to claim 1, characterized in that, After pre-calibrating the piston of the servo electric cylinder to a safe range, the method further includes the following steps: Step S50: Collect real-time radial offset data of the pre-calibrated servo electric cylinder piston component; Step S51: Based on the real-time radial offset data, determine whether the real-time radial offset distance of the piston component still exceeds the safe range. If not, return to step S50 and reacquire the real-time radial offset data of the piston component. If yes, generate the radial adjustment data of the piston component based on the real-time radial offset data. Step S52: Based on the radial adjustment data, the offset correction unit moves towards the piston component, bringing the output end of the offset correction unit into contact with the piston component, thereby increasing the pushing force of the offset correction unit and adjusting the radial offset distance of the piston component to a safe range.

10. A servo electric cylinder piston offset correction system, comprising the servo electric cylinder piston offset correction method according to any one of claims 1-9, characterized in that, The system includes: Servo electric cylinder; The status detection unit is installed at the output end of the servo electric cylinder and is used to acquire the status data of the piston and radial offset data of the servo electric cylinder. The offset correction unit includes a mounting ring, a correction drive, a contact member, and a rolling member. The mounting ring is mounted on the output end of the servo cylinder and has adjustment holes, with several adjustment holes arranged in an array along the circumference of the mounting ring. The correction drive is mounted on the mounting ring and is at least partially inserted into the adjustment holes. The contact member is connected to the output end of the correction drive and at least partially passes through the adjustment holes, and the contact member has a rolling groove. The rolling member is mounted in the rolling groove and abuts against the piston of the servo cylinder to be corrected. Control unit, including: The model training module is used to generate a training set through the difference operation method, and continuously train the training set using the random forest method to obtain the trained prediction model. Then, the prediction model is embedded into the operating system of the servo electric cylinder to form a radial offset prediction model of the piston of the servo electric cylinder. The offset prediction module is used to predict the radial offset data of the servo electric cylinder piston based on the real-time motion data of the servo electric cylinder piston. The offset correction module determines whether the radial offset data of the servo cylinder piston exceeds the safe range. If so, it moves the offset correction unit toward the servo cylinder piston to pre-correct the piston to the safe range. If not, it reacquires the real-time motion data of the servo cylinder piston.

11. The servo electric cylinder piston offset correction system according to claim 10, characterized in that, The status detection unit includes: Eddy current sensor is used to acquire radial offset data of the piston component of a servo electric cylinder; Speed ​​sensor, used to acquire motion speed data of the piston component of servo electric cylinder; An accelerometer is used to acquire motion acceleration data of the piston component in a servo electric cylinder. Displacement sensors are used to acquire motion position data of the piston components in servo electric cylinders; Vibration detection sensors are used to collect vibration signal data during the movement of the piston components in a servo electric cylinder.

Images