High-speed mechanical execution end anti-shake calculation control method, system and processing system

Abstract

The present invention relates to the technical field of mechanical jitter analysis, and specifically to a method, system and processing system for calculating and controlling the anti-jitter of the terminal of a high-speed machine. The method includes obtaining the trajectory image data of the motor shaft; identifying the feature points in the trajectory image data by the Harris corner point detection algorithm, and marking the feature points with features and normality to divide the feature trajectory points and normal trajectory points; predicting the feature points in the image data by a machine learning model, regulating the judgment threshold of the feature points in the image data determined by the Harris corner point detection algorithm, re-identifying and extracting the feature points to construct the running trajectory of the motor shaft; determining the jitter state of the motor shaft according to the constructed running trajectory, and controlling the anti-jitter of the motor shaft based on the jitter state. The running trajectory constructed by the feature points can be used to analyze the jitter condition of the motor shaft, and a prediction of the jitter degree of the motor shaft in the subsequent time can be given, so as to facilitate the suppression and adjustment of the jitter of the motor shaft.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical jitter analysis, and in particular to a method, system and processing system for calculating and controlling anti-jitter execution of a high-speed mechanical terminal. Background Art

[0002] During the operation of the motor, its own motor shaft will have certain jitters due to various fault factors. At present, the motion trajectory of the motor shaft can be identified by obtaining the image data of the motor shaft. However, the motion trajectory of the motor shaft is formed by obtaining the feature points in the image data. However, for the feature points in the image data, such as the feature points currently selected by the Harris corner detection algorithm, the Harris corner detection algorithm determines the feature points in the image by setting a threshold. However, because the threshold is usually set manually, there will be certain errors in the Harris corner detection algorithm, which will affect the subsequent construction of the motor shaft's running trajectory.

[0003] Moreover, after the running trajectory of the motor shaft is constructed, the jitter condition of the motor shaft can be displayed. Therefore, when the jitter condition is determined, how to accurately predict and analyze the degree of jitter of the motor shaft in subsequent situations is a problem that needs to be solved at present. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, the present invention provides a high-speed mechanical end-stage anti-shake calculation and control method, system and processing system, which can effectively solve the problem of inaccurate judgment of the jitter condition caused by errors in the prior art when analyzing the motor shaft jitter condition.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0006] The present invention provides a high-speed mechanical execution terminal anti-shake calculation and control method, comprising:

[0007] Obtain the trajectory image data of the motor shaft during operation;

[0008] The Harris corner detection algorithm is used to identify the feature points in the trajectory image data, and the feature points are extracted to create a feature set. The feature points in the feature set are marked with features and normality to divide the feature trajectory points and normal trajectory points.

[0009] The feature points in the image data are predicted by the machine learning model, thereby adjusting the threshold of the feature points in the image data determined by the Harris corner detection algorithm, so that the Harris corner detection algorithm can re-identify the feature points in the extracted image data and construct the running trajectory of the motor shaft;

[0010] The vibration state of the motor shaft is determined according to the running trajectory of the constructed motor shaft, and the motor shaft is controlled to prevent vibration based on the vibration state, wherein:

[0011] The operating parameters of the motor shaft in the shaking state are determined, and the shaking states of other motor shafts are predicted through the meta-model based on the operating parameters, so as to perform anti-shake control on the motor shaft.

[0012] Furthermore, when acquiring the trajectory image data of the motor shaft, a collection period is set to collect the trajectory image data in different collection periods.

[0013] Furthermore, the formula of the Harris corner detection algorithm is as follows:

[0014] Determine the points in the image , determine its autocorrelation function:

[0015] ;

[0016] Determine the corner response function:

[0017] ;

[0018] Where: is the autocorrelation function of the image window, is the weight function, Indicates that the image is The pixel value of is the corner point response function, is a point in the image window, is a point outside the image window, is a point in the image, u and v are relative to The horizontal and vertical offsets of

[0019] when When , it is determined as a feature point, and T is the judgment threshold.

[0020] Furthermore, the machine learning model is a support vector machine, and its algorithm formula is as follows:

[0021] ;

[0022] in, , represents 0 or 1, indicating that the feature point is a normal trajectory point or a feature trajectory point, w is the weight vector, b is the bias vector, is the regularization parameter, is the square of the Euclidean norm of the weight vector, represents the dot product between the weight vector and the feature points in the image, yes Loss function, is the feature point in the original image Kernel functions that map to high-dimensional space;

[0023] ;

[0024] in, is the output value of the decision function, according to , is the pixel threshold.

[0025] Furthermore, the jitter state includes:

[0026] Amplitude jitter mode, waveform jitter mode, frequency jitter mode, phase jitter mode, where:

[0027] By creating a comparison image for different shaking modes in the shaking state and generating a comparison set, the comparison set is compared with the shaking state of the motor shaft to determine the shaking state of the motor shaft.

[0028] Furthermore, the hybrid model is obtained by fusing a linear regression model and a random forest model, wherein the linear regression model is:

[0029] ;

[0030] Where: Y is the predicted jitter value, is the independent variable that affects the jitter value, is the intercept term, which represents the expected value of the jitter amplitude when all independent variables are zero. are regression coefficients, which represent the relationship between the respective variables and the jitter value, is the error term.

[0031] Furthermore, the random forest model is:

[0032] ;

[0033] in: is the predicted jitter value, is the average value of the motor shaft jitter predicted by all decision trees, is the jitter value predicted by the kth decision tree, is the weight of the kth decision tree;

[0034] Therefore, the meta-model after the fusion of linear regression model and random forest is:

[0035] ;

[0036] in: is the predicted jitter value after the linear regression model and random forest fusion, is the weight coefficient. The jitter value of the motor shaft is predicted by combining the meta-model with the linear regression model and the random forest model. The subsequent jitter degree of the motor shaft can be calculated more accurately, and the error caused by the jitter analysis of the motor shaft through the linear regression model and the random forest model can be reduced. It is possible to provide a more accurate visual analysis of the jitter status and make adjustments more easily according to different jitter states of the motor shaft.

[0037] The present invention also provides an anti-shake calculation control system, which is applied to any one of the above-mentioned anti-shake calculation control methods for executing the terminal of a high-speed machine, comprising:

[0038] An image acquisition module is used to obtain trajectory image data of the motor shaft during operation;

[0039] A feature extraction module is used to identify feature points in trajectory image data using the Harris corner detection algorithm;

[0040] An adaptive control module is used to control the threshold of the feature points in the image data determined by the Harris corner detection algorithm, so that the Harris corner detection algorithm can re-identify the feature points in the extracted image data and construct the running trajectory of the motor shaft;

[0041] The vibration control module is used to determine the vibration state according to the running trajectory and to control the motor shaft to prevent vibration;

[0042] The jitter prediction module is used to determine the operating parameters of the motor shaft in the jitter state, and predict the jitter state of other motor shafts through the meta-model, thereby performing anti-jitter control on the motor shaft.

[0043] The present invention also provides a processing system for the anti-shake calculation and control method, comprising:

[0044] A memory and a processor, wherein the memory stores computer program instructions that can be loaded by the processor and execute any of the methods described above.

[0045] Compared with the known prior art, the technical solution provided by the present invention has the following beneficial effects:

[0046] By combining the Harris detection algorithm with the machine learning model, the feature points of the motor shaft trajectory image data can be accurately extracted and the Harris detection algorithm can be optimized, so that the subsequent operation trajectory constructed by the feature points is more accurate, which is convenient for analyzing the jitter condition of the motor shaft. By extracting the operating parameters from the operation trajectory, the associated prediction of the motor shaft jitter degree is re-made based on the operating parameters, so as to give a prediction of the motor shaft jitter degree in the subsequent time, so as to facilitate the suppression and adjustment of the motor shaft jitter. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 The figure is a schematic diagram of the control method of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] The present invention will be further described below in conjunction with the embodiments.

[0051] Example 1 (see Figure 1 ): A high-speed mechanical execution end anti-shake calculation and control method, including:

[0052] Obtain the trajectory image data of the motor shaft during operation;

[0053] The Harris corner detection algorithm is used to identify feature points in the trajectory image data, extract feature points to create a feature set, and mark the feature points in the feature set with features and normality to divide feature trajectory points and normal trajectory points. In this embodiment, the feature trajectory points are corner points determined by the Harris corner detection algorithm, and the normal trajectory points are feature points that are determined by the Harris corner detection algorithm as not being corner points. Thus, the running trajectory of the motor shaft can be constructed through the determined feature trajectory points. However, it is considered that there will be a certain inaccuracy in determining the feature trajectory points by the Harris corner detection algorithm, resulting in the neglect of some related feature trajectory points needed. Thus:

[0054] The feature points in the image data are predicted by the machine learning model, thereby adjusting the threshold of the feature points in the image data determined by the Harris corner detection algorithm, so that the Harris corner detection algorithm can re-identify the feature points in the extracted image data and construct the running trajectory of the motor shaft;

[0055] Determine the jitter state of the constructed motor shaft according to its running trajectory, and adjust the motor shaft for anti-jitter based on the jitter state. Learn and predict the feature trajectory points and normal trajectory points in the trajectory image data through the machine learning model, and determine which areas in the trajectory image data will have feature trajectory points through the machine learning model. Therefore, the normal trajectory points can be analyzed through the judgment of the machine learning model to avoid the Harris corner detection algorithm from demarcating possible feature trajectory points as normal trajectory points. After the prediction and judgment of the machine learning model, adjust the judgment threshold of the trajectory feature points in the area predicted by the machine learning model through the Harris corner detection algorithm, so that the Harris corner detection algorithm can extract the trajectory feature points more accurately, so as to facilitate the subsequent construction of the running trajectory of the motor shaft.

[0056] It should be noted that the operating parameters of the motor shaft in the shaking state are determined, and the shaking states of other motor shafts are predicted through the meta-model based on the operating parameters, and thus the motor shaft is subjected to anti-shake control; after the running trajectory of the motor shaft is constructed through the above-mentioned precise determination, the running trajectory of the motor shaft can be extracted to determine the shaking control trajectory matching the running trajectory, so as to realize the shaking analysis of the running trajectory of the motor shaft, thereby determining the shaking cause of the motor shaft, which is conducive to adjusting the shaking condition of the motor shaft. In this embodiment, the shaking state in the running trajectory of the motor shaft refers to the following;

[0057] According to the above, after determining the running trajectory of the motor shaft and judging the jitter state, the running parameters of the motor shaft in the jitter state are determined, and other motor shafts with the same state are determined through the running parameters. Based on the running parameters, a predictive analysis of the jitter state of the motor shaft under the running parameters is made through the meta-model to determine the jitter values ​​(referring to the severity of the jitter) of other motor shafts under the changed running parameters, so as to determine what kind of condition the jitter of the motor shaft will produce within a specified time, and then, to facilitate the subsequent adaptive control operation of the jitter of the motor shaft.

[0058] In summary, the above technical solution is further explained as follows:

[0059] In the above, when obtaining the trajectory image data of the motor shaft, the acquisition period can be set to collect the trajectory image data in different acquisition periods. For the motor shaft, it will be affected by many factors during its operation. Therefore, by defining different acquisition periods, the trajectory image data of the motor shaft at different times can be collected, so as to conduct a deep analysis of the jitter state of the motor shaft later.

[0060] Furthermore, the formula of Harris corner detection algorithm is as follows:

[0061] Determine the point in the image , determine its autocorrelation function:

[0062] ;

[0063] Determine the corner response function:

[0064] ;

[0065] Where: is the autocorrelation function of the image window, is the weight function, Indicates that the image is The pixel value of is the corner point response function, is a point in the image window, is a point outside the image window, is a point in the image, u and v are relative to The horizontal and vertical offsets of

[0066] when When , it is determined as a feature point, T is the determination threshold, and in this embodiment, a specific method for using the Harris corner point detection algorithm to determine feature points in trajectory image data is given, thereby facilitating more accurate extraction of feature points from trajectory image data.

[0067] The machine learning model is a support vector machine, and its algorithm formula is as follows:

[0068] ;

[0069] in, , represents 0 or 1, indicating that the feature point is a normal trajectory point or a feature trajectory point, w is the weight vector, b is the bias vector, is the regularization parameter, is the square of the Euclidean norm of the weight vector, represents the dot product between the weight vector and the feature points in the image, yes Loss function, is the feature point in the original image Kernel functions that map to high-dimensional space;

[0070] ;

[0071] in, is the output value of the decision function, according to , is the pixel threshold. According to the above, the pixel threshold is determined for each pixel point in the above image as a feature trajectory point through the decision function in the support vector machine. The corner points and non-corner points are distinguished according to the feature trajectory points and the normal trajectory points (the feature trajectory points are corner points, and the normal trajectory points are non-corner points). In this way, the area where the feature trajectory points exist can be predicted in the image of the motor shaft. When the subsequent image is extracted with the Harris corner point detection algorithm, the Harris algorithm is used to perform preliminary feature point detection in advance, that is, feature trajectory point detection, and all candidate normal trajectory points are recorded. Then, the support vector machine is used to determine whether the normal trajectory point is in the area where the feature trajectory point is predicted to exist. When it exists, the detection threshold of the Harris corner point detection algorithm in the area is adjusted to adjust the determination threshold of the Harris corner point detection algorithm, so that the Harris corner point detection algorithm can accurately detect and extract the feature trajectory points of the motor shaft in the above image, so as to facilitate the subsequent accurate construction of the running trajectory of the motor shaft.

[0072] Among them, when adjusting the decision threshold T in the Harris corner detection algorithm through the support vector machine, it is necessary to calculate the descriptor of the feature point, so it is as follows:

[0073] Determine the descriptor of the feature point:

[0074] ;

[0075] in, is the descriptor, p is the position of the corner point, e is the direction of the corner point, M and C are the sizes of the local area of ​​the image, It is the gradient strength of the local area, and the calculated descriptor is used as the input feature of the support vector machine, so as to facilitate training and classification by the support vector machine to determine the difference between feature trajectory points and normal trajectory points.

[0076] The jitter states include:

[0077] Amplitude jitter mode, waveform jitter mode, frequency jitter mode, phase jitter mode, where:

[0078] By creating a comparison image for different shaking modes in the shaking state and generating a comparison set, the comparison set is compared with the shaking state of the motor shaft to determine the shaking state of the motor shaft. In this embodiment, by creating a comparison image that is compared with the shaking state, and thus after obtaining the shaking state of the motor shaft, by inputting it into the comparison set, the shaking state of the motor shaft, that is, the shaking mode, can be quickly identified and analyzed, so as to clarify the cause of the shaking of the motor shaft. Generally speaking, the cause of the large-amplitude shaking mode is that the motor shaft has a major imbalance, excessive excitation force or insufficient structural stiffness, and the cause of the small-amplitude shaking mode is due to a slight imbalance. The waveform shaking mode is usually related to single-frequency or multi-frequency resonance for sinusoidal waveform vibration, which is due to the natural frequency or excitation frequency of the system. The jitter caused by the frequency, square wave or pulse waveform is related to the electrical operating characteristics of the motor, such as switching frequency or motor control strategy, random waveform vibration may be related to a variety of factors, including mechanical looseness, multi-frequency excitation or environmental interference; frequency jitter mode, including fundamental frequency jitter, which is usually caused by motor rotor imbalance, bearing failure or gear wear, etc., subharmonic jitter (integer multiples of the fundamental frequency) is caused by nonlinear elements (such as saturation or nonlinear stiffness) in the system, high-frequency jitter is related to the electrical components of the motor (such as motor windings, drivers), or is caused by system resonance; phase jitter mode, usually jitter caused by phase difference and phase jump, based on the above, the jitter cause caused by different jitter states can be determined, so as to form a suppression adjustment for the motor shaft jitter according to the corresponding jitter cause.

[0079] Specifically, the hybrid model is obtained by fusing the linear regression model and the random forest model, where the linear regression model is:

[0080] ;

[0081] Where: Y is the predicted jitter value, is the independent variable that affects the jitter value (i.e., operating parameters, which can be speed, load, temperature, lubrication, bearing loss, motor aging, humidity, and duration), is the intercept term, which represents the expected value of the jitter amplitude when all independent variables are zero. are regression coefficients, which represent the relationship between the respective variables and the jitter value, It is an error term. On the basis of the above, when multiple jitter modes are clarified and the jitter state of the motor shaft is determined, the operating parameters in each jitter mode are determined, and the jitter degree of the motor shaft under the jitter mode is predicted by a linear regression model by combining multiple independent variable operating parameters.

[0082] The random forest model is:

[0083] ;

[0084] in: is the predicted jitter value, is the average value of the motor shaft jitter predicted by all decision trees, is the jitter value predicted by the kth decision tree, is the weight of the kth decision tree;

[0085] Therefore, the meta-model after the fusion of linear regression model and random forest is:

[0086] ;

[0087] in: is the predicted jitter value after the linear regression model and random forest fusion, is the weight coefficient. The fused meta-model can accurately derive the subsequent jitter degree of other motor axes under the influencing factors. In this way, the jitter conditions of other motor axes under the jitter state can be clarified, so as to facilitate anti-jitter suppression control of other motor axes.

[0088] The present invention also provides an anti-shake calculation control system, which is applied to the anti-shake calculation control method for executing the terminal of the high-speed machinery, comprising:

[0089] An image acquisition module is used to obtain trajectory image data of the motor shaft during operation;

[0090] A feature extraction module is used to identify feature points in trajectory image data using the Harris corner detection algorithm;

[0091] An adaptive control module is used to control the threshold of the feature points in the image data determined by the Harris corner detection algorithm, so that the Harris corner detection algorithm can re-identify the feature points in the extracted image data and construct the running trajectory of the motor shaft;

[0092] The vibration control module is used to determine the vibration state according to the running trajectory and to control the motor shaft to prevent vibration;

[0093] The jitter prediction module is used to determine the operating parameters of the motor shaft in the jitter state, and predict the jitter state of other motor shafts through the meta-model, thereby performing anti-jitter control on the motor shaft.

[0094] The present invention also provides a processing system for the anti-shake calculation and control method, comprising:

[0095] A memory and a processor, wherein the memory stores computer program instructions that can be loaded by the processor and execute any of the methods described above.

[0096] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-speed mechanical execution end anti-shake calculation and control method, characterized in that: include: Obtain the trajectory image data of the motor shaft during operation; The Harris corner detection algorithm is used to identify the feature points in the trajectory image data, and the feature points are extracted to create a feature set. The feature points in the feature set are marked with features and normality to divide the feature trajectory points and normal trajectory points. The feature points in the image data are predicted by the machine learning model, thereby adjusting the threshold of the feature points in the image data determined by the Harris corner detection algorithm, so that the Harris corner detection algorithm can re-identify the feature points in the extracted image data and construct the running trajectory of the motor shaft; The vibration state of the motor shaft is determined according to the running trajectory of the constructed motor shaft, and the motor shaft is anti-vibration controlled based on the vibration state, wherein: Determine the operating parameters of the motor shaft in the shaking state, and predict the shaking states of other motor shafts through the meta-model based on the operating parameters, and thereby perform anti-shake control on the motor shafts; The formula of the Harris corner detection algorithm is as follows: Determine the point in the image , determine its autocorrelation function: ； Determine the corner response function: ； Where: is the autocorrelation function of the image window, is the weight function, Indicates that the image is The pixel value of is the corner point response function, is a point in the image window, is a point outside the image window, is a point in the image, and for relative to The horizontal and vertical offsets, is the output value of the decision function, is the input value of the decision function, where: when When , it is determined as a feature point. is the judgment threshold; The machine learning model is a support vector machine, and its algorithm formula is as follows: ； in, , Indicates 0 or 1. When it is 0, it means that the feature point is a normal trajectory point; when When it is 1, it means that the feature point is a feature trajectory point. is the weight vector, It is a bias top. is the regularization parameter, is the square of the Euclidean norm of the weight vector, is the weight vector, is the feature point in the original image Kernel functions that map to high-dimensional space; ； in, is the output value of the decision function, according to , is a pixel threshold. According to the above, a pixel threshold is determined for each pixel point in the above image as a feature trajectory point through a decision function in a support vector machine. Corner points and non-corner points are distinguished according to the feature trajectory points and normal trajectory points in the above. The feature trajectory points are corner points and the normal trajectory points are non-corner points. In this way, the area where the feature trajectory points exist can be predicted in the image of the motor shaft. When the subsequent image is extracted with the Harris corner point detection algorithm, a preliminary feature point detection is performed in advance with the Harris algorithm, that is, feature trajectory point detection, and all candidate normal trajectory points are recorded. Then, the support vector machine is used to determine whether the normal trajectory point is in the area where the feature trajectory point is predicted to exist. If so, the detection threshold of the Harris corner point detection algorithm in the area is adjusted to adjust the determination threshold of the Harris corner point detection algorithm, so that the Harris corner point detection algorithm can accurately detect and extract the feature trajectory points of the motor shaft, so as to facilitate the subsequent accurate construction of the running trajectory of the motor shaft. Among them, the decision threshold in the Harris corner detection algorithm is calculated by support vector machine. When making adjustments, it is necessary to calculate the descriptors of the feature points, so, as follows: Determine the descriptors of feature points: ； in, is the descriptor, is the position of the corner point, is the direction of the corner point, M and C are the sizes of the local area of ​​the image, It is the gradient strength of the local area, and the calculated descriptor is used as the input feature of the support vector machine, so as to facilitate training and classification by the support vector machine to determine the difference between feature trajectory points and normal trajectory points.

2. The high-speed mechanical execution end anti-shake calculation and control method according to claim 1 is characterized in that: When acquiring the trajectory image data of the motor shaft, a collection period is set to collect the trajectory image data in different collection periods.

3. The high-speed mechanical execution end anti-shake calculation and control method according to claim 1 is characterized in that: The jitter state includes: Amplitude jitter mode, waveform jitter mode, frequency jitter mode, phase jitter mode, where: By creating a comparison image for different shaking modes in the shaking state and generating a comparison set, the comparison set is compared with the shaking state of the motor shaft to determine the shaking state of the motor shaft.

4. The high-speed mechanical execution end anti-shake calculation and control method according to claim 1 is characterized in that: The meta-model is obtained by fusing the linear regression model and the random forest model, wherein the linear regression model is: ； in: is the predicted jitter value, is the independent variable that affects the jitter value, is the intercept term, which represents the expected value of the jitter amplitude when all independent variables are zero. are regression coefficients, which represent the relationship between the respective variables and the jitter value, is the error term.

5. Anti-shake calculation and control system, applied to the anti-shake calculation and control method for high-speed mechanical execution end described in claim 1, characterized in that: include: An image acquisition module is used to obtain trajectory image data of the motor shaft during operation; A feature extraction module is used to identify feature points in trajectory image data using the Harris corner detection algorithm; An adaptive control module is used to control the threshold of the feature points in the image data determined by the Harris corner detection algorithm, so that the Harris corner detection algorithm can re-identify the feature points in the extracted image data and construct the running trajectory of the motor shaft; The vibration control module is used to determine the vibration state according to the running trajectory and to control the motor shaft to prevent vibration; The jitter prediction module is used to determine the operating parameters of the motor shaft in the jitter state, and predict the jitter state of other motor shafts through the meta-model, thereby performing anti-jitter control on the motor shaft.

6. A processing system for an anti-shake calculation and control method, characterized in that: include: A memory and a processor, wherein the memory stores computer program instructions that can be loaded by the processor and execute the method according to any one of claims 1 to 4.

Images