Big Data-Based Stepper Motor Life Prediction and Analysis Method and System

Abstract

The present invention relates to the technical field of data processing. More specifically, the present invention relates to a method and system for predicting and analyzing the service life of a stepping motor based on big data. The method includes: obtaining a plurality of operation data sequences during the operation of the motor, and dividing each operation data sequence into a plurality of data segments with the same length based on a preset length; constructing a regression model for each operation data through support vector regression, and predicting the service life of the motor according to the predicted values output by all regression models; during the construction process of the regression model, taking the adjustment factor of each data segment of any operation data as a weight, weighting the insensitive loss of the corresponding data segment, and performing data regression according to the weighted insensitive loss. The present invention can improve the accuracy of the constructed regression model and realize accurate prediction of the service life of the motor.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing. More specifically, the present invention relates to a method and system for predicting the service life of a stepping motor based on big data. Background Art

[0002] Due to its high precision, good controllability, and excellent reliability, stepping motors are widely used in various precision machinery. However, during long-term high-load operation, stepping motors are faced with problems such as wear, aging, and potential failures. These factors not only affect the operating efficiency of the equipment but also may pose a threat to production safety. In order to improve the accuracy and effectiveness of maintenance, by using big data analysis technology to monitor the operating status of the motor in real time and combining with a prediction model to evaluate the motor life, potential failures can be effectively predicted, the maintenance timing can be optimized, thereby improving the reliability, production efficiency, and preventive maintenance level of the equipment.

[0003] In related technologies, for example, in the patent application document with the publication number CN111382546A, a method for predicting the service life of a generator insulation system based on support vector machine modeling is disclosed. The method includes: taking the aging time as the input to establish six prediction models for non-destructive electrical parameters of generator stator bars respectively. Taking every five aging cycles as a unit, predict the six non-destructive electrical parameters in turn, and then bring the results into the established life prediction model of the remaining breakdown field strength of the generator insulation based on support vector machine until the prediction result is lower than the critical remaining field strength to obtain the generator life.

[0004] However, when constructing the support vector machine regression model in the above solution, the influence of data with different importance levels is not considered, and it is impossible to make important data have a higher fitting accuracy, resulting in a lower accuracy of the constructed support vector machine regression model, a lower accuracy of the predicted value determined based on the support vector machine regression model, and thus it is impossible to accurately predict the life of the generator. Summary of the Invention

[0005] In order to solve the problem that the motor life cannot be accurately predicted due to the low accuracy of the prediction model, the present invention provides a method and system for predicting the service life of a stepping motor based on big data.

[0006] According to the first aspect of the present invention, a method for predicting the service life of a stepping motor based on big data is provided, including:

[0007] Obtain multiple operation data sequences during the operation of the motor, and divide each operation data sequence into several data segments with the same length based on a preset length;

[0008] Construct a regression model for each item of operation data through support vector regression, and predict the motor life according to the predicted values output by all regression models; in the process of constructing the regression model, use the adjustment factor of each data segment of any item of operation data as the weight, weight the insensitive loss of the corresponding data segment, and perform data regression according to the weighted insensitive loss; the method for obtaining the adjustment factor includes:

[0009] Use any data segment as the test set for cross-validation, obtain the residuals of each data in this data segment, and calculate the degree of fluctuation of this data segment: ; is the degree of fluctuation of the th data segment of the th item of operation data; is the standard deviation of this data segment; is the average standard deviation of all data segments of this item of operation data; is the residual of the th data in this data segment; is the data volume of this data segment; is the natural exponential function;

[0010] Quantify the difference between the ratio of the average residual of this data segment and the corresponding data segment in the remaining item of operation data during the corresponding period and the ratio of the overall residuals of the corresponding two items of operation data, obtain the abnormality degree of this data segment, and calculate the adjustment factor of this data segment. The adjustment factor is positively correlated with both the degree of fluctuation and the abnormality degree.

[0011] When constructing a regression model for each item of operation data through support vector regression, the present invention can use the adjustment factor of the data segment of each item of operation data to adaptively adjust the insensitive loss of the corresponding data segment, so that the data segment with a smaller adjustment factor passes through a lower insensitive loss, improving the regression accuracy, and the data segment with a larger adjustment factor passes through a higher insensitive loss, reducing the influence of the regression result of the corresponding data segment on the overall regression result, ensuring the accuracy of the constructed regression model, and thus enabling accurate prediction of the motor life.

[0012] Preferably, the abnormality degree satisfies the following relational expression:

[0013] ;

[0014] In the formula, is the abnormality degree of the th data segment of the th item of operation data; is the average residual of the th data segment of the th item of operation data; is the th item of operation data of the The average residual of each data segment; is the average residual of all data segments of the th item of operation data; is the average residual of all data segments of the th item of operation data; is the total number of items of operation data.

[0015] When calculating the degree of abnormality, the present invention utilizes the feature that the regression effects between data segments with the same ordinal number in each item of operation data are similar, so as to accurately evaluate the possibility of abnormal data existing in each data segment.

[0016] Preferably, when constructing the regression model of each item of operation data through support vector regression, the regression label of each regression model is the remaining life of the motor.

[0017] The present invention can predict the motor life through different types of operation data, thus ensuring the accuracy of the predicted motor life.

[0018] Preferably, when performing cross-validation of support vector regression by using any data segment of any item of operation data as the test set, the method further includes:

[0019] Using the remaining all data segments of this item of operation data as the training set, and determining the cross-validation result of the test set with the regression model trained based on the training set.

[0020] Preferably, the method for obtaining the adjustment factor includes:

[0021] Obtaining the degree of fluctuation and the degree of abnormality of any data segment, and performing multiplication operation on the degree of fluctuation and the degree of abnormality of this data segment to obtain the adjustment factor of this data segment.

[0022] The present invention can set a larger adjustment factor for data segments with larger degrees of fluctuation and abnormality, so as to reduce the influence of abnormal data by increasing the insensitive loss of the corresponding data segments.

[0023] Preferably, the multiple operation data sequences include a current data sequence, a temperature data sequence, a rotational speed data sequence, a vibration intensity data sequence, and an output torque data sequence.

[0024] Preferably, predicting the motor life according to the predicted values output by all regression models includes:

[0025] Obtaining all the constructed regression models, inputting the current, temperature, rotational speed, vibration intensity, and output torque at the current moment during the motor operation into the corresponding regression models, and outputting the predicted values of each item of operation data;

[0026] The average value of the motor life determined based on the predicted values of each item of operating data is used as the final motor life.

[0027] The present invention can improve the accuracy of the constructed regression model, thereby achieving accurate prediction of the motor life.

[0028] According to the second aspect of the present invention, there is provided a stepping motor life prediction and analysis system based on big data. The system includes a memory and a processor. A computer program is stored on the memory, and the processor executes the computer program to implement the steps of the first aspect of the present invention.

[0029] The present invention has the following effects:

[0030] 1. When constructing the regression model of each item of operating data through support vector regression, the present invention can adaptively adjust the insensitive loss of each data segment of each item of operating data, so that the data segment with a larger adjustment factor has a larger sensitive loss, thereby reducing the influence of abnormal data, and the data segment with a smaller adjustment factor has a smaller sensitive loss, thereby improving the regression accuracy of the corresponding data segment, ensuring the accuracy of the constructed regression model, and thus accurately predicting the motor life.

[0031] 2. The present invention synthesizes data from multiple aspects and calculates the adjustment factors of each data segment, ensuring the accuracy of the adjustment factors, and thus enabling precise self-adaptation to the insensitive loss during the construction of the regression model. Description of the Drawings

[0032] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0033] Figure 1 It is a schematic flowchart of the steps of the method for predicting and analyzing the life of a stepping motor based on big data according to an embodiment of the present invention. Detailed Embodiments

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] The following will describe in detail the specific embodiments of the present invention with reference to the accompanying drawings.

[0036] Refer to Figure 1, A method for predicting the service life of a stepping motor based on big data, including steps S1 - S2, specifically as follows:

[0037] S1: Obtain multiple operation data sequences during the operation of the motor, and divide each operation data sequence into several data segments with the same length based on a preset length.

[0038] In an exemplary embodiment of the present invention, the multiple operation parameter sequences include a current data sequence, a temperature data sequence, a rotation speed data sequence, a vibration intensity data sequence, and an output torque data sequence.

[0039] It should be noted that the current, temperature, rotation speed, vibration intensity, and output torque during the operation of the motor can reflect the operation stability of the stepping motor. Therefore, the data that can reflect the operation stability of the stepping motor can be used as the data basis to predict the service life of the stepping motor. Of course, appropriate types of data can also be selected according to specific situations. This embodiment does not make special limitations on the selected data types, as long as they can reflect the operation stability of the stepping motor.

[0040] Specifically, within a certain sampling time period (such as within one month), at a certain sampling frequency (such as 1 time / hour), collect the current, temperature, rotation speed, vibration intensity, and output torque during the operation of the motor to obtain multiple operation data sequences during the operation of the motor. This embodiment does not make special limitations on the sampling time period and the sampling frequency.

[0041] Optionally, the current data during the operation of the motor can be collected by an ammeter arranged in the motor operation power supply system, the temperature data during the operation of the motor can be collected by a thermometer arranged around the motor, the rotation speed data during the operation of the motor can be measured by a tachometer, the vibration intensity data during the operation of the motor can be collected by a vibration sensor, and the output torque during the operation of the motor can be obtained by a torque sensor. This embodiment does not make special limitations on the installation positions and types of the devices for collecting data.

[0042] Optionally, the preset length can be denoted as , and then use the preset length Starting from the first data of each operation data sequence, intercept data segments with a length of to divide each operation data sequence into multiple data segments with the same length, and obtain multiple data segments of each operation data. In this embodiment = 10. This embodiment does not make special limitations on the size of the preset length.

[0043] It should be noted that the data segments with the same ordinal number in each operation data correspond to the same sampling time period.

[0044] S2: Construct a regression model for each item of operating data through support vector regression, and predict the motor life according to the predicted values output by all regression models; during the construction of the regression model, use the adjustment factor of each data segment of any item of operating data as the weight, weight the insensitive loss of the corresponding data segment, and perform data regression according to the weighted insensitive loss.

[0045] In an exemplary embodiment of the present invention, when constructing a regression model for each item of operating data through support vector regression, the regression label of each regression model is the remaining life of the motor.

[0046] It should be noted that when predicting the motor life through support vector regression, the traditional support vector regression algorithm usually directly uses all the collected data for regression modeling to obtain the prediction of the motor life by the regression model. However, during the construction of the regression model, directly using all the data for regression does not fully consider the importance of some important features in the data, resulting in a low fitting accuracy of key data, which in turn affects the overall accuracy of the regression model and makes it impossible to accurately predict the motor life based on the obtained regression model. Therefore, the present invention improves the process of constructing the regression model through support vector regression. The specific improvement content is as follows: using the calculated adjustment factor to adaptively adjust the insensitive loss of each data segment of any item of operating data, so that the data segments with lower adjustment factors can obtain higher regression accuracy through lower insensitive loss, thereby being able to construct a regression model with higher accuracy based on the data regression results of each item of operating data and achieve accurate prediction of the motor life.

[0047] Among them, the insensitive loss (epsilon-insensitive loss) is a component in the support vector regression (SVR) algorithm, which defines the tolerated error range. In the insensitive loss function, only when the error between the predicted value and the actual value exceeds the specified tolerated range (i.e., the insensitive loss), will the loss be penalized. That is, the lower the insensitive loss, the higher the regression accuracy.

[0048] Specifically, the determination of the adjustment factor of any data segment can be achieved through the following steps:

[0049] Step 1: Use any data segment as a test set for cross-validation, obtain the residuals of each data in this data segment, and calculate the degree of fluctuation of this data segment;

[0050] It should be noted that the result of cross-validation of any data segment of any operating data as a test set can reflect the degree of fit between the data segment and all other data segments of the operating data. The data segment with a higher degree of fit indicates that it performs better in the regression model, and the fluctuation of the data will affect the fitting effect of the data, which is specifically manifested as: the more drastic the fluctuation of the data in the data segment, the greater the residual of each data in the data segment. Therefore, the present invention utilizes this feature to evaluate the fluctuation of the corresponding data segment by obtaining the residual of each data in each data segment. Among them, the process of cross-validation of support vector regression based on the test set is a prior art, and this embodiment will not be described in detail here.

[0051] In an exemplary embodiment of the present invention, when any data segment of any item of running data is used as a test set for cross-validation of support vector regression, the method further includes:

[0052] All remaining data segments of the running data are used as training sets to determine the cross-validation results of the test set based on the regression model trained on the training set.

[0053] Furthermore, after determining the residuals of each data in each data segment, the degree of fluctuation of any data segment can be calculated. Specifically, the degree of fluctuation satisfies the following relationship:

[0054] ;

[0055] In the formula, For the The first item of operation data The degree of fluctuation of each data segment; is the standard deviation of the data segment; The average standard deviation of all data segments for this run; When the data segment is used as a test set, the first The residual of the data; is the amount of data in this data segment; is a natural exponential function, where the natural exponential function refers to a function with a natural constant An exponential function with base .

[0056] in, Reflects the The first item of operation data The standard deviation of the data segment is The ratio of the average standard deviation of all data in the data segment. The larger the value, the more drastic the fluctuation of the data in the data segment, and the greater the fluctuation degree of the corresponding data segment.

[0057] Reflects the The first item of operation data The normalized value of the residual of the th data in a data segment; It reflects the sum of the normalized residuals when the th item of operating data's th data segment is used as the test set for cross-validation. The larger this value, the worse the fit of the data in this data segment to all other data segments of this item of operating data, that is, the worse the performance of this data segment in the regression model, and further it can indicate that the data in this data segment fluctuates more violently, corresponding to a larger degree of fluctuation of this data segment.

[0058] Optionally, when the degree of fluctuation of any data segment is large, the possibility of abnormal data existing in this data segment is relatively large. Therefore, when performing support vector regression based on this data segment, a larger insensitive loss should be set to avoid using too precise regression for this data segment and affecting the accuracy of the overall regression model.

[0059] Step 2: Quantify the difference between the ratio of the average residual of this data segment and the data segments in the corresponding time period of the remaining item of operating data and the ratio of the overall residuals of the corresponding two items of operating data to obtain the degree of abnormality of this data segment;

[0060] It should be noted that for each data segment of each item of operating data, curve regression can be performed on the motor life according to each item of operating data, and the regression results obtained according to each item of operating data should be similar. If there is a significant difference in the regression effect between any data segment of any item of operating data and the remaining data segments in the corresponding time period, then there is a high possibility that abnormal data exists in this any data segment. Therefore, the present invention utilizes this feature to calculate the degree of abnormality of each data segment.

[0061] Specifically, the degree of abnormality of any data segment satisfies the following relational expression:

[0062] ;

[0063] In the formula, is the degree of abnormality of the th data segment of the th item of operating data; is the average residual of the th data segment of the th item of operating data; is the average residual of the th data segment of the th item of operating data; is the average residual of all data segments of the th item of operating data; is the average residual of all data segments of the th item of operating data; is the total number of items of operation data.

[0064] Among them, reflects the ratio of the average residual of the th item of operation data to the th item of operation data for the th data segment; reflects the ratio of the overall residual of the th item of operation data to the th item of operation data. When the value of is relatively large, it indicates that has a relatively large difference from , and further indicates that the relationship between the th data segments of these two items of operation data does not match the overall relationship of these two items of operation data.

[0065] quantifies the difference between the ratio of the average residual of the th data segment of the th item of operation data and the th data segment of other items of data, and the ratio of the overall residual of the corresponding two items of operation data. The larger this value is, the greater the difference in the regression effect between the th data segment of the th item of operation data and the th data segment of other items of operation data. Further, it indicates that the th data segment of the th item of operation data is likely to have abnormal data, and the degree of abnormality of the corresponding data segment is relatively large.

[0066] Optionally, when the degree of abnormality of any data segment of any item of operation data is relatively large, a relatively large insensitive loss should be set when performing support vector regression based on this data segment to reduce the impact of abnormal data on the regression result.

[0067] In another embodiment, it is also possible to quantify the difference between the ratio of the average residual of any data segment of any item of operation data and the data segment of the corresponding time period in the remaining items of operation data, and the ratio of the overall residual of the corresponding two items of operation data in other ways. For example, it can be quantified by taking the absolute value and then averaging.

[0068] Step 3: Calculate the adjustment factor of this data segment. The adjustment factor is positively correlated with both the degree of fluctuation and the degree of abnormality.

[0069] In an exemplary embodiment of the present invention, the determination of the adjustment factor of any data segment can be achieved through the following steps:

[0070] Obtain the fluctuation degree and abnormality degree of any data segment of any item of operation data, perform a multiplication operation on the fluctuation degree of this data segment and the abnormality degree of this data segment to obtain the adjustment factor of this data segment.

[0071] Optionally, when the fluctuation degree of any data segment of any item of operation data is large and the abnormality degree of this data segment is also large, it indicates that the data in this data segment deviates greatly from all other data of this item of operation data. Setting a large adjustment factor can appropriately increase the insensitive loss of this data segment, thereby preventing the result of support vector regression based on this data segment from affecting the overall regression accuracy.

[0072] In another embodiment, the adjustment factor of each data segment of each item of operation data can also be calculated through other calculation methods, such as being determined through a summation calculation method instead of a multiplication calculation method.

[0073] Furthermore, the determination method of the adjustment factor can be used to calculate the adjustment factor of each data segment of each item of operation data, and then the adjustment factor of each data segment of any item of operation data is used as a weight to weight the insensitive loss of the corresponding data segment to obtain the insensitive loss of each data segment of each item of operation data. In this embodiment, the initial value of the insensitive loss of all data segments is , and the initial value of the insensitive loss of each data segment in this embodiment is not particularly limited.

[0074] Next, the process of constructing a regression model for each item of operation data through support vector regression will be described:

[0075] First, based on the fluctuation degree and abnormality degree of each data segment of each item of operation data, calculate the adjustment factor of the corresponding data segment, and then adaptively obtain the adjusted insensitive loss of each data segment of each item of operation data based on the adjustment factor.

[0076] After that, all data segments of any item of operation data are used as a training set, the remaining motor life corresponding to each data in each data segment is used as a regression label, and the adjusted insensitive loss of each data segment is used as an input feature to perform data regression based on the training set, regression label, and input feature to obtain the regression model of this item of operation data, and then determine the regression models of all items of operation data.

[0077] It should be noted that the process of constructing a regression model for data through support vector regression is a prior art, and this embodiment will not elaborate on it here.

[0078] Further, after obtaining the regression model for each item of operation data, the motor life can be predicted based on the predicted values output by all the regression models. In an exemplary embodiment of the present invention, the prediction of the motor life can be achieved through the following steps:

[0079] Obtain all the constructed regression models, input the current moment current, temperature, rotation speed, vibration intensity, and output torque during the motor operation into the corresponding regression models, and output the predicted motor life for each item of operation data; take the average value of the predicted motor life based on each item of operation data as the final life of the motor.

[0080] The present invention can adaptively adjust the insensitive loss when performing support vector regression for different data segments, enabling data segments without abnormal data to obtain higher regression accuracy through a lower insensitive loss, and enabling data segments with abnormal data to reduce the impact of abnormal data on the regression result through a higher insensitive loss, ensuring the accuracy of the constructed regression model, and thus enabling accurate prediction of the motor life.

[0081] The present invention also provides a stepping motor life prediction and analysis system based on big data. The system includes a memory and a processor, and a computer program is stored on the memory. The computer program integrates the functions of the stepping motor life prediction and analysis method based on big data. When the computer program is executed, the accuracy of the constructed regression model can be improved through the stepping motor life prediction and analysis method based on big data, and accurate prediction of the motor life can be achieved.

[0082] In the description of this specification, the meanings of "a plurality of" and "several" are at least two, such as two, three, or more, etc., unless otherwise clearly and specifically defined.

[0083] Although this specification has shown and described multiple embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will think of many changes, alterations, and alternative ways without departing from the spirit and concept of the present invention. It should be understood that various alternative solutions to the embodiments of the present invention described herein can be adopted in the practice of the present invention.

Claims

1. A stepper motor life prediction and analysis method based on big data, characterized in that: include: Acquire multiple operation data sequences during the operation of the motor, and divide each operation data sequence into a number of data segments of the same length based on a preset length; A regression model for each operating data is constructed by support vector regression, and the motor life is predicted according to the predicted values ​​output by all regression models; in the process of constructing the regression model, the adjustment factor of each data segment of any operating data is used as a weight, the insensitive loss of the corresponding data segment is weighted, and data regression is performed according to the weighted insensitive loss; the method for obtaining the adjustment factor includes: Use any data segment as a test set for cross-validation, obtain the residuals of each data in the data segment, and calculate the degree of fluctuation of the data segment: ; For the The first item of operation data The degree of fluctuation of each data segment; is the standard deviation of the data segment; The average standard deviation of all data segments for this run; For the data segment The residual of the data; is the amount of data in this data segment; is the natural exponential function; Quantify the difference between the ratio of the average residuals of the data segment and the data segment of the corresponding period in the remaining operation data and the ratio of the overall residuals of the corresponding two operation data to obtain the abnormality of the data segment and calculate the adjustment factor of the data segment. The adjustment factor is positively correlated with the fluctuation degree and the abnormality degree. The abnormality degree satisfies the following relationship: ； In the formula, For the The first item of operation data The degree of abnormality of each data segment; For the The first item of operation data The average residual of the data segment; For the The first item of operation data The average residual of the data segment; For the The average residual of all data segments of the running data; For the The average residual of all data segments of the running data; is the total number of running data items.

2. The stepper motor life prediction and analysis method based on big data according to claim 1 is characterized in that: When constructing a regression model for each operating data item through support vector regression, the regression label of each regression model is the remaining life of the motor.

3. The stepper motor life prediction and analysis method based on big data according to claim 1 is characterized in that: When any data segment of any item of running data is used as a test set for cross-validation of support vector regression, the method further includes: All remaining data segments of the running data are used as training sets, so as to determine the cross-validation result of the test set based on the regression model trained by the training set.

4. The stepper motor life prediction and analysis method based on big data according to claim 1 is characterized in that: The method for obtaining the adjustment factor includes: The fluctuation degree and abnormality degree of any data segment of any operation data are obtained, and the fluctuation degree of the data segment and the abnormality degree of the data segment are multiplied to obtain the adjustment factor of the data segment.

5. The stepper motor life prediction and analysis method based on big data according to claim 1 is characterized in that: The plurality of operation data sequences include a current data sequence, a temperature data sequence, a rotation speed data sequence, a vibration intensity data sequence and an output torque data sequence.

6. The stepper motor life prediction and analysis method based on big data according to claim 5 is characterized in that: The motor life is predicted based on the predicted values ​​output by all regression models, including: Obtain all the regression models constructed, input the current, temperature, speed, vibration intensity and output torque of the motor at the current moment during operation into the corresponding regression model, and output the motor life predicted by each operation data; The average value of the motor life predicted based on each operating data is taken as the final life of the motor.

7. The stepper motor life prediction and analysis system based on big data is characterized by: The stepper motor life prediction and analysis system based on big data includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of the stepper motor life prediction and analysis method based on big data as described in any one of claims 1-6.

Images