Fault detection method for spindle gearbox of numerical control machine tool
By training the minimum volume coverage ellipsoid model and building an exception scoring function, the problem of lack of fault samples in fault detection of spindle transmission of CNC machine tool is solved, and accurate fault detection is achieved when only normal samples are available.
Patent Information
- Application Number
- CN202411920786.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-25
- Publication Date
- 2025-04-11
AI Technical Summary
CNC machine spindle transmission is prone to failure due to its high load and fast speed working state, but due to its high reliability in design, it is difficult to obtain fault samples, making it difficult for conventional machine learning methods to detect faults.
By obtaining the normal state vibration data of the spindle transmission of the CNC machine tool, preprocessing is performed to obtain the optimal feature subset, training the minimum volume coverage ellipsoid model (MVCE), obtaining the ellipsoid center vector and scatter matrix, and constructing an exception scoring function to determine whether there is a fault in the test sample.
It realizes accurate detection of faults of CNC machine spindle transmission when only normal samples exist, avoiding dependence on abnormal samples and improving the reliability of fault detection.
Smart Images
Figure CN119807963A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of fault detection, and in particular to a method for detecting a fault of a spindle gearbox of a CNC machine tool. Background Art
[0002] The spindle gearbox is an important component of CNC machine tool equipment that transmits power. Due to the harsh and changeable operating environment, its components are often in a high-load, high-speed working state, which is prone to various abnormal problems, such as gear wear, bearing damage, shaft imbalance, etc. It is necessary to timely and accurately detect the faults of the spindle gearbox of CNC machine tools to avoid abnormal deterioration or serious failures.
[0003] The CNC machine tool spindle gearbox fault detection method in the related art usually adopts a machine learning method, using a large number of normal samples and a large number of fault samples to train a fault detection model for fault judgment.
[0004] However, CNC machine tools are usually designed with high reliability, and the probability of spindle gearbox failure is low, resulting in only normal samples but no fault samples in the samples obtained from the spindle gearbox of CNC machine tools. In the absence of fault samples, conventional machine learning methods cannot be used for fault detection. Summary of the invention
[0005] The present disclosure provides a method for detecting a fault in a spindle gearbox of a CNC machine tool, which can accurately detect a fault in a spindle gearbox of a CNC machine tool when only normal samples are present. The technical solution at least includes the following solutions: In a first aspect, a method for detecting a fault of a spindle gearbox of a CNC machine tool is provided, comprising: acquiring a plurality of vibration data of a spindle gearbox of a CNC machine tool, wherein the plurality of vibration data are vibration data of the spindle gearbox of the CNC machine tool in a normal state; preprocessing the plurality of vibration data to obtain an optimal feature subset; training a minimum volume covering ellipsoid model MVCE based on the optimal feature subset to obtain an ellipsoid center vector and a scatter matrix of the MVCE; acquiring a test sample, wherein the test sample is real-time vibration data of the spindle gearbox of the CNC machine tool; constructing an anomaly scoring function based on the ellipsoid center vector and the scatter matrix; and determining an anomaly score of the test sample based on the anomaly scoring function, wherein the anomaly score is used to indicate whether the spindle gearbox of the CNC machine tool has a fault.
[0006] Optionally, based on the optimal feature subset, the minimum volume covering ellipsoid model MVCE is trained to obtain the ellipsoid center vector and scatter matrix of the MVCE, including: constructing the objective function of the MVCE; using the Lagrange multiplier method to transform the solution of the objective function of the MVCE into a dual problem; solving the dual problem to obtain the Lagrange multiplier vector; based on the Lagrange multiplier vector, determining the ellipsoid center vector and scatter matrix of the MVCE.
[0007] Optionally, the abnormality scoring function is expressed by the following formula:
[0008] in, is the anomaly scoring function, For the test sample, is the ellipsoid center vector, is the scatter matrix.
[0009] Optionally, the preprocessing of the multiple vibration data to obtain the optimal feature subset includes: respectively calculating the time domain features, frequency domain features and time-frequency domain features of the multiple vibration data to obtain N types of features of each vibration data, constructing the features of the same type into a feature vector to obtain N feature vectors; normalizing the N feature vectors to obtain a normalized multi-domain feature set; obtaining the Laplace score of each feature vector in the normalized multi-domain feature set, and arranging the N feature vectors from small to large according to the Laplace score to obtain a first arrangement; constructing the first k feature vectors in the first arrangement into a first vector set; calculating the mutual correlation coefficient of each feature vector in the first vector set, and screening out m feature vectors from the first vector set based on the mutual correlation coefficient, and the m feature vectors are the feature vectors in the optimal feature subset; wherein N, k, and m are all integers, N is greater than k, and k is greater than m.
[0010] Optionally, the calculating the mutual correlation coefficient of each eigenvector in the first vector set and selecting m eigenvectors from the first vector set based on the mutual correlation coefficient includes: obtaining the average absolute mutual correlation coefficient of each eigenvector in the first vector set; and removing a eigenvector with the largest average absolute mutual correlation coefficient from the first vector set.
[0011] Optionally, obtaining the average absolute mutual correlation coefficient of each feature vector in the first vector set includes: respectively calculating the Pearson correlation coefficient between the i-th feature vector in the first vector set and other feature vectors except the i-th feature vector, to obtain multiple Pearson correlation coefficients of the i-th feature vector; and obtaining the average absolute mutual correlation coefficient of the i-th feature based on the multiple Pearson correlation coefficients of the i-th feature vector.
[0012] In the second aspect, a CNC machine tool spindle gearbox fault detection device is also provided, including: a first acquisition module, used to acquire multiple vibration data of the CNC machine tool spindle gearbox, and the multiple vibration data are the vibration data of the CNC machine tool spindle gearbox in a normal state; a preprocessing module, used to preprocess the multiple vibration data to obtain an optimal feature subset; a training module, used to train a minimum volume covering ellipsoid model MVCE based on the optimal feature subset to obtain the ellipsoid center vector and scatter matrix of the MVCE; a second acquisition module, used to acquire a test sample, and the test sample is the real-time vibration data of the CNC machine tool spindle gearbox; a scoring function construction module, used to construct an abnormal scoring function based on the ellipsoid center vector and the scatter matrix; a scoring module, used to determine the abnormal score of the test sample based on the abnormal scoring function, and the abnormal score is used to indicate whether the CNC machine tool spindle gearbox has a fault.
[0013] Optionally, the training module is also used to construct the objective function of the MVCE; use the Lagrange multiplier method to transform the solution of the objective function of the MVCE into a dual problem; solve the dual problem to obtain a Lagrange multiplier vector; based on the Lagrange multiplier vector, determine the ellipsoid center vector and scatter matrix of the MVCE.
[0014] Optionally, in the scoring function building module, the abnormal scoring function is expressed by the following formula:
[0015] in, is the anomaly scoring function, For the test sample, is the ellipsoid center vector, is the scatter matrix.
[0016] Optionally, the preprocessing module is also used to respectively calculate the time domain features, frequency domain features and time-frequency domain features of the multiple vibration data to obtain N types of features for each vibration data, construct features of the same type into a feature vector to obtain N feature vectors; normalize the N feature vectors to obtain a normalized multi-domain feature set; obtain the Laplace score of each feature vector in the normalized multi-domain feature set, and arrange the N feature vectors from small to large according to the Laplace score to obtain a first arrangement; construct the first k feature vectors in the first arrangement into a first vector set; calculate the mutual correlation coefficient of each feature vector in the first vector set, and filter out m feature vectors from the first vector set based on the mutual correlation coefficient, and the m feature vectors are the feature vectors in the optimal feature subset; wherein N, k, and m are all integers, N is greater than k, and k is greater than m.
[0017] Optionally, the preprocessing module is further used to obtain an average absolute cross-correlation coefficient of each eigenvector in the first vector set; and remove an eigenvector with the largest average absolute cross-correlation coefficient from the first vector set.
[0018] Optionally, the preprocessing module is also used to calculate the Pearson correlation coefficient between the i-th eigenvector in the first vector set and other eigenvectors except the i-th eigenvector, respectively, to obtain multiple Pearson correlation coefficients of the i-th eigenvector; based on the multiple Pearson correlation coefficients of the i-th eigenvector, obtain the average absolute correlation coefficient of the i-th feature.
[0019] In a third aspect, a computer device is also provided, comprising: a memory and a processor, wherein at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor, thereby executing the CNC machine tool spindle gearbox fault detection method described in the above embodiment.
[0020] In a fourth aspect, a computer-readable storage medium is also provided, wherein at least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by a processor, thereby executing the CNC machine tool spindle gearbox fault detection method described in the above embodiment.
[0021] In a fifth aspect, a computer program product is provided, comprising a computer program / instruction, wherein when the computer program / instruction is executed by a processor, the method described in the first aspect is implemented.
[0022] The beneficial effects brought by the technical solution provided by the embodiments of the present disclosure include at least: In the disclosed embodiment, by obtaining multiple vibration data of the CNC machine tool spindle gearbox under normal conditions, and using the multiple vibration data to train the MVCE, the obtained MVCE can represent the minimum range of normal vibration data. The subsequent abnormal scoring function constructed by the ellipsoid center vector and scatter matrix of the MVCE is essentially to calculate the distance between the input test sample and the MVCE, and the distance can indicate whether the input test sample is within the MVCE range. If the input test sample is within the MVCE range, it means that the test sample belongs to normal vibration data, that is, the CNC machine tool spindle gearbox is fault-free; if the input test sample is outside the MVCE range, it means that the test sample belongs to abnormal vibration data, that is, the CNC machine tool spindle gearbox has a fault, thus achieving accurate CNC machine tool spindle gearbox fault detection. And the whole process only involves vibration data samples under normal conditions, and does not involve samples under abnormal conditions, that is, it can accurately detect CNC machine tool spindle gearbox faults when only normal samples exist. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0024] Figure 1 A flow chart of a method for detecting a spindle gearbox fault of a CNC machine tool provided by an exemplary embodiment of the present disclosure is shown; Figure 2 A flow chart of a method for detecting a fault of a spindle gearbox of a CNC machine tool provided by another exemplary embodiment of the present disclosure is shown; Figure 3 A schematic structural diagram of a CNC machine tool spindle gearbox fault detection device provided by an exemplary embodiment of the present disclosure is shown; Figure 4 It is a schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION
[0025] Unless otherwise defined, the technical or scientific terms used herein shall have the usual meanings understood by persons of ordinary skill in the field to which the present disclosure belongs. The words "first", "second", "third" and similar words used in the patent application specification and claims of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate a quantitative limitation, but rather indicate the presence of at least one. Words such as "include" or "comprise" and similar words mean that the elements or objects appearing before "include" or "comprises" cover the elements or objects listed after "include" or "comprises" and their equivalents, and do not exclude other elements or objects.
[0026] In order to make the objectives, technical solutions and advantages of the present disclosure more clear, the embodiments of the present disclosure will be further described in detail below with reference to the accompanying drawings.
[0027] Figure 1 A flowchart of a method for detecting a fault in a spindle gearbox of a CNC machine tool provided by an exemplary embodiment of the present disclosure is shown. The method can be executed by a computer device. Figure 1 , the method comprising: In step 101, a plurality of vibration data of a spindle gearbox of a CNC machine tool is obtained.
[0028] Optionally, any vibration data is 3-channel vibration data, wherein the 3-channel vibration data includes horizontal radial vibration data, vertical radial vibration data and axial vibration data of the CNC machine tool spindle gearbox housing. The 3-channel vibration data can more comprehensively reflect the vibration condition of the CNC machine tool spindle gearbox.
[0029] When acquiring vibration data of three channels, three acceleration sensors can be arranged in the horizontal radial direction, vertical radial direction and axial direction of the CNC machine tool spindle gearbox housing. Although there is a single acceleration sensor that can acquire vibration data of three channels, such acceleration sensors are relatively expensive. In the disclosed embodiment, by arranging three acceleration sensors at three positions, it is possible to acquire vibration data of three channels simply and at low cost.
[0030] Here, the vibration data acquired in step 101 are all vibration data of the CNC machine tool spindle gearbox in a normal state, and the plurality of vibration data are all historical vibration data of the CNC machine tool spindle gearbox.
[0031] In step 102, a plurality of vibration data are preprocessed to obtain an optimal feature subset.
[0032] In step 103, the MVCE is trained based on the optimal feature subset to obtain the ellipsoid center vector and scatter matrix of the MVCE.
[0033] The Minimum Volume Covering Ellipsoid (MVCE) model is an optimization model. The optimal feature subset is the optimal feature in normal vibration data. Therefore, the MVCE trained based on the optimal feature subset can represent the minimum range of normal vibration data.
[0034] In step 104, a test sample is obtained.
[0035] The test sample is the real-time vibration data of the CNC machine tool spindle gearbox.
[0036] Here, the test samples obtained in step 104 also need to be preprocessed in the same manner as in step 102 before being scored in step 106 .
[0037] In step 105, an anomaly scoring function is constructed based on the ellipsoid center vector and the scatter matrix.
[0038] The input of the anomaly scoring function is the preprocessed test sample, and the output is the score. The essence of the anomaly scoring function is to calculate the distance between the input test sample and the MVCE, which can indicate whether the input test sample is within the MVCE range. If the input test sample is within the MVCE range, it means that the test sample is normal vibration data, that is, the CNC machine tool spindle gearbox is fault-free; if the input test sample is outside the MVCE range, it means that the test sample is abnormal vibration data, that is, the CNC machine tool spindle gearbox is faulty.
[0039] In step 106 , an anomaly score of the test sample is determined based on the anomaly scoring function.
[0040] The anomaly score is used to indicate whether there is a fault in the CNC machine tool spindle gearbox.
[0041] In the disclosed embodiment, by obtaining multiple vibration data of the CNC machine tool spindle gearbox under normal conditions, and using the multiple vibration data to train the MVCE, the obtained MVCE can represent the minimum range of normal vibration data. The subsequent abnormal scoring function constructed by the ellipsoid center vector and scatter matrix of the MVCE is essentially to calculate the distance between the input test sample and the MVCE, and the distance can indicate whether the input test sample is within the MVCE range. If the input test sample is within the MVCE range, it means that the test sample belongs to normal vibration data, that is, the CNC machine tool spindle gearbox is fault-free; if the input test sample is outside the MVCE range, it means that the test sample belongs to abnormal vibration data, that is, the CNC machine tool spindle gearbox has a fault, thus achieving accurate CNC machine tool spindle gearbox fault detection. And the whole process only involves vibration data samples under normal conditions, and does not involve samples under abnormal conditions, that is, it can accurately detect CNC machine tool spindle gearbox faults when only normal samples exist.
[0042] Figure 2 A flowchart of a method for detecting a fault in a spindle gearbox of a CNC machine tool provided by another exemplary embodiment of the present disclosure is shown. The method can be executed by a computer device. Figure 2 , the method comprising: In step 201, a plurality of vibration data of a spindle gearbox of a CNC machine tool is obtained.
[0043] The relevant contents of step 201 refer to the aforementioned step 101, and the detailed description is omitted here.
[0044] In step 202, a plurality of vibration data are preprocessed to obtain an optimal feature subset.
[0045] Optionally, step 202 includes the following steps ae.
[0046] Step a, respectively calculating the time domain features, frequency domain features and time-frequency domain features of multiple vibration data to obtain N types of features for each vibration data, constructing the features of the same type into a feature vector to obtain N feature vectors.
[0047] Wherein, N is a positive integer.
[0048] Optionally, when the vibration data includes vibration data of 3 channels, 29 types of features can be extracted from the vibration data of each channel, that is, a total of 87 types of features can be extracted from one vibration data (equivalent to N being 87). For ease of understanding, the following description is based on the case where N is 87.
[0049] The following is an explanation of the characteristics of the 29 types.
[0050] There are 14 types of time domain features, including 5 dimensional features: mean, standard deviation, square root amplitude, root mean square value and peak value, 6 dimensionless features: waveform index, peak index, pulse index, margin index, skewness and kurtosis, and 3 nonlinear features: approximate entropy, sample entropy and fuzzy entropy. There are many calculation methods for these 14 types of time domain features in related technologies, so they are omitted here.
[0051] There are 5 types of frequency domain features, including centroid frequency, mean square frequency, frequency standard deviation, frequency skewness and frequency kurtosis. When calculating frequency domain features, you first need to calculate the power spectrum of the vibration data of a certain channel, and then calculate these 5 types of frequency domain features based on the power spectrum. There are many implementation methods for calculating these 5 types of frequency domain features based on the power spectrum in related technologies, so detailed description is omitted here.
[0052] There are a total of 10 types of time-frequency domain features. Before calculating the time-frequency domain features of the vibration data of a certain channel, it is necessary to first perform a three-layer wavelet packet decomposition transform on the vibration data of the channel to obtain 8 frequency components, and then calculate the wavelet packet energy spectrum features of these 8 frequency components respectively to obtain 8 wavelet packet energy spectrum features (that is, 8 types of time-frequency domain features are obtained). Then calculate the energy entropy features and singular entropy features of these 8 frequency components as a whole to obtain the energy entropy features and singular entropy features of the vibration data of the channel (that is, 2 types of time-frequency domain features are obtained). There are many methods for calculating wavelet packet decomposition transform, wavelet packet energy spectrum features, energy entropy features and singular entropy features in the related technology, so detailed description is omitted here.
[0053] After 29 types of features are calculated for the vibration data of the three channels of a certain vibration data, it is equivalent to obtaining three 1×29 vectors. By concatenating these three vectors, 87 types of features of the vibration data can be obtained.
[0054] The vibration data of the three channels of any vibration data are calculated in the above manner, and 87 types of features of each vibration data can be obtained. In the case of n vibration data, these n vibration data can be expressed as a matrix ,in That is the pth vibration data, The dimension is 87 times 1, which means that the p-th vibration data has 87 types of features. p is an integer, and the value range of p is 1 to n.
[0055] In step 201, a total of n vibration data are obtained, and each vibration data has 87 types of features. The same type of features are constructed into a feature vector, that is, the matrix Each row of is a feature vector, so 87 feature vectors can be obtained, each of which has n dimensions.
[0056] Step b: normalize the N feature vectors to obtain a normalized multi-domain feature set.
[0057] Optionally, the Z-score method is used to normalize the N feature vectors to obtain a normalized multi-domain feature set. There are many implementations of the Z-score method in the related art, and detailed description is omitted here.
[0058] Normalization does not change the dimension of the feature vector, so the normalized multi-domain feature set also includes 87 feature vectors, which can be expressed in matrix form To represent the 87 feature vectors in the normalized multi-domain feature set.
[0059] Although the dimensions of these 87 eigenvectors are the same, the dimensions of these 87 eigenvectors are obviously different. Some eigenvectors even have no dimensions, so they need to be normalized before subsequent calculations.
[0060] Step c: obtaining the Laplace score of each feature vector in the normalized multi-domain feature set, and arranging the N feature vectors from small to large according to the Laplace score to obtain a first arrangement.
[0061] There are many methods for calculating Laplace scores in the related art, and detailed description is omitted here.
[0062] After calculating the Laplace scores of the 87 eigenvectors respectively, the 87 eigenvectors can be sorted in ascending order of the Laplace scores to obtain the first arrangement.
[0063] Step d: construct the first k feature vectors in the first arrangement into a first vector set.
[0064] Here, k is an integer and k is less than N. The value of k can be set according to requirements, and the embodiment of the present disclosure does not limit this, and it only needs to satisfy that k is less than N.
[0065] The 87 eigenvectors in the first arrangement are arranged in ascending order. The first k eigenvectors in the first arrangement in step d are actually the first k eigenvectors with smaller Laplace scores, and the other 87-k eigenvectors with larger Laplace scores are eliminated. A larger Laplace score indicates that the eigenvector is a redundant vector, which has little effect on fault prediction and needs to be eliminated.
[0066] Through step d, the first screening of 87 feature vectors is achieved, thereby filtering out some features that have a smaller impact on fault prediction and retaining features that have a greater impact on fault prediction.
[0067] Step e: calculating the mutual correlation coefficient between each eigenvector in the first vector set and other eigenvectors, and screening k eigenvectors based on the mutual correlation coefficients to obtain m eigenvectors.
[0068] These m feature vectors are the feature vectors in the optimal feature subset.
[0069] Wherein, m is an integer and k is greater than m. The value of m can be set according to requirements, and the embodiment of the present disclosure does not limit this, and it only needs to satisfy that k is greater than m.
[0070] Optionally, step e includes the following steps e1-e2: Step e1, obtaining the average absolute cross-correlation coefficient of each eigenvector in the first vector set.
[0071] Optionally, for the i-th eigenvector in the first vector set in step e1, the average absolute cross-correlation coefficient of the i-th eigenvector is calculated using the following two steps.
[0072] In the first step, the Pearson correlation coefficients between the ith eigenvector in the first vector set and other eigenvectors except the ith eigenvector are calculated respectively to obtain multiple Pearson correlation coefficients of the ith eigenvector.
[0073] Optionally, the Pearson correlation coefficient between the ith eigenvector and the jth eigenvector is calculated using formula (1), where the jth eigenvector is any eigenvector in the first vector set except the ith eigenvector.
[0074] (1) In formula (1), is the i-th eigenvector and the jth eigenvector The Pearson correlation coefficient between them; n is an integer, which represents the number of multiple vibration data obtained in step 201. middle, represents the pth component of the i-th eigenvector, that is, the component of the p-th vibration data in the i-th eigenvector, p is an integer, and the value range of p is 1 to n. middle, It represents the pth component of the jth eigenvector, that is, the component of the pth vibration data in the jth eigenvector, p is an integer, and the value range of p is 1 to n. is the mean of the i-th eigenvector, ; is the mean of the jth eigenvector, .
[0075] The above formula (1) can be used to calculate the Pearson correlation coefficient between the ith eigenvector and each eigenvector except the ith eigenvector, thereby obtaining multiple Pearson correlation coefficients of the ith eigenvector. When there are k eigenvectors in the first vector set, k-1 Pearson correlation coefficients can be calculated for the ith eigenvector.
[0076] In the second step, based on multiple Pearson correlation coefficients of the ith eigenvector, the average absolute cross-correlation coefficient of the ith feature is obtained.
[0077] Optionally, the average absolute cross-correlation coefficient of the i-th eigenvector is calculated using formula (2).
[0078] (2) In formula (2), is the average absolute correlation coefficient of the i-th eigenvector, and s represents the number of eigenvectors in the first vector set except the i-th eigenvector. In the first iteration, s is equal to k-1. As the iterations progress, the number of eigenvectors in the first vector set decreases, and the value of s also decreases. The meanings of other parameters in formula (2) are the same as those in formula (1), and their detailed description is omitted here.
[0079] By performing the first step to the second step on each eigenvector in the first vector set, the average absolute cross-correlation coefficient of each eigenvector in the first vector set can be obtained.
[0080] Step e2: remove the eigenvector with the largest average absolute cross-correlation coefficient from the first vector set.
[0081] Optionally, formula (3) is used to determine the maximum average absolute cross-correlation coefficient.
[0082] (3) In formula (3), The meanings of other parameters in formula (3) are the same as those in formula (2), and their detailed description is omitted here.
[0083] The eigenvector corresponding to the largest average absolute cross-correlation coefficient is a eigenvector in the optimal feature subset, which can be expressed as After determining a After that, you need to remove it from the first vector set (At this time, the total number of feature vectors in the first vector set is reduced by 1), and then the above steps a to d are repeated until the total number of feature vectors in the first vector set is equal to m. At this time, the first vector set is the optimal feature subset.
[0084] Here, if the average absolute correlation coefficient of a feature vector is the largest, it means that the feature vector is redundant with respect to other feature vectors and can be eliminated. By continuously eliminating redundant vectors, the feature vector in the optimal feature subset that is finally retained is the feature vector that is most relevant to fault detection.
[0085] Here, each feature vector is actually a type of feature, and the m feature vectors in the optimal feature subset indicate m types of features. When preprocessing the test sample later, only the m types of features of the test sample need to be calculated, without calculating 87 types of features.
[0086] Through step e, the vectors in the first vector set are screened, which is equivalent to a second screening of the 87 feature vectors. In this way, the feature vectors with less impact on fault prediction are effectively filtered, and the feature vectors most relevant to fault detection are retained.
[0087] Through steps ae, two feature selections are performed by combining Laplace scores and cross-correlation, so that the most discriminative and least redundant features can be selected. This method does not rely on data labels and human experience, and can construct the optimal feature set under unsupervised conditions.
[0088] In step 203, the MVCE is trained based on the optimal feature subset to obtain the ellipsoid center vector and scatter matrix of the MVCE.
[0089] Optionally, step 203 includes the following step fi.
[0090] Step f, construct the objective function of MVCE.
[0091] Optionally, the objective function of MVCE is expressed using formula (4).
[0092] (4) In formula (4), is the scatter matrix, is the ellipsoid center vector, is the penalty factor, is the slack variable. , represents the pth vibration data vector in the optimal feature subset. n and p are integers, the value range of p is 1 to n, and n represents the number of the multiple vibration data obtained in step 201.
[0093] Here, after the optimal feature subset is filtered to only m feature vectors, it is equivalent to filtering out m feature types that are most relevant to fault prediction, and these m feature vectors can be restored to multiple vibration data. For example, the matrix to represent the m eigenvectors in the optimal feature subset. Then the matrix In the above formula, it means that the characteristic components belonging to the same vibration data are constructed as one vibration data, that is, the matrix Each column of is constructed as a vibration data, thereby obtaining n vibration data, each vibration data having m dimensions. These n vibration data can be expressed in the following matrix form: .
[0094] Optionally, an intermediate variable can be introduced and , to simplify formula (4). , , then formula (4) can be simplified to formula (5).
[0095] (5) The meanings of the parameters in formula (5) are the same as those in formula (4), and their detailed description is omitted here.
[0096] Step g: Use the Lagrange multiplier method to transform the solution of the objective function of MVCE into a dual problem.
[0097] Optionally, the transformed dual problem is expressed using formula (6).
[0098] (6) In formula (6), , represents the matrix in formula (4) The augmented dataset of , is a vector with n elements and all n elements are 1; is the Lagrange multiplier vector, Indicated by The resulting diagonal matrix. m represents the number of eigenvectors in the optimal feature subset, that is, the matrix The number of elements in any one vibration data.
[0099] Step h, solve the dual problem and obtain the Lagrange multiplier vector.
[0100] The dual problem can be solved based on the solution of the SDP (Semidefinite Programming) problem to obtain the Lagrange multiplier vector. In implementation, the solution of the SDP problem in the open source package CVX can be used to solve the dual problem.
[0101] Step i, based on the Lagrange multiplier vector, determine the ellipsoid center vector and scatter matrix of the MVCE.
[0102] Optionally, step i is expressed using formula (7).
[0103] (7) The meanings of the parameters in formula (7) are the same as those in formulas (4) to (6), and their detailed description is omitted here.
[0104] In step 204, a test sample is obtained.
[0105] The test sample is the real-time vibration data of the spindle gearbox of the CNC machine tool.
[0106] For the test sample, it is only necessary to calculate the m types of features included in the optimal feature subset in step 202.
[0107] In step 205, an anomaly scoring function is constructed based on the ellipsoid center vector and the scatter matrix.
[0108] Optionally, the anomaly scoring function is expressed using formula (8).
[0109] (8) In formula (8), is the anomaly scoring function, is the test sample. The meanings of other parameters in formula (8) are the same as those in formula (4), and their detailed description is omitted here.
[0110] In step 205, based on the anomaly scoring function, an anomaly score of the test sample is determined.
[0111] The anomaly score is used to indicate whether there is a fault in the CNC machine tool spindle gearbox.
[0112] Before executing step 205, the test samples need to be normalized. The normalization process is expressed by formula (9).
[0113] (9) In formula (9), is the normalized test sample, For the test sample, For the matrix The mean vector of multiple vibration data in , For the matrix The variance vector of multiple vibration data in .
[0114] After the test sample is normalized, the normalized test sample can be input into the anomaly scoring function to obtain the score output by the anomaly scoring function for the test sample.
[0115] In the disclosed embodiment, for a test sample input into the abnormal scoring function, if the score output by the abnormal scoring function is greater than 0, it means that the test sample is outside the MVCE range and is abnormal vibration data, that is, there is a fault in the spindle gearbox of the CNC machine tool; if the score output by the abnormal scoring function is less than or equal to 0, it means that the test sample is within the MVCE range and is normal vibration data, that is, there is no fault in the spindle gearbox of the CNC machine tool.
[0116] The following are device embodiments of the present application. For details not described in detail in the device embodiments, reference may be made to the above method embodiments.
[0117] Figure 3 FIG. 1 is a schematic diagram showing a structure of a CNC machine tool spindle gearbox fault detection device provided by an exemplary embodiment of the present disclosure. Figure 3 The CNC machine tool spindle gearbox fault detection device 300 includes: a first acquisition module 301, a preprocessing module 302, a training module 303, a second acquisition module 304, a scoring function construction module 305 and a scoring module 306.
[0118] The first acquisition module 301 is used to acquire a plurality of vibration data of a spindle gearbox of a CNC machine tool, wherein the plurality of vibration data are vibration data of the spindle gearbox of the CNC machine tool in a normal state.
[0119] The preprocessing module 302 is used to preprocess the multiple vibration data to obtain the optimal feature subset.
[0120] The training module 303 is used to train the minimum volume covering ellipsoid model MVCE based on the optimal feature subset, and obtain the ellipsoid center vector and scatter matrix of the MVCE.
[0121] The second acquisition module 304 is used to acquire a test sample, where the test sample is real-time vibration data of a spindle gearbox of a CNC machine tool.
[0122] The scoring function construction module 305 is used to construct an abnormal scoring function based on the ellipsoid center vector and the scatter matrix.
[0123] The scoring module 306 is used to determine the abnormality score of the test sample based on the abnormality scoring function, and the abnormality score is used to indicate whether there is a fault in the spindle gearbox of the CNC machine tool.
[0124] Optionally, the training module 303 is also used to construct the objective function of the MVCE; use the Lagrange multiplier method to transform the solution of the objective function of the MVCE into a dual problem; solve the dual problem to obtain the Lagrange multiplier vector; based on the Lagrange multiplier vector, determine the ellipsoid center vector and scatter matrix of the MVCE.
[0125] Optionally, in the scoring function building module, the anomaly scoring function is expressed by the following formula:
[0126] in, is the anomaly scoring function, For the test sample, is the ellipsoid center vector, is the scatter matrix.
[0127] Optionally, the preprocessing module 302 is also used to respectively calculate the time domain features, frequency domain features and time-frequency domain features of multiple vibration data to obtain N types of features for each vibration data, construct features of the same type into a feature vector to obtain N feature vectors; normalize the N feature vectors to obtain a normalized multi-domain feature set; obtain the Laplace score of each feature vector in the normalized multi-domain feature set, and arrange the N feature vectors in ascending order according to the Laplace score to obtain a first arrangement; construct the first k feature vectors in the first arrangement into a first vector set; calculate the mutual correlation coefficient of each feature vector in the first vector set, and filter out m feature vectors from the first vector set based on the mutual correlation coefficient, and the m feature vectors are the feature vectors in the optimal feature subset; wherein N, k, and m are all integers, N is greater than k, and k is greater than m.
[0128] Optionally, the preprocessing module 302 is further configured to obtain an average absolute cross-correlation coefficient of each eigenvector in the first vector set; and remove an eigenvector with the largest average absolute cross-correlation coefficient from the first vector set.
[0129] Optionally, the preprocessing module 302 is also used to calculate the Pearson correlation coefficient between the i-th eigenvector in the first vector set and other eigenvectors except the i-th eigenvector, respectively, to obtain multiple Pearson correlation coefficients of the i-th eigenvector; based on the multiple Pearson correlation coefficients of the i-th eigenvector, obtain the average absolute correlation coefficient of the i-th feature.
[0130] It should be noted that: when the CNC machine tool spindle gearbox fault detection device provided in the above embodiment performs CNC machine tool spindle gearbox fault detection, only the division of the above functional modules is used as an example. In actual application, the above functional distribution can be completed by different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the CNC machine tool spindle gearbox fault detection device provided in the above embodiment and the CNC machine tool spindle gearbox fault detection method embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.
[0131] The division of modules in the embodiments of the present disclosure is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present disclosure may be integrated into a processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.
[0132] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a terminal device (which can be a personal computer, mobile phone, or communication device, etc.) or a processor (processor) to execute all or part of the steps of the method of each embodiment of the present disclosure. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.
[0133] Figure 4 Schematic diagram of the structure of the computer device provided by the embodiment of the present disclosure. Figure 4 As shown, the computer device 400 includes: a processor 401 and a memory 402 .
[0134] The processor 401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 401 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 401 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 401 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 401 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.
[0135] The memory 402 may include one or more computer-readable storage media, which may be non-transitory. The memory 402 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 402 is used to store at least one instruction, which is used to be executed by the processor 401 to implement the CNC machine tool spindle gearbox fault detection method provided in the embodiment of the present disclosure.
[0136] Those skilled in the art will understand that Figure 4 The structure shown in the figure does not constitute a limitation on the computer device 400, and the computer device 400 may include more or less components than those shown in the figure, or combine some components, or adopt a different arrangement of components.
[0137] The embodiment of the present disclosure also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by the processor of a computer device, the computer device can execute the CNC machine tool spindle gearbox fault detection method provided in the embodiment of the present disclosure.
[0138] The embodiments of the present disclosure also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the CNC machine tool spindle gearbox fault detection method provided in the embodiments of the present disclosure.
[0139] The above description is only an optional embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.
Claims
1. A method for detecting a fault in a spindle gearbox of a CNC machine tool, characterized in that: The method comprises: Acquire a plurality of vibration data of a spindle gearbox of a CNC machine tool, wherein the plurality of vibration data are vibration data of the spindle gearbox of the CNC machine tool in a normal state; Preprocessing the plurality of vibration data to obtain an optimal feature subset; Based on the optimal feature subset, a minimum volume covering ellipsoid model MVCE is trained to obtain an ellipsoid center vector and a scatter matrix of the MVCE; Acquire a test sample, wherein the test sample is real-time vibration data of the spindle gearbox of the CNC machine tool; Constructing an anomaly scoring function based on the ellipsoid center vector and the scatter matrix; Based on the anomaly scoring function, an anomaly score of the test sample is determined, and the anomaly score is used to indicate whether there is a fault in the CNC machine tool spindle gearbox.
2. The method according to claim 1, characterized in that The method of training a minimum volume coverage ellipsoid model MVCE based on the optimal feature subset to obtain an ellipsoid center vector and a scatter matrix of the MVCE includes: Constructing the objective function of the MVCE; The solution of the objective function of the MVCE is transformed into a dual problem by using the Lagrange multiplier method; Solving the dual problem to obtain a Lagrange multiplier vector; Based on the Lagrange multiplier vector, the ellipsoid center vector and the scatter matrix of the MVCE are determined.
3. The method according to claim 1, characterized in that The anomaly scoring function is expressed by the following formula: in, is the anomaly scoring function, For the test sample, is the ellipsoid center vector, is the scatter matrix.
4. The method according to claims 1 to 3, characterized in that The preprocessing of the plurality of vibration data to obtain an optimal feature subset includes: Respectively calculating the time domain features, frequency domain features, and time-frequency domain features of the plurality of vibration data to obtain N types of features of each vibration data, and constructing features of the same type into a feature vector to obtain N feature vectors; Normalizing the N feature vectors to obtain a normalized multi-domain feature set; Obtaining the Laplace score of each feature vector in the normalized multi-domain feature set, and arranging the N feature vectors from small to large according to the Laplace scores to obtain a first arrangement; Constructing the first k feature vectors in the first arrangement into a first vector set; Calculating the mutual correlation coefficient of each feature vector in the first vector set, and selecting m feature vectors from the first vector set based on the mutual correlation coefficient, wherein the m feature vectors are feature vectors in the optimal feature subset; Wherein, N, k, and m are all integers, N is greater than k, and k is greater than m.
5. The method according to claim 4, characterized in that The calculating the mutual correlation coefficient of each feature vector in the first vector set, and selecting m feature vectors from the first vector set based on the mutual correlation coefficient, includes: Obtaining the average absolute cross-correlation coefficient of each eigenvector in the first vector set; The eigenvector with the largest average absolute cross-correlation coefficient is removed from the first vector set.
6. The method according to claim 5, characterized in that The obtaining of the average absolute correlation coefficient of each eigenvector in the first vector set comprises: Respectively calculating the Pearson correlation coefficients between the ith eigenvector in the first vector set and other eigenvectors except the ith eigenvector, to obtain a plurality of Pearson correlation coefficients of the ith eigenvector; Based on the multiple Pearson correlation coefficients of the i-th feature vector, an average absolute cross-correlation coefficient of the i-th feature is obtained.
7. A CNC machine tool spindle gearbox fault detection device, characterized in that: The device comprises: A first acquisition module is used to acquire a plurality of vibration data of a spindle gearbox of a CNC machine tool, wherein the plurality of vibration data are vibration data of the spindle gearbox of the CNC machine tool in a normal state; A preprocessing module, used for preprocessing the plurality of vibration data to obtain an optimal feature subset; A training module, used for training a minimum volume covering ellipsoid model MVCE based on the optimal feature subset, to obtain an ellipsoid center vector and a scatter matrix of the MVCE; A second acquisition module is used to acquire a test sample, wherein the test sample is real-time vibration data of the spindle gearbox of the CNC machine tool; A scoring function construction module, used to construct an abnormal scoring function based on the ellipsoid center vector and the scatter matrix; The scoring module is used to determine the abnormality score of the test sample based on the abnormality scoring function, and the abnormality score is used to indicate whether the spindle gearbox of the CNC machine tool has a fault.
8. A computer device, characterized in that: The computer device comprises: a memory and a processor, wherein at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one computer program, and the at least one computer program is loaded and executed by a processor to implement the method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.
Citation Information
Cited By
Barrier equipment real-time state monitoring system and method based on Internet of Things
CN122310149A