CN114358059A - Screw rod part qualification inspection method based on moment singularity analysis - Google Patents
Screw rod part qualification inspection method based on moment singularity analysis Download PDFInfo
- Publication number
- CN114358059A CN114358059A CN202111554836.2A CN202111554836A CN114358059A CN 114358059 A CN114358059 A CN 114358059A CN 202111554836 A CN202111554836 A CN 202111554836A CN 114358059 A CN114358059 A CN 114358059A
- Authority
- CN
- China
- Prior art keywords
- screw rod
- moment
- method based
- characteristic
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012797 qualification Methods 0.000 title claims abstract description 40
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000004458 analytical method Methods 0.000 title claims abstract description 24
- 238000007689 inspection Methods 0.000 title claims abstract description 21
- 239000013598 vector Substances 0.000 claims abstract description 31
- 238000013145 classification model Methods 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 18
- 238000012706 support-vector machine Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 17
- 230000006835 compression Effects 0.000 claims abstract description 12
- 238000007906 compression Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 7
- 239000006185 dispersion Substances 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 241001598107 Imperata Species 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000012850 discrimination method Methods 0.000 claims 1
- 230000017105 transposition Effects 0.000 claims 1
- 238000004422 calculation algorithm Methods 0.000 abstract description 3
- 230000004044 response Effects 0.000 abstract description 2
- 238000001514 detection method Methods 0.000 description 27
- 238000002474 experimental method Methods 0.000 description 12
- 238000004088 simulation Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 6
- 238000013016 damping Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000035772 mutation Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000011946 reduction process Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a screw rod part qualification inspection method based on moment singularity analysis, which comprises the following steps of: s1, obtaining a screw rod rotation moment through the tail end of the robot; s2, denoising the torque signal; s3, extracting singularity features and statistical features of the moment signals; s4, performing feature compression processing; and S5, inputting the training sample consisting of the processed characteristic vectors into a support vector machine classification model for training to obtain a classification model capable of detecting the product qualification. Compared with the wavelet theory algorithm, the method has the advantages of giving consideration to quick response, having better robust performance, and not needing a large number of training samples and training time.
Description
Technical Field
The invention belongs to the field of product qualification inspection, and particularly relates to a screw rod part qualification inspection and analysis method based on moment singularity analysis.
Background
With the increasing demand of enterprises for automation, robots are more and more widely applied to various industrial scenes. Traditional product qualification nature detects and mostly relies on the manual work to detect, and the robot can use and makes detection efficiency and rate of accuracy promote greatly. In the qualification inspection of screw rod type parts, the screw rod type parts are twisted and detected by a robot, and whether the parts are qualified or not is judged based on a torque signal acquired from the tail end of the robot.
The common methods for extracting the singular characteristics of the signal include wavelet transformation and intelligent detection. Wavelet transform is considered as the most important tool for detecting signal changes due to the characteristic of reflecting local characteristics of signals. Mallat applies wavelet transform to singularity detection of signals for the first time, and lays a theoretical foundation for singularity analysis of detection signals. Because wavelet transform has a series of problems of spectrum aliasing, large calculation amount, offset defect, difficulty in selecting wavelet base and the like, researchers successively improve wavelet transform from the wavelet theory itself. The learner eliminates the frequency band dislocation phenomenon by exchanging the order of the nodes at the even number positions, namely the two nodes after wavelet packet decomposition, and introduces two operators to respectively remove the frequency components outside the ideal passband range of the high-frequency sub-band and the low-frequency sub-band so as to eliminate the frequency band overlapping phenomenon. The scholars provide a signal change point positioning method based on second-generation wavelet transform and multi-level hypothesis test, and the signal change point can be quickly positioned by only needing fewer samples. Some researchers have proposed a joint positioning method combining correlation and wavelet transform modulo maximum to reduce the positioning error. Although the wavelet theory is optimized to a certain extent by the research method, the offset defect of wavelet transformation cannot be completely eliminated, and the calculation process is complex, so that the system cannot be stopped accurately and timely, and further damage to products is caused.
In order to further improve the accuracy and rapidity of the singularity detection of the signal, a relevant learner develops a series of research works by using an intelligent algorithm. Some researchers provide a detection method based on a convolutional neural network, and the convolutional neural network model is utilized to directly learn useful information for detection from a working signal, so that the process of manually preprocessing original signal data in the traditional method can be omitted, end-to-end detection is realized, and the detection accuracy rate exceeds 93%. The researchers have proposed a low latency detection method based on long short term memory network (LSTM), which has smaller storage space occupation and lower computation latency. The learners provide a deep neural network recognition algorithm based on a stack sparse autoencoder and Softmax, the stack sparse autoencoder can effectively extract high-dimensional features of working signals, the built deep neural network can effectively recognize the leakage state of a hydraulic pump, and the recognition accuracy reaches 97.6%. Although the research method is greatly improved in the aspects of data preprocessing, detection precision and the like compared with the wavelet theory, a large amount of sample data is needed for model training, and the early working time cost is greatly increased.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a screw rod part qualification inspection and analysis method based on moment singularity analysis.
In order to achieve the purpose of the invention, the screw rod part qualification inspection and analysis method based on moment singularity analysis provided by the invention comprises a robot and a control cabinet thereof, an industrial personal computer, a driver and a simulation detection experiment platform, and comprises the following steps:
s1, adjusting the pose of the robot to enable the tail end of the robot to be aligned and matched with the screw rod screwing end on the simulation detection experiment platform, and smoothly screwing the screw rod when the system is started;
s2, preprocessing the torque signal acquired in the step 1, and performing real-time noise reduction processing on the torque signal subjected to sliding window iteration by using a wavelet threshold noise reduction method based on bayes;
s3, calculating and extracting singularity characteristics and statistical characteristics of the noise-reduced moment signals in the step 2, wherein the statistical characteristics comprise a maximum value, a minimum value, an average value, a skewness, a kurtosis and a standard deviation;
s4, performing feature compression processing on the features obtained by calculation in the step 3 to obtain final feature vectors formed by the features with high correlation;
s5, inputting the training sample composed of the feature vectors processed in the step 4 into a classification model of a support vector machine for training to obtain a classification model capable of detecting the product qualification, and using the classification model to realize the qualification inspection of the screw parts.
Further, the step S1 specifically includes the following steps:
s11, inputting the coordinates of the screw rod screwing end on the simulation detection experiment platform into the robot, and adjusting the position and the posture of the robot to enable the tail end of the robot to be matched with the screw rod screwing end on the simulation detection experiment platform in an aligning manner;
s12, starting the qualification testing robot system, driving the screw rod on the simulation detection experiment platform to rotate by the robot end device, and sending a torque signal acquired by the servo motor to the industrial personal computer in real time.
Further, the step S2 specifically includes the following steps:
s21, carrying out sliding window iterative preprocessing on the collected torque signal, and sliding the window every 1msAdding new data of 1ms and removing the data added in the earliest 1ms, and keeping the real-time performance of the data;
s22, carrying out real-time noise reduction processing on the torque signal after the sliding window iteration by using a wavelet threshold noise reduction method based on bayes, removing high-frequency noise and avoiding excessive interference on an effective signal part.
Further, the step S22 specifically includes:
s221, in order to weaken the boundary effect generated in the noise reduction process, the sliding window is subjected toObtaining W by symmetric continuation processingτ(i);
S222, for the Wτ(i) Wavelet transformation is carried out to obtain wavelet coefficientsj,kWτ(i) Further, the variance of the noisy signal is obtained:
where j is the scale and k is time.
S223, the collected moment signal is composed of an original signal and noise, and the variance of the noise is estimated as:
wherein, mean (& gtY & ltj,kWτ(i) |) is the median of the absolute values of the wavelet coefficients in the wavelet decomposition node. The variance σ of the original signals(j, k, i) is:
s224, further, we calculate a Bayes threshold:
further, the step S3 specifically includes:
s31, calculating a Lipschitz index of the noise-reduced torque signal in the step S2, wherein the index is a singular characteristic value of the torque signal;
and S32, extracting statistical characteristics of the moment signal and the Lipschitz index to form a characteristic vector TZ, wherein the statistical characteristics comprise a maximum value, a minimum value, an average value, a skewness, a kurtosis and a standard deviation.
Further, step S4 specifically includes:
s41, the feature compression is to perform feature compression processing on the features TZ obtained by the calculation in step S3 by using a Fisher discriminant method, where the Fisher discriminant index is:
wherein n is the serial number of the characteristic element, SBnIs the dispersion between n features; SWnIs the dispersion within the feature, M is the number of classes of the feature,is of class LpThe mean value of the medium-sized features n,andare calculated by the following formulas, respectively:
wherein L ispIs the p-th class of feature elements, x is the class LpCharacteristic of (1)pIs the number of feature element samples in the type. Sorting the calculated Fisher discrimination indexes according to the magnitude of the numerical values, and taking the characteristic elements with the top rank to form a vector
S42, sorting the calculated Fisher discrimination indexes, wherein the characteristic with high correlation is the characteristic with the top rank.
Further, the step S5 specifically includes:
s51, forming a sample set and a test set by the processed feature vectors in the step S4;
and S52, inputting the sample set and the test set into a Support Vector Machine (SVM) classification model for training to obtain a classification model capable of detecting the product qualification.
Further, the step S52 specifically includes:
s521, the Support Vector Machine (SVM) model can classify normal moment signals and abrupt moment signals by defining an optimal hyperplane as a division plane of a moment signal characteristic sample, and the hyperplane equation is as follows:
where ω is the hyperplane normal vector and b is the displacement term of the hyperplane.
S522, each sample corresponds to a label yiE { -e, e }, wherein e, -e represent different sample classes, respectively, if yiWhen is + e, then there isIf yiWhen is-e, then there isLet the equation:
s523, a sample for establishing the equation in step S522 is referred to as a support vector, and the sum of the distances from the two heterogeneous support vectors to the hyperplane is referred to as a distance γ:
s524, solving the optimal hyperplane needs to satisfy the following equation:
s525, constructing an objective function through a Lalang Ri multiplier method, solving through Sequence Minimum Optimization (SMO), and finally determining the optimal hyperplane as follows:
wherein, w*Is a normal vector found by the optimal lagrange multiplier,is an arbitrary support vector.
Compared with the prior art, the invention has the following advantages and effects:
1. the invention realizes the qualification inspection of screw rod parts based on a robot, trains a classification model capable of being effectively identified by extracting the singularity and statistical characteristics of the acquired torque signal, and performs subsequent communication, calculation and processing through an industrial personal computer, has simple device structure and easy system maintenance, realizes automatic acquisition and processing of data through the industrial personal computer, and can effectively improve the efficiency of data processing;
2. the screw rod automatic detection device is high in automation degree, the screw rod is subjected to screwing detection by using a robot, the qualification of screw rod parts is automatically detected, and the detection efficiency is greatly improved;
3. the problem of poor robustness in the traditional signal identification method is solved, and meanwhile, the rapidity of identification is kept.
Drawings
FIG. 1 is a schematic diagram of the overall structure of a screw rod type part qualification inspection robot system according to an embodiment of the invention;
FIG. 2 is a flow chart of a screw rod part qualification inspection method based on moment singularity analysis according to an embodiment of the present invention;
FIG. 3(a) is a torque signal diagram of a qualified screw-type part collected according to an embodiment of the present invention;
FIG. 3(b) is a torque signal diagram of an unqualified screw part according to the embodiment of the invention;
FIG. 4(a) is a diagram showing the results of inspecting and identifying qualified leading screw parts according to the embodiment of the present invention;
FIG. 4(b) is a graph showing the results of testing and identifying an unqualified lead screw part according to the embodiment of the present invention;
FIG. 4(c) is a diagram showing the results of detecting and identifying qualified lead screw parts under the damping interference according to the embodiment of the present invention;
FIG. 4(d) is a diagram showing the results of detecting and identifying defective lead screw parts under the damping interference according to the embodiment of the present invention;
in the figure: 1-a robot; 2-a terminal device; 3-simulating a detection experiment platform; 4-a screw rod; 5-a driver; 6-an industrial personal computer; 7-control cabinet.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
The screw rod part qualification inspection and analysis method based on the moment singularity analysis is implemented through a robot qualification inspection system. Referring to fig. 1, the robot qualification inspection system includes a robot 1, a terminal device 2, a simulation test experiment platform 3, a screw rod 4, a driver 5, an industrial personal computer 6, and a control cabinet 7. The end device 2 is fixed on the tail end of the robot 1, the screw rod 4 is fixed on the simulation detection experiment platform 3, the industrial personal computer 6 is connected with the control cabinet 7 through an Ethernet cable, the industrial personal computer 6 is connected with the driver 5 through the Ethernet cable, and the driver 5 is connected with the end device 2 through a signal line and a power line.
When the screw rod type parts are unqualified, the moment signals collected by the robot qualification testing system are mutation signals. Namely, when the mutation signal is detected, judging that the product is unqualified; and when the signal is finished and the mutation signal is not detected yet, judging that the product is qualified.
As shown in fig. 2, the method for inspecting and analyzing the eligibility of the screw rod type part based on the moment singularity analysis, provided by the invention, comprises the following steps:
s1, adjusting the pose of the robot 1 to enable the robot end device 2 to be aligned and matched with the screwing end of the screw rod 4 on the simulation detection experiment platform 3, enabling the screw rod to be smoothly screwed when the robot qualification testing system is started, and collecting a screw rod rotation torque signal by the end device 2 when the screw rod is screwed at the robot end position 2.
Specifically, in some embodiments of the present invention, the step S1 specifically includes:
s11, inputting the coordinates of the screwing end of the screw rod 4 on the simulation detection experiment platform 3 into the robot 1, and adjusting the position and the posture of the robot 1 to enable the terminal device 2 to be aligned and matched with the screwing end of the screw rod 4 on the simulation detection experiment platform 3;
and S12, starting, driving the screw rod 4 on the simulation detection experiment platform 3 to rotate by the end device 2 of the robot, and sending a moment signal acquired by the end device 2 of the robot to the industrial personal computer 6 in real time.
S2, preprocessing the torque signal acquired in the step 1, and denoising the preprocessed torque signal in real time, wherein in some embodiments of the invention, a wavelet threshold denoising method based on bayes is adopted for denoising.
Specifically, in some embodiments of the present invention, the step S2 specifically includes:
s21, carrying out sliding window iterative preprocessing on the collected torque signal, and sliding the window every 1msAdding new data of 1ms and removing the data added in the earliest 1ms, and keeping the real-time performance of the data;
s22, carrying out real-time noise reduction processing on the torque signal after the sliding window iteration by using a wavelet threshold noise reduction method based on bayes, and removing high-frequency noise to avoid excessive interference on an effective signal part.
Specifically, in some embodiments of the present invention, the step S22 specifically includes:
s221, in order to weaken the boundary effect generated in the noise reduction process, the sliding window is subjected toObtaining a processed sliding window W for symmetric continuation processingτ(i);
S222, for the Wτ(i) Performing waveletTransforming to obtain wavelet coefficientj,kWτ(i) Further, the variance of the noisy signal is obtained:
where j is the scale, k is time, i is a time instant, and N is the total data length.
S223, the collected moment signal includes an original signal and noise, and the variance of the noise is estimated as:
wherein, mean (& gtY & ltj,kWτ(i) |) is the median of the absolute values of the wavelet coefficients in the wavelet decomposition node. The variance σ of the original signals(j, k, i) is:
s224, calculating to obtain a Bayes threshold value:
and S3, calculating and extracting singularity characteristics and statistical characteristics of the noise-reduced moment signal in the S2, wherein the statistical characteristics comprise a maximum value, a minimum value, an average value, a skewness, a kurtosis and a standard deviation.
Specifically, in some embodiments of the present invention, the step S3 specifically includes:
s31, calculating a Lipschitz index of the noise-reduced torque signal in the step S2, wherein the index is a singular characteristic value of the torque signal;
and S32, extracting statistical characteristics of the moment signal and the Lipschitz index to form a characteristic vector TZ, wherein the statistical characteristics comprise a maximum value, a minimum value, an average value, a skewness, a kurtosis and a standard deviation.
And S4, performing feature compression processing on the features obtained by calculation in the S3 to obtain a final feature vector formed by the features with high correlation with the quality of the screw rod type parts.
Specifically, in some embodiments of the present invention, the step S4 specifically includes:
s41, the feature compression is to perform feature compression processing on the feature vector TZ calculated in step S3 by using a Fisher discriminant method, where the Fisher discriminant index is:
wherein n is the serial number of the characteristic element, p and q are the class numbers, SBnIs the dispersion between n features; SWnIs the dispersion within the feature, M is the number of classes of the feature,is of class LpThe mean value of the medium-sized features n,is of class LqThe mean value of the medium-sized features n,is of class LpThe variance of the middle-range feature n,andare calculated by the following formulas, respectively:
wherein L ispIs the p-th class of feature elements, x is the class LpCharacteristic of (1)pIs the number of feature element samples in the type. Sorting the calculated Fisher discrimination indexes according to the magnitude of the numerical values, and taking the characteristic elements with the top rank to form a vector
S42, sorting the calculated Fisher discrimination indexes, wherein the characteristic with high correlation is the characteristic with the top rank. In some embodiments of the invention, the top 5 features are taken as the features with high relevance.
In other embodiments, mutual information methods may also be used for feature compression.
S5, inputting the training sample composed of the feature vectors processed in the step 4 into a classification model of a support vector machine for training to obtain a classification model capable of detecting the product qualification, and using the classification model to realize the qualification inspection of the screw parts.
Specifically, step S5 specifically includes:
s51, forming the processed feature vectors in the step S4 into a sample set and a test set;
and S52, inputting the sample set and the test set into a Support Vector Machine (SVM) classification model for training to obtain a classification model capable of detecting the product qualification.
Specifically, step S52 specifically includes:
s521, the Support Vector Machine (SVM) model can classify normal moment signals and abrupt moment signals by defining an optimal hyperplane as a division plane of a moment signal characteristic sample, and the hyperplane equation is as follows:
where ω is the hyperplane normal vector and b is the displacement term of the hyperplane.
S522, each sample corresponds to a label yiE { -e, e }, where e, -e represent different sample classes, respectively. If yiWhen is + e, then there isIf yiWhen is-e, then there isLet the equation:
wherein the content of the first and second substances,is the ith feature after feature compression.
In one embodiment of the present invention, e has a value of 1. It will be appreciated that + e and-e are used only to distinguish between positive and negative samples, and in other embodiments other values may be used.
S523, a sample for establishing the equation in step S522 is referred to as a support vector, and the sum of the distances from the two heterogeneous support vectors to the hyperplane is referred to as a distance γ:
s524, solving the optimal hyperplane needs to satisfy the following equation:
where m is the number of samples.
S525, constructing an objective function through a Lalang Ri multiplier method, solving through Sequence Minimum Optimization (SMO), and finally determining the optimal hyperplane as follows:
wherein, w*Is a normal vector found by the optimal lagrange multiplier,is an arbitrary support vector and is,is the optimal lagrangian multiplier.
And finishing training after the optimal hyperplane is obtained, and obtaining a final support vector machine classification model.
And S53, repeating the steps S1, S2 and S3, wherein in the step S3, the effective features extracted in the step S4 are directly selected and input into the trained support vector machine classification model, when the classification result is-e, the screw rod parts are unqualified, and when the classification result is e, the screw rod parts are qualified, so that the qualification detection of the screw rod parts is completed. As shown in fig. 4(a) and (b), qualified lead screw parts and unqualified lead screw parts can be accurately identified even under the damping interference, as shown in fig. (c) and (d).
The detection method provided by the embodiment of the invention has better robustness while giving consideration to quick response, and does not need a large number of training samples and training time.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents and are included in the scope of the present invention.
Claims (10)
1. A screw rod part qualification inspection method based on moment singularity analysis is characterized by comprising the following steps:
s1, obtaining a screw rod rotation torque signal through the tail end of the robot;
s2, preprocessing the acquired torque signal in the S1, and then carrying out real-time noise reduction processing on the preprocessed torque signal;
s3, calculating and extracting the singularity characteristic and the statistical characteristic of the noise-reduced moment signal in the S2;
s4, performing feature compression processing on the features obtained through calculation in the S3 to obtain a final feature vector formed by the features with high correlation with the quality of the screw rod type parts;
and S5, inputting the training sample composed of the processed characteristic vectors in the step S4 into a support vector machine classification model for training to obtain the support vector machine classification model capable of detecting the product qualification, and using the model to realize the qualification inspection of the screw parts.
2. The screw rod type part qualification testing method based on the moment singularity analysis as claimed in claim 1, wherein the step S2 specifically comprises:
s21, carrying out sliding window iterative preprocessing on the collected torque signal, and sliding the window at each time interval delta tAdding data of a new time interval delta t and removing the data added at the earliest time interval delta t;
and S22, carrying out real-time denoising processing on the torque signal after the sliding window iteration by using a wavelet threshold denoising method based on bayes.
3. The screw rod type part qualification inspection method based on moment singularity analysis according to the full force requirement 2, wherein the step S22 specifically comprises:
s221, aiming at the sliding windowObtaining a processed sliding window W for symmetric continuation processingτ(i);
S222, the processed sliding window Wτ(i) Wavelet transformation is carried out to obtain wavelet coefficientsj,kWτ(i) Further, the variance of the noisy signal is obtained:
where j is the scale, k is time, i is a certain time, and N is the total data length;
s223, the collected moment signal is composed of an original signal and noise, and the variance of the noise is estimated as:
wherein, mean (& gtY & ltj,kWτ(i) |) is the median of the absolute values of the wavelet coefficients in the wavelet decomposition node, the variance σ of the original signals(j, k, i) is:
s224, calculating to obtain a Bayes threshold value:
4. the screw rod type part qualification testing method based on the moment singularity analysis as claimed in claim 1, wherein the step S3 specifically comprises:
s31, calculating a Lipschitz index of the noise-reduced torque signal in the step S2, wherein the index is a singular characteristic value of the torque signal;
and S32, extracting the moment signal and the statistical characteristics of the Lipschitz index to form a characteristic vector TZ.
5. The screw rod type part qualification testing method based on moment singularity analysis according to claim 1, wherein the statistical features comprise a maximum value, a minimum value, an average value, a skewness, a kurtosis and a standard deviation.
6. The screw rod type part qualification testing method based on moment singularity analysis according to claim 1, wherein in step S4, a mutual information method or a Fisher discrimination method is adopted for feature compression.
7. The screw rod type part qualification inspection method based on moment singularity analysis according to claim 1, wherein the step S4 specifically comprises:
s41, performing feature compression processing on the feature TZ obtained by calculation in the step S3 by using a Fisher discriminant method, wherein the Fisher discriminant index is as follows:
wherein n is the serial number of the characteristic element, SBnIs the dispersion between n features; SWnIs the dispersion within the feature, M is the number of classes of the feature,is of class LpThe mean value of the medium-sized features n,is of class LqMean of medium features n;
s42, sorting the calculated Fisher discrimination indexes, wherein the characteristic with high correlation is the characteristic with the top rank.
8. The screw rod type part qualification inspection method based on moment singularity analysis according to claim 7,andare calculated by the following formulas, respectively:
wherein L ispIs the p-th class of feature elements, x is the class LpCharacteristic of (1)pIs the number of feature element samples in the type.
9. The screw rod type part qualification testing method based on the moment singularity analysis according to any one of claims 1 to 8, wherein the step S5 specifically comprises:
s51, forming the processed feature vectors in the step S4 into a sample set and a test set;
and S52, inputting the sample set and the test set into a classification model of a support vector machine for training to obtain a classification model capable of detecting the product qualification.
10. The screw rod type part qualification testing method based on the moment singularity analysis of claim 9, wherein the step S52 specifically comprises:
s521, the support vector machine model can classify normal torque signals and abrupt torque signals by defining an optimal hyperplane as a dividing plane of a torque signal characteristic sample, and the hyperplane equation is as follows:
wherein, omega is a hyperplane normal vector, b is a hyperplane displacement term;
s522, each sample corresponds to a label yiE { -e, e }, wherein e, -e represent different sample classes, respectively, if yiWhen is + e, then there isIf yiWhen is-e, then there isLet the equation:
s523, a sample for establishing the equation in step S522 is referred to as a support vector, and the sum of the distances from the two heterogeneous support vectors to the hyperplane is referred to as a distance γ:
s524, solving the optimal hyperplane needs to satisfy the following equation:
s525, constructing an objective function through a Lalang Ri multiplier method, obtaining the objective function through a sequence minimum optimization solution, and finally determining the optimal hyperplane as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111554836.2A CN114358059A (en) | 2021-12-17 | 2021-12-17 | Screw rod part qualification inspection method based on moment singularity analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111554836.2A CN114358059A (en) | 2021-12-17 | 2021-12-17 | Screw rod part qualification inspection method based on moment singularity analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114358059A true CN114358059A (en) | 2022-04-15 |
Family
ID=81100223
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111554836.2A Pending CN114358059A (en) | 2021-12-17 | 2021-12-17 | Screw rod part qualification inspection method based on moment singularity analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114358059A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101833671A (en) * | 2010-03-30 | 2010-09-15 | 西安理工大学 | Support vector machine-based surface electromyogram signal multi-class pattern recognition method |
CN110472288A (en) * | 2019-07-17 | 2019-11-19 | 武汉科技大学 | A kind of prediction technique, system, device and the medium of Bolt Torque decaying |
CN112036435A (en) * | 2020-07-22 | 2020-12-04 | 温州大学 | Brushless direct current motor sensor fault detection method based on convolutional neural network |
CN112036296A (en) * | 2020-08-28 | 2020-12-04 | 合肥工业大学 | Motor bearing fault diagnosis method based on generalized S transformation and WOA-SVM |
-
2021
- 2021-12-17 CN CN202111554836.2A patent/CN114358059A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101833671A (en) * | 2010-03-30 | 2010-09-15 | 西安理工大学 | Support vector machine-based surface electromyogram signal multi-class pattern recognition method |
CN110472288A (en) * | 2019-07-17 | 2019-11-19 | 武汉科技大学 | A kind of prediction technique, system, device and the medium of Bolt Torque decaying |
CN112036435A (en) * | 2020-07-22 | 2020-12-04 | 温州大学 | Brushless direct current motor sensor fault detection method based on convolutional neural network |
CN112036296A (en) * | 2020-08-28 | 2020-12-04 | 合肥工业大学 | Motor bearing fault diagnosis method based on generalized S transformation and WOA-SVM |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Chen et al. | Graph convolutional network-based method for fault diagnosis using a hybrid of measurement and prior knowledge | |
CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
CN112284736B (en) | Convolutional neural network fault diagnosis method based on multi-channel attention module | |
CN113567130A (en) | Bearing fault diagnosis method based on multiple working conditions of equipment | |
CN109934422B (en) | Neural network wind speed prediction method based on time series data analysis | |
CN108435819B (en) | Energy consumption abnormity detection method for aluminum profile extruder | |
CN113076920B (en) | Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model | |
CN111275108A (en) | Method for performing sample expansion on partial discharge data based on generation countermeasure network | |
CN107886539A (en) | High class gear visible detection method under a kind of industrial scene | |
CN105447506A (en) | Gesture recognition method based on interval distribution probability characteristics | |
CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
CN113554621A (en) | Tool wear state identification system and method based on wavelet scale map and depth migration learning | |
CN117009916A (en) | Actuator fault diagnosis method based on multi-sensor information fusion and transfer learning | |
CN114754973A (en) | Wind tunnel force measurement test data intelligent diagnosis and analysis method based on machine learning | |
CN116735170A (en) | Intelligent fault diagnosis method based on self-attention multi-scale feature extraction | |
CN113506239A (en) | Strip steel surface defect detection method based on cross-stage local network | |
CN114118219A (en) | Data-driven real-time abnormal detection method for health state of long-term power-on equipment | |
CN112151067B (en) | Digital audio tampering passive detection method based on convolutional neural network | |
CN113392475A (en) | Intelligent fault identification method for speed reducer of industrial robot | |
CN114358059A (en) | Screw rod part qualification inspection method based on moment singularity analysis | |
CN113523904A (en) | Cutter wear detection method | |
CN116630728A (en) | Machining precision prediction method based on attention residual error twin network | |
CN108960347B (en) | System and method for evaluating effect of handwriting recognition sequencing stability of convolutional neural network | |
CN110610203A (en) | Electric energy quality disturbance classification method based on DWT and extreme learning machine | |
CN116452963A (en) | Multi-dimensional feature-based seabed substrate type classification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |