CN114248152A - Cutter wear state evaluation method based on optimization features and lion group optimization SVM - Google Patents
Cutter wear state evaluation method based on optimization features and lion group optimization SVM Download PDFInfo
- Publication number
- CN114248152A CN114248152A CN202111674361.0A CN202111674361A CN114248152A CN 114248152 A CN114248152 A CN 114248152A CN 202111674361 A CN202111674361 A CN 202111674361A CN 114248152 A CN114248152 A CN 114248152A
- Authority
- CN
- China
- Prior art keywords
- cutter
- svm
- lion
- wear state
- feature
- 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
- 241000282320 Panthera leo Species 0.000 title claims abstract description 50
- 238000005457 optimization Methods 0.000 title claims abstract description 33
- 238000011156 evaluation Methods 0.000 title claims abstract description 17
- 238000001914 filtration Methods 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 21
- 238000013210 evaluation model Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 230000002159 abnormal effect Effects 0.000 claims description 25
- 238000005520 cutting process Methods 0.000 claims description 23
- 230000009467 reduction Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 13
- 238000000354 decomposition reaction Methods 0.000 claims description 8
- 239000013598 vector Substances 0.000 claims description 8
- 230000005856 abnormality Effects 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 7
- 238000000513 principal component analysis Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 4
- 238000004519 manufacturing process Methods 0.000 abstract description 11
- 238000005299 abrasion Methods 0.000 abstract description 5
- 238000003801 milling Methods 0.000 description 10
- 238000003754 machining Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000003556 assay Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 241000287127 Passeridae Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 235000021178 picnic Nutrition 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000003825 pressing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23Q—DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING; MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
- B23Q17/00—Arrangements for observing, indicating or measuring on machine tools
- B23Q17/09—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool
- B23Q17/0952—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool during machining
- B23Q17/0957—Detection of tool breakage
Abstract
The invention relates to a cutter wear state evaluation method based on an optimized feature and a lion group optimization SVM, which comprises the following steps of: s1, preprocessing data; s2, feature extraction; s3, filtering the features to obtain a preferred feature subset with highest correlation and minimum redundancy; s4, constructing a cutter wear evaluation model; and S5, evaluating the wear state of the cutter. The feature subset obtained by the method has higher degree of correlation and minimum redundancy, the SVM classifier optimized by the lion group algorithm has low time complexity, stronger optimization ability, faster iteration, lower classification error rate and 97.125% of recognition rate, and the method can improve the evaluation speed and stability of the abrasion state of the cutter, improve the production efficiency and reduce the production cost.
Description
Technical Field
The invention relates to a technology in the technical field of machining, in particular to a cutter wear state evaluation method based on an optimized feature and a lion group optimized SVM.
Background
Modern manufacturing industry is gradually developing towards intellectualization, and the performance state perception of main parts of the numerical control machine tool is particularly important in the processing process. The evaluation of the state of wear of the tool is therefore very important during machining, studies have shown that the downtime of a numerically controlled machine tool caused by the failure of the tool represents about 20% of its total downtime. The interruption of the processing process may cause the scrapping of the workpiece and even the breakdown of the whole production system, which affects the production efficiency.
The traditional tool replacement is mostly based on the empirical life analysis of the tool, and the used tool is replaced when the used tool reaches the empirical life value. However, changing the tool according to empirical life entails two problems: premature replacement of the tool results in increased tool use costs; failure to replace the dull-ground tool in time results in a reduction in workpiece quality. In addition, the conventional fixed cutting method based on experience selection cannot ensure that the tool is fully utilized within a limited service life, which easily causes waste of the tool and increase of production cost. The cutting parameters are dynamically optimized according to the tool wear information, and the tool is replaced in time, so that the important effects of fully using the tool, reducing the production cost and preventing machining accidents are achieved. The existing tool wear state monitoring methods are generally divided into direct methods and indirect methods. In contrast, direct assays are more costly and most scholars are working on indirect assays.
At present, the indirect method for evaluating the abrasion state of the cutter at home and abroad mainly comprises the following steps: salgado et al analyzed cutting force and sound signals during workpiece machining to determine the current wear state of the tool in 2018 by a method of combining LS-SVM and Singular Spectrum Analysis (SSA). In 2019 Chen Jiahua and the like, an Intrinsic Mode Function (IMF) energy distance method is combined with an SVM, and the current abrasion condition of a cutter of the shearing machine is judged according to working noise during field processing. In 2019, the evaluation model of the original SVM is optimized by using a PSO algorithm, such as Garcia-Nieto and the like, so that the accuracy rate of the evaluation on the wear state of the cutter reaches 95%. In 2020, the plum-shun et al uses GA to optimize the SVM model thereof and then judges the current cutter wear state, thereby obtaining good effect. In 2021, picnic will optimize SVM by using sparrow optimization algorithm, so that the accuracy of evaluating the wear state of the cutter reaches 96%. However, the most concerned points of these evaluation methods are the optimization of various types of classifiers by various optimization algorithms, and the research on the feature value extraction and selection is single, but the signal feature extraction and optimization selection are the key to realize the rapid and effective classification of the tool wear state. Therefore, in addition to focusing on the optimization of the back-end classifier, the feature optimization selection of the front-end is also worth attention.
Disclosure of Invention
Aiming at the defects of the prior art, a cutter wear state evaluation method based on optimized features and a lion group optimized SVM is provided, the front end uses multilayer Filter type filtering to select features to obtain a group of optimized feature subsets with highest correlation and lowest redundancy, and the rear end uses a lion group optimization algorithm to optimize the SVM to construct a cutter wear evaluation model.
The technical scheme for realizing the invention is as follows:
a cutter wear state evaluation method based on a preferred feature and a lion group optimization SVM comprises the following steps:
s1, preprocessing data: preprocessing data of an empirical data set of the numerical control machine tool, wherein the empirical data set of the numerical control machine tool comprises sensor data and tool wear condition data of 7 channels, and the sensor data of the 7 channels are cutting force in the x direction, the y direction and the z direction and vibration and acoustic emission signals in the x direction, the y direction and the z direction respectively;
s1.1, quantile abnormality detection: respectively carrying out quantile abnormality detection on the empirical data of the numerical control machine tool, and removing abnormal values;
s1.2, Hampel filtering: performing Hampel filtering on the data after the quantile abnormality detection of the step S1.1 is completed and abnormal values are removed;
s2, extracting time domain characteristics, frequency domain characteristics and time-frequency domain characteristics from the data subjected to the step S1;
s3, carrying out multi-layer Filter type filtering on the features extracted in the step S2 to obtain a preferred feature subset with highest correlation and minimum redundancy;
s4, optimizing the SVM by using a lion group optimization algorithm to the preferred feature subset obtained in the step S3 to construct a cutter wear evaluation model;
and S5, collecting cutting force in the x direction, the y direction and the z direction of the real-time numerical control machine tool, vibration in the x direction, the y direction and the z direction and acoustic emission signal data, and carrying out tool wear state evaluation by using the tool wear evaluation model constructed in the step S4.
Further: the method aims to achieve the purposes of higher optimization speed of an evaluation model, stronger convergence, faster iteration, lower classification error rate, effective reduction of production cost and improvement of production efficiency. The concrete steps of optimizing the SVM by the lion group optimization algorithm in the step S4 to construct the cutter wear evaluation model are as follows:
s4.1, taking the optimal characteristic subset obtained in the step S3 as a training set, taking the cutter wear condition as a basis for label division, and dividing the cutter wear condition into a slight wear state, a normal wear state and a rapid wear state;
s4.2, initializing each parameter of the lion group optimization algorithm;
s4.3, updating the positions of the lion king, the female lion and the young lion;
s4.4, calculating a fitness value, updating a global optimal position and a historical optimal position, and training by using a training set to obtain two parameters c and g of the SVM;
and S4.5, judging whether the optimal condition is achieved or not, if not, continuing iteration, and if so, outputting the trained LSO-SVM classifier to obtain a final tool wear state evaluation model.
Further: the specific steps of the multilayer Filter type filtering selection feature in the step S3 are as follows:
s3.1, Pearson correlation coefficient filtering: calculating Pearson correlation coefficient r values among the features, wherein the closer the absolute value of the r values is to 1, the stronger the correlation is, and selecting a strong correlation feature subset with the correlation between 0.8 and r < 1;
s3.2, PCA dimensionality reduction filtering: PCA principal component analysis is to rotate the coordinates formed by the original feature vectors, select the vector with the maximum variance to form a new coordinate system, perform dimensionality reduction mapping on the original feature set to a low-dimensional space to realize dimensionality reduction, and select the first m feature values with the principal component contribution rate of more than 85% to form a new feature subset;
s3.3, filtering the mRMR search criterion: the mRMR criterion takes mutual information as a basic standard for measuring the correlation degree between variables, and performs feature search through a maximum correlation criterion and a minimum redundancy criterion to form a final preferred feature subset.
Pearson correlation coefficients can effectively filter out most invalid features, but are rough; PCA can obtain projections of a data low-dimensional space, but the projections do not contain class label information; the mRMR search criterion considers not only the correlation between the features and the label, but also the correlation between the features, but is not suitable for large-scale data and has high time complexity; the three are combined to form multi-layer Filter type filtration, so that the defects of redundancy and the problem of overhigh complexity can be overcome, and the optimal feature subset with highest correlation and minimum redundancy is obtained finally.
Further: the specific steps of feature extraction in the step S2 are as follows:
s2.1, extracting time domain features: extracting time domain characteristics on a three-axis cutting force and a three-axis vibration signal of the cutter; including peak, maximum, minimum, peak-to-peak, mean, average amplitude, square root amplitude, variance, standard deviation, root mean square, kurtosis, skewness, form factor, crest factor, pulse factor, margin factor, and clearance factor;
s2.2, extracting frequency domain features: extracting frequency domain characteristics including average frequency, center-of-gravity frequency, frequency root mean square and frequency standard deviation from three-axis cutting force and three-axis vibration signals of the cutter;
s2.3, decomposing and extracting the time-frequency domain characteristics by the wavelet packet: and extracting wavelet packet energy characteristics by utilizing three-layer decomposition of a wavelet packet on the triaxial cutting force and triaxial vibration signals of the cutter.
Besides common time domain and frequency domain characteristics, the time domain and frequency domain characteristics are extracted by wavelet packet transformation, the wavelet packet transformation can be used for not only continuously decomposing the signals at low frequency, but also carrying out high-precision analysis on the high-frequency region of the signals, and conditions are provided for energy characteristic extraction on the full frequency band. By using wavelet packet decomposition, the time-frequency domain information of the original signal can be decomposed into a plurality of frequency bands, and when a cutter fault occurs, the vibration signal and the cutting force signal can be directly changed in frequency distribution, so that the energy value of each frequency band after the wavelet packet decomposition is changed. Based on the above, the current tool wear state can be judged.
Further: the method has the advantages of simple and easily accessible data preprocessing and low time complexity. The data preprocessing in the step S1 includes the following steps:
s1.1, detecting a quantile abnormal value: all the numerical values are firstly arranged from small to large, then are equally divided into 4 parts, and then an upper quartile Q1, a middle quartile Q2 and a lower quartile Q3 are sequentially arranged, so that the quartile distance IQR is equal to Q3-Q1, and the judgment basis of the abnormal value is the numerical value which is larger than Q1+1.5 XIRQ or smaller than Q3-1.5 XIQR;
s1.2, Hampel filtering: setting the number k of samples on two sides of the sample, setting the window size to be 2k +1, setting an upper boundary coefficient n _ delta and a lower boundary coefficient n _ delta, calculating the local standard deviation x _ delta and the local estimated median x _ m of each sample based on a sliding window, then calculating the abnormal value upper boundary (x _ m + n _ delta x _ delta) and the abnormal value lower boundary (x _ m-n _ delta x _ delta) of the sample, and replacing the sample by using the estimated median x _ m if the sample value is larger than the abnormal value upper boundary or smaller than the abnormal value lower boundary.
The invention has the beneficial effects that:
the invention not only focuses on the optimization of the rear-end classifier, but also carries out multi-layer Filter type filtering on the feature selection of the front end; compared with the prior art, the obtained feature subset has higher degree of correlation and minimum redundancy, the SVM classifier optimized by the lion group algorithm has low time complexity, stronger optimization ability, faster iteration, lower classification error rate and 97.125% of recognition rate, so that the evaluation speed and stability of the wear state of the cutter can be improved, the production efficiency is improved, and the production cost is reduced.
Drawings
FIG. 1 is a flow chart of the tool wear state assessment method based on the preferred feature and the lion group optimization SVM of the invention;
FIG. 2 is a visualization of data preprocessing;
FIG. 3 shows the results of principal component analysis;
FIG. 4 is a diagram of the classification results of a test set;
FIG. 5 is a comparison graph of classification error rates of PSO-SVM and LSO-SVM.
Detailed Description
The following detailed description of specific embodiments of the invention is made with reference to the accompanying drawings in which:
a cutter wear state evaluation method based on a preferred feature and a lion group optimization SVM comprises the following steps:
s1, preprocessing data: preprocessing data of an empirical data set of the numerical control machine tool, wherein the empirical data set of the numerical control machine tool comprises sensor data and tool wear condition data of 7 channels, and the sensor data of the 7 channels are cutting force in the x direction, the y direction and the z direction and vibration and acoustic emission signals in the x direction, the y direction and the z direction respectively;
s1.1, detecting a quantile abnormal value: all the numerical values are firstly arranged from small to large, then are equally divided into 4 parts, and then an upper quartile Q1, a middle quartile Q2 and a lower quartile Q3 are sequentially arranged, so that the quartile distance IQR is equal to Q3-Q1, and the judgment basis of the abnormal value is the numerical value which is larger than Q1+1.5 XIRQ or smaller than Q3-1.5 XIQR;
s1.2, Hampel filtering: setting the number k of samples on two sides of the sample, setting the window size to be 2k +1, setting an upper boundary coefficient n _ delta and a lower boundary coefficient n _ delta, calculating the local standard deviation x _ delta and the local estimated median x _ m of each sample based on a sliding window, then calculating the abnormal value upper boundary (x _ m + n _ delta x _ delta) and the abnormal value lower boundary (x _ m-n _ delta x _ delta) of the sample, and replacing the sample by using the estimated median x _ m if the sample value is larger than the abnormal value upper boundary or smaller than the abnormal value lower boundary.
S2, extracting time domain features, frequency domain features and time-frequency domain features from the data after the step S1:
s2.1, extracting time domain features: extracting time domain characteristics on a three-axis cutting force and a three-axis vibration signal of the cutter; including peak, maximum, minimum, peak-to-peak, mean, average amplitude, square root amplitude, variance, standard deviation, root mean square, kurtosis, skewness, form factor, crest factor, pulse factor, margin factor, and clearance factor;
s2.2, extracting frequency domain features: extracting frequency domain characteristics including average frequency, center-of-gravity frequency, frequency root mean square and frequency standard deviation from three-axis cutting force and three-axis vibration signals of the cutter;
s2.3, decomposing and extracting the time-frequency domain characteristics by the wavelet packet: and extracting wavelet packet energy characteristics and wavelet packet energy spectrum entropy characteristics by utilizing three-layer decomposition of wavelet packets on three-axis cutting force and three-axis vibration signals of the cutter.
S3, performing multi-layer Filter type filtering on the features extracted in the step S2 to obtain a preferred feature subset with highest correlation and minimum redundancy:
s3.1, Pearson correlation coefficient filtering: calculating the value of a Pearson correlation coefficient r among the features, wherein the closer the absolute value of the r value is to 1, the stronger the correlation is, and selecting a strong correlation feature subset with the r being more than 0.8 and less than 1;
s3.2, PCA dimensionality reduction filtering: PCA principal component analysis is to rotate the coordinates formed by the original feature vectors, select the vector with the maximum variance to form a new coordinate system, perform dimensionality reduction mapping on the original feature set to a low-dimensional space to realize dimensionality reduction, and select the first m feature values with the contribution rate of the principal components larger than 85% to form a new feature subset;
s3.3, filtering the mRMR search criterion: the mRMR criterion takes mutual information as a basic standard for measuring the correlation degree between variables, and performs feature search through a maximum correlation criterion and a minimum redundancy criterion to form a final preferred feature subset.
S4, optimizing the SVM by using a lion group optimization algorithm to the preferred feature subset obtained in the step S3 to construct a cutter wear evaluation model;
s4.1, taking the optimal characteristic subset obtained in the step S3 as a training set, taking the cutter wear condition as a basis for label division, and dividing the cutter wear condition into a slight wear state, a normal wear state and a rapid wear state;
s4.2, initializing each parameter of the lion group optimization algorithm;
s4.3, updating the positions of the lion king, the female lion and the young lion;
s4.4, calculating a fitness value, updating a global optimal position and a historical optimal position, and training by using a training set to obtain two parameters c and g of the SVM;
and S4.5, judging whether the optimal condition is achieved or not, if not, continuing iteration, and if so, outputting the trained LSO-SVM classifier to obtain a final tool wear state evaluation model.
And S5, collecting cutting force in the x direction, the y direction and the z direction of the real-time numerical control machine tool, vibration in the x direction, the y direction and the z direction and acoustic emission signal data, and carrying out tool wear state evaluation by using the tool wear evaluation model constructed in the step S4.
The method is verified by using a data set published by prediction competition of the health state of a cutter of a high-speed milling machine tool in the United states PHM Society (2010), a numerical control milling machine in an experiment adopts an end milling mode, sensor data of 7 channels are collected, wherein the sensor data are milling force in x, y and z directions and vibration and acoustic emission signals in x, y and z directions respectively, the milling cutter has 315 continuous end face milling data records, the abrasion loss of the cutting edge of the milling cutter after each milling is also stored as a basis for training a prediction model, and C6 is one milling cutter data set which runs independently for 315 continuous times and contains an abrasion loss label;
quantile abnormality detection: the method is based on statistics, all numerical values are arranged from small to large, then the numerical values are equally divided into 4 parts, an upper quartile Q1, a middle quartile Q2 and a lower quartile Q3 are sequentially arranged, the quartile distance IQR is equal to Q3-Q1, and the judgment basis of the abnormal value is the numerical value which is larger than Q1+1.5 multiplied by IRQ or smaller than Q3-1.5 multiplied by IQR;
hampel filtering: setting the number k of samples on both sides of the sample, setting the window size to be 2k +1, and setting an upper and lower bound coefficient nδCalculating the local standard deviation x of each sample based on the sliding windowδLocally estimating median xmThen, the upper bound of the abnormal value of the sample is calculated as xm+nδ×xδAnd the lower bound of the outlier xm-nδ×xδIf the sample value is larger than the upper limit of the abnormal value or smaller than the lower limit of the abnormal value, the estimated median value x is usedmReplacing the sample;
the feature extraction comprises the following steps:
1) extracting time domain features and frequency domain features: extracting time domain characteristics such as average amplitude, standard deviation, peak value, root mean square and the like, and frequency domain characteristics such as average frequency, frequency root mean square and the like, from the three-axis cutting force and the three-axis vibration signal of the cutter;
2) wavelet packet decomposition extraction time-frequency domain characteristics: extracting wavelet packet energy characteristics by utilizing three-layer decomposition of a wavelet packet on a triaxial cutting force and a triaxial vibration signal of the cutter, and for k-layer wavelet packet decomposition, node energy numbered m is as follows:
the step of selecting features for multi-layer Filter-like filtering includes:
1) pearson correlation coefficient filtering: and calculating the value of a Pearson correlation coefficient r between the characteristics, wherein the calculation formula of the value of r is as follows:
the more the absolute value of the r value is close to 1, the stronger the correlation is, and a strong correlation characteristic subset with the r being more than 0.8 and less than 1 is selected;
2) PCA dimensionality reduction filtration: PCA principal component analysis is to rotate the coordinates formed by the original feature vectors, select the vector with the maximum variance to form a new coordinate system, perform dimensionality reduction mapping on the original feature set to a low-dimensional space to realize dimensionality reduction, and select the first m feature values with the contribution rate of the principal components larger than 85% to form a new feature subset;
3) mRMR search criteria filtering: the mRMR criterion takes mutual information as a basic standard for measuring the correlation between variables, wherein the theory of mutual information is defined as follows:
given two random variables x and y, their probability densities and joint probability densities, p (x), p (y), p (x, y), the formula for mutual information I between the two variables is defined as:
and performing feature search through a maximum correlation and minimum redundancy criterion to form a final preferred feature subset, wherein the maximum correlation criterion is defined as:
the minimum redundancy criterion is defined as:
the mRMR algorithm combines the maximum correlation and minimum redundancy criteria to obtain the following two search metrics, namely, a mutual information difference Metric (MID) and a mutual information quotient Metric (MIQ), where MID is defined as:
MIQ is defined as:
wherein S is a feature subset, | S | is the number of feature quantities, fiIs the ith feature, and c is the target class;
the method for optimizing the SVM by the lion group optimization algorithm to construct the cutter wear evaluation model comprises the following steps:
1) taking the obtained optimal characteristic subset as a training set, and taking the wear band width VB of the cutter back tool face as a basis for label division, wherein 0-80 microns is in a slight wear state, 80-120 microns is in a normal wear state, and >120 microns is in a rapid wear state;
2) initializing the position x of a lion in a lion groupiThe number N, the maximum iteration number T, the dimension space D and the scale factor beta of the adult lion to the lion group. Calculating the number of the lion king and the parent lion in the lion group, and setting the historical optimal position of the individual with the rest being the young lionInitializing a group optimal position for the current position of each lion and setting the group optimal position as a lion king position;
3) updating the position of the lion king according to the following formula, calculating the fitness value,
the mother lion is followed by pressing
Adjust the position of the user, press the young lion
Adjust its position. Where γ is a random number generated according to a normal distribution N (0,1), αfIs a disturbance factor of the motion range of the female lion, alphacIs a disturbance factor of a young lion,
is the historical best position of the ith lion at the kth generation,
is the historical best position of a hunting cooperative partner randomly selected from the k-th generation female lion group, gkIs the optimal position in the kth generation population, g'kIs the position where the ith young lion is driven within the hunting range;
4) calculating a fitness value, updating a global optimal position and a historical optimal position, and training by using a training set to acquire two parameters c and g of the SVM;
5) judging whether the optimal condition is reached, if not, continuing iteration, and if so, outputting a trained LSO-SVM classifier to obtain a final tool wear state evaluation model;
6) and (3) randomly selecting a part of data sets in the data sets published by the prediction competition of the health state of the tool of the high-speed milling machine tool in the United states PHM Society (2010) as test sets, and verifying the identification rate and the performance of the model.
Referring to fig. 1, the method of the present invention is implemented by the following steps:
step one, preprocessing data.
Taking the Y-axis cutting force signal of sample No. 1 of the C6 dataset as an example, the visualization result shown in fig. 2 is obtained after a series of data preprocessing.
And step two, feature extraction.
TABLE 1 all extracted features
And step three, multilayer Filter type filtration.
Firstly, Z normalization processing is carried out on data, then a Pearson coefficient among all features is calculated by utilizing SPSS, a high-correlation feature subset with the r value of 0.8-1 is selected, and first-layer filtering is completed; performing PCA dimension reduction, wherein the result of the principal component is shown in figure 3, and finishing the second-layer filtration; and finally, selecting the final preferable feature subset by utilizing a search rule of an mRMR algorithm to complete the third-layer filtering.
And fourthly, optimizing the SVM by using the lion group optimization algorithm to construct a cutter wear evaluation model.
Initializing each parameter of LSO, then obtaining parameters c and g of SVM by optimizing, then judging whether the optimal parameter is reached, if not, continuing iteration until the optimal parameter is found, if so, outputting a trained LSO-SVM classifier to obtain a final cutter wear state evaluation model, then verifying the recognition rate and the performance of the model by using a test set, wherein the classification recognition rate is 97.125% as shown in FIG. 4, and comparing the classification error rates of the PSO-SVM and the LSO-SVM, as shown in FIG. 5, the upper classification error rate of the PSO-SVM is shown, and the lower classification error rate of the LSO-SVM is shown, so that the LSO-SVM classification error rate is lower, and the iteration speed is higher.
Claims (5)
1. A cutter wear state evaluation method based on a preferred feature and a lion group optimization SVM is characterized by comprising the following steps:
s1, preprocessing data: preprocessing data of an empirical data set of the numerical control machine tool, wherein the empirical data set of the numerical control machine tool comprises sensor data and tool wear condition data of 7 channels, and the sensor data of the 7 channels are cutting force in the x direction, the y direction and the z direction and vibration and acoustic emission signals in the x direction, the y direction and the z direction respectively;
s1.1, quantile abnormality detection: respectively carrying out quantile abnormality detection on the empirical data of the numerical control machine tool, and removing abnormal values;
s1.2, Hampel filtering: performing Hampel filtering on the data after the quantile abnormality detection of the step S1.1 is completed and abnormal values are removed;
s2, extracting time domain characteristics, frequency domain characteristics and time-frequency domain characteristics from the data subjected to the step S1;
s3, carrying out multi-layer Filter type filtering on the features extracted in the step S2 to obtain a preferred feature subset with highest correlation and minimum redundancy;
s4, optimizing the SVM by using a lion group optimization algorithm to the preferred feature subset obtained in the step S3 to construct a cutter wear evaluation model;
and S5, collecting cutting force in the x direction, the y direction and the z direction of the real-time numerical control machine tool, vibration in the x direction, the y direction and the z direction and acoustic emission signal data, and carrying out tool wear state evaluation by using the tool wear evaluation model constructed in the step S4.
2. The cutter wear state evaluation method based on the optimized features and the lion group optimized SVM as claimed in claim 1, wherein the lion group optimization algorithm in the step S4 optimizes the SVM to construct the cutter wear evaluation model by the specific steps of:
s4.1, taking the optimal characteristic subset obtained in the step S3 as a training set, taking the cutter wear condition as a basis for label division, and dividing the cutter wear condition into a slight wear state, a normal wear state and a rapid wear state;
s4.2, initializing each parameter of the lion group optimization algorithm;
s4.3, updating the positions of the lion king, the female lion and the young lion;
s4.4, calculating a fitness value, updating a global optimal position and a historical optimal position, and training by using a training set to obtain two parameters c and g of the SVM;
and S4.5, judging whether the optimal condition is achieved or not, if not, continuing iteration, and if so, outputting the trained LSO-SVM classifier to obtain a final tool wear state evaluation model.
3. The cutter wear state assessment method based on the preferred feature and the lion-group optimization SVM as claimed in claim 1, wherein the specific steps of selecting the features by multi-layer Filter in the step S3 are as follows:
s3.1, Pearson correlation coefficient filtering: calculating Pearson correlation coefficient r values among the features, wherein the closer the absolute value of the r values is to 1, the stronger the correlation is, and selecting a strong correlation feature subset with the correlation between 0.8 and r < 1;
s3.2, PCA dimensionality reduction filtering: PCA principal component analysis is to rotate the coordinates formed by the original feature vectors, select the vector with the maximum variance to form a new coordinate system, perform dimensionality reduction mapping on the original feature set to a low-dimensional space to realize dimensionality reduction, and select the first m feature values with the principal component contribution rate of more than 85% to form a new feature subset;
s3.3, filtering the mRMR search criterion: the mRMR criterion takes mutual information as a basic standard for measuring the correlation degree between variables, and performs feature search through a maximum correlation criterion and a minimum redundancy criterion to form a final preferred feature subset.
4. The cutter wear state assessment method based on the preferred feature and the lion-group optimization SVM as claimed in claim 1, wherein the specific steps of feature extraction in the step S2 are as follows:
s2.1, extracting time domain features: extracting time domain characteristics on a three-axis cutting force and a three-axis vibration signal of the cutter; including peak, maximum, minimum, peak-to-peak, mean, average amplitude, square root amplitude, variance, standard deviation, root mean square, kurtosis, skewness, form factor, crest factor, pulse factor, margin factor, and clearance factor;
s2.2, extracting frequency domain features: extracting frequency domain characteristics including average frequency, center-of-gravity frequency, frequency root mean square and frequency standard deviation from three-axis cutting force and three-axis vibration signals of the cutter;
s2.3, decomposing and extracting the time-frequency domain characteristics by the wavelet packet: and extracting wavelet packet energy characteristics by utilizing three-layer decomposition of a wavelet packet on the triaxial cutting force and triaxial vibration signals of the cutter.
5. The cutter wear state assessment method based on the preferred feature and lion group optimization SVM as claimed in claim 1, wherein the data preprocessing in the step S1 comprises the following specific steps:
s1.1, detecting a quantile abnormal value: all the numerical values are firstly arranged from small to large, then are equally divided into 4 parts, and then an upper quartile Q1, a middle quartile Q2 and a lower quartile Q3 are sequentially arranged, so that the quartile distance IQR is equal to Q3-Q1, and the judgment basis of the abnormal value is the numerical value which is larger than Q1+1.5 XIRQ or smaller than Q3-1.5 XIQR;
s1.2, Hampel filtering: setting the number k of samples on both sides of the sample, setting the window size to be 2k +1, and setting an upper and lower bound coefficient nδCalculating the local standard deviation x of each sample based on the sliding windowδLocally estimating median xmThen, the upper bound of the abnormal value of the sample is calculated as xm+nδ×xδAnd the lower bound of the outlier xm-nδ×xδIf the sample value is larger than the upper limit of the abnormal value or smaller than the lower limit of the abnormal value, the estimated median value x is usedmThe sample is replaced.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111674361.0A CN114248152A (en) | 2021-12-31 | 2021-12-31 | Cutter wear state evaluation method based on optimization features and lion group optimization SVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111674361.0A CN114248152A (en) | 2021-12-31 | 2021-12-31 | Cutter wear state evaluation method based on optimization features and lion group optimization SVM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114248152A true CN114248152A (en) | 2022-03-29 |
Family
ID=80799064
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111674361.0A Pending CN114248152A (en) | 2021-12-31 | 2021-12-31 | Cutter wear state evaluation method based on optimization features and lion group optimization SVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114248152A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117631599A (en) * | 2024-01-26 | 2024-03-01 | 深圳一嘉智联科技有限公司 | Industrial control computer data transmission method and system based on data analysis |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020017139A1 (en) * | 2000-05-15 | 2002-02-14 | Kluft Werner Wilhelm | Method and device for monitoring the wear condition of a tool |
| US20170320182A1 (en) * | 2016-05-06 | 2017-11-09 | Massachusetts Institute Of Technology | Method and Apparatus for Efficient Use of CNC Machine Shaping Tool Including Cessation of Use No Later than the Onset of Tool Deterioration By Monitoring Audible Sound During Shaping |
| CN108581635A (en) * | 2018-05-03 | 2018-09-28 | 哈尔滨理工大学 | A kind of milling cutter side edge cutting edge abrasion three-dimensional detection device and method |
| CN110509109A (en) * | 2019-07-16 | 2019-11-29 | 西安交通大学 | Tool Wear Monitoring method based on multiple dimensioned depth convolution loop neural network |
| CN110561193A (en) * | 2019-09-18 | 2019-12-13 | 杭州友机技术有限公司 | Cutter wear assessment and monitoring method and system based on feature fusion |
| JP2020042347A (en) * | 2018-09-06 | 2020-03-19 | 東芝機械株式会社 | Machine tool and tool abnormality determination method |
| CN111300146A (en) * | 2019-11-29 | 2020-06-19 | 上海交通大学 | Numerical control machine tool cutter abrasion loss online prediction method based on spindle current and vibration signal |
| CN112372371A (en) * | 2020-10-10 | 2021-02-19 | 上海交通大学 | Method for evaluating abrasion state of numerical control machine tool cutter |
| CN112433507A (en) * | 2019-08-26 | 2021-03-02 | 电子科技大学 | LSO-LSSVM (least squares support vector machine) -based five-axis numerical control machine tool thermal error comprehensive modeling method |
| CN112757053A (en) * | 2020-12-25 | 2021-05-07 | 清华大学 | Model fusion tool wear monitoring method and system based on power and vibration signals |
| CN113618491A (en) * | 2021-08-23 | 2021-11-09 | 浙江工业大学 | Method for establishing broach wear state recognition model |
-
2021
- 2021-12-31 CN CN202111674361.0A patent/CN114248152A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020017139A1 (en) * | 2000-05-15 | 2002-02-14 | Kluft Werner Wilhelm | Method and device for monitoring the wear condition of a tool |
| US20170320182A1 (en) * | 2016-05-06 | 2017-11-09 | Massachusetts Institute Of Technology | Method and Apparatus for Efficient Use of CNC Machine Shaping Tool Including Cessation of Use No Later than the Onset of Tool Deterioration By Monitoring Audible Sound During Shaping |
| CN108581635A (en) * | 2018-05-03 | 2018-09-28 | 哈尔滨理工大学 | A kind of milling cutter side edge cutting edge abrasion three-dimensional detection device and method |
| JP2020042347A (en) * | 2018-09-06 | 2020-03-19 | 東芝機械株式会社 | Machine tool and tool abnormality determination method |
| CN110509109A (en) * | 2019-07-16 | 2019-11-29 | 西安交通大学 | Tool Wear Monitoring method based on multiple dimensioned depth convolution loop neural network |
| CN112433507A (en) * | 2019-08-26 | 2021-03-02 | 电子科技大学 | LSO-LSSVM (least squares support vector machine) -based five-axis numerical control machine tool thermal error comprehensive modeling method |
| CN110561193A (en) * | 2019-09-18 | 2019-12-13 | 杭州友机技术有限公司 | Cutter wear assessment and monitoring method and system based on feature fusion |
| CN111300146A (en) * | 2019-11-29 | 2020-06-19 | 上海交通大学 | Numerical control machine tool cutter abrasion loss online prediction method based on spindle current and vibration signal |
| CN112372371A (en) * | 2020-10-10 | 2021-02-19 | 上海交通大学 | Method for evaluating abrasion state of numerical control machine tool cutter |
| CN112757053A (en) * | 2020-12-25 | 2021-05-07 | 清华大学 | Model fusion tool wear monitoring method and system based on power and vibration signals |
| CN113618491A (en) * | 2021-08-23 | 2021-11-09 | 浙江工业大学 | Method for establishing broach wear state recognition model |
Non-Patent Citations (2)
| Title |
|---|
| 李亚;黄亦翔;赵路杰;刘成良;: "基于t分布邻域嵌入与XGBoost的刀具多工况磨损评估", 机械工程学报, vol. 56, no. 01 * |
| 黄智;刘永超;邓涛;周涛;祝云;: "一种五轴数控机床热误差建模方法", 中国机械工程, no. 13 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117631599A (en) * | 2024-01-26 | 2024-03-01 | 深圳一嘉智联科技有限公司 | Industrial control computer data transmission method and system based on data analysis |
| CN117631599B (en) * | 2024-01-26 | 2024-04-12 | 深圳一嘉智联科技有限公司 | Industrial control computer data transmission method and system based on data analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113664612A (en) | Numerical control machine tool milling cutter abrasion real-time monitoring method based on deep convolutional neural network | |
| CN107877262B (en) | A kind of numerical control machine tool wear monitoring method based on deep learning | |
| CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
| CN109858345B (en) | Intelligent fault diagnosis method suitable for pipe expansion equipment | |
| CN116237817B (en) | Five-axis linkage numerical control machine intelligent monitoring system based on Internet of things | |
| CN114248152A (en) | Cutter wear state evaluation method based on optimization features and lion group optimization SVM | |
| CN111126255A (en) | Numerical control machine tool cutter wear value prediction method based on deep learning regression algorithm | |
| CN114619292A (en) | Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network | |
| CN116070527B (en) | Milling cutter residual life prediction method based on degradation model | |
| CN114273977A (en) | MES-based cutter wear detection method and system | |
| CN113752089A (en) | Cutter state monitoring method based on singularity Leersian index | |
| CN114871850B (en) | Tool wear state assessment method based on vibration signals and BP neural network | |
| CN110555243B (en) | Two-dimensional map construction method for vibration information of machine tool spindle in milling process | |
| CN114749996A (en) | Tool residual life prediction method based on deep learning and time sequence regression model | |
| CN113485244A (en) | Numerical control machine tool control system and method based on cutter wear prediction | |
| CN113618491A (en) | Method for establishing broach wear state recognition model | |
| Wang et al. | Thread quality classification of a tapping machine based on machine learning | |
| Xiao et al. | Deep learning based modeling for cutting energy consumed in CNC turning process | |
| CN112884027B (en) | Cutting process real-time state monitoring method and device based on pattern recognition | |
| CN112885080B (en) | Construction method for driving condition of new energy automobile | |
| Yang et al. | A novel decentralized weighted ReliefF-PCA method for fault detection | |
| CN110543908B (en) | Control chart pattern recognition method based on dynamic observation window | |
| CN112907032A (en) | Efficiency-based evaluation method for large electromagnetic ejection device | |
| Zheng et al. | An efficient method based on conditional generative adversarial networks for imbalanced fault diagnosis of rolling bearing | |
| CN117332224A (en) | Cross-domain migration-based generation countermeasure network signal generation and quality evaluation 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 |