CN113919396A - Vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning - Google Patents
Vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning Download PDFInfo
- Publication number
- CN113919396A CN113919396A CN202111191193.XA CN202111191193A CN113919396A CN 113919396 A CN113919396 A CN 113919396A CN 202111191193 A CN202111191193 A CN 202111191193A CN 113919396 A CN113919396 A CN 113919396A
- Authority
- CN
- China
- Prior art keywords
- wear
- image
- tool
- vibration signal
- cutter
- 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
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000012544 monitoring process Methods 0.000 title claims abstract description 25
- 238000013507 mapping Methods 0.000 claims abstract description 22
- 238000012545 processing Methods 0.000 claims description 24
- 238000005299 abrasion Methods 0.000 claims description 23
- 230000011218 segmentation Effects 0.000 claims description 23
- 238000005259 measurement Methods 0.000 claims description 12
- 230000004927 fusion Effects 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 230000007423 decrease Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000002596 correlated effect Effects 0.000 claims description 3
- 230000000875 corresponding effect Effects 0.000 claims description 3
- 230000003628 erosive effect Effects 0.000 claims description 3
- 238000007500 overflow downdraw method Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000000691 measurement method Methods 0.000 description 7
- 238000003754 machining Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000003801 milling Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000009529 body temperature measurement Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/04—Ageing analysis or optimisation against ageing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Signal Processing (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a method for monitoring the wear state of a machine tool cutter by using vibration signals and image characteristics through semi-supervised learning, which comprises the following steps: (1) establishing a mapping model from a tool wear image to a wear area by taking an accurate matching machine tool wear area value as a target; (2) establishing a discrimination model based on the vibration signal, and outputting the usable life of the cutter; (3) and fusing two discrimination results and decision making to realize accurate prediction of the wear state of the tool of the machine tool. Compared with a single judging means, the method can obtain a more accurate judging result, improves the service life utilization rate of the machine tool cutter, and has good adaptability and robustness for different types of machine tools and cutters.
Description
Technical Field
The invention relates to a vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning, and belongs to the field of mechanical fault diagnosis.
Background
The rapid development of the tool machining and manufacturing industry promotes the continuous development of the machine tool machining technology in China. Because the cutter can cause inevitable wear and even damage of the cutter due to natural factors or human factors and the like in the milling process, the method is of great importance for the research of more intelligent, lower-cost and more reliable detection technology for monitoring the wear state of the cutter. The detection method not only relates to the working condition of the cutter, but also more importantly meets the product precision requirement and quality requirement of the workpiece to be machined, and further, the detection of the abrasion state of the cutter is a powerful guarantee that the production efficiency is improved, the service life of machinery is prolonged, and the industrial production is not influenced.
The tool life management technology has become a key technology consistently determined at home and abroad, which not only greatly reduces the labor cost and the material cost, but also drives the progress of the machining and manufacturing industries. The efficient, simple and quick tool wear state detection technology can keep the consistency of a processing line in the high-speed milling process, and can also improve the efficiency of the whole high-speed milling process fundamentally, so that the practical application value is wider.
The existing method for monitoring the wear state of the tool of the machine tool can be divided into a direct measurement method and an indirect measurement method. Direct measurement methods such as resistance measurement, distance measurement, radiation measurement, and optical measurement are easily affected by a processing environment during measurement or are limited to a specific scene, and therefore, the accuracy of measurement is easily affected and the universality is poor. Indirect measurement methods reflect the wear degree of the tool by measuring parameters related to the tool, such as cutting force measurement methods, acoustic emission measurement methods, vibration signal measurement methods, cutting temperature measurement methods and the like, and these methods do not directly measure the state of the tool in the measurement, but indirectly represent the state of the tool by detecting other signals, and the measured signals contain a large number of interference factors, which affect the final judgment result.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning.
The invention establishes a mapping model by using the worn cutter image and the worn area value, combines the mapping model with the vibration signal and fuses multiple judgment results. The adopted measures are beneficial to realizing accurate judgment of the accurate cutter abrasion state, single judgment basis errors can be better avoided, and the service life utilization rate of the machine tool cutter is improved. The invention combines the advantages of direct measurement and indirect measurement, performs the fusion of decision level, and can be suitable for more scenes and has more stable judgment results.
Interpretation of terms:
1. pearson correlation coefficient: pearson product-moment correction coefficient, also known as PPMCC or PCCs, commonly denoted r or Pearson's r, for measuring a distance between two variables X and YCorrelation(linear correlation) with a value between-1 and 1.
2. BP network: the bp (back propagation) neural network is a multi-layer feedforward neural network trained according to an error back propagation algorithm. The BP network comprises an input layer, a hidden layer and an output layer, the square of the network error is used as an objective function, the minimum value of the objective function is calculated by adopting a gradient descent method, and the basic structure of the BP network is shown in the first half part of fig. 3.
3. Softmax classifier: the softmax function, whose input value is a vector, where the elements in the vector are the scoring values of any real number, outputs a vector, where each element value is between 0 and 1, and the sum of all elements is 1 (normalized classification probability).
To achieve the above object, the present invention adopts the following solutions:
a vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning comprises the following steps:
(1) establishing a mapping of a tool wear image and a wear area: sequentially carrying out graying, denoising, segmentation, binarization and opening operation processing on the tool wear image, and establishing mapping of the pixel number and the wear value of an image wear region;
(2) judging the wear state of the tool by using the vibration signal: the method comprises the following steps:
(2-1) preprocessing the vibration signal;
(2-2) extracting the characteristics of the vibration signal preprocessed in the step (2-1);
(2-3) performing feature selection;
(2-4) building a network model;
(3) and (3) obtaining a tool wear image prediction result through the step (1), obtaining a vibration signal prediction result through the step (2), and outputting a final result after fusion.
Preferably, in step (1), the graying process comprises the following steps:
setting the resolution of the tool wear image to be M multiplied by N, wherein three-channel pixel values at any position (i, j) are respectively R (i, j), G (i, j) and B (i, j), and after the tool wear image is subjected to graying operation, the single-channel pixel values are as follows:the gray scale k of the cutter abrasion image belongs to [0,255 ]]The number of pixel points per gray level is hk。
Preferably, in step (1), the denoising pre-processing procedure is as follows:
for any pixel I (I, j), the gray values of 8 pixels in the neighborhood are ordered from large to small as { I0,I1,…,I7Is then the pixel I (I, j) is represented as
Preferably, in step (1), the segmentation process is as follows:
dividing the tool wear image subjected to denoising pretreatment into a target part and a background part by a segmentation threshold T, and recording the number of pixel points belonging to the target as N0The number of pixels belonging to the background is recorded as N1The number of pixels belonging to the target being proportional to the whole wear image of the toolThe pixel number belonging to the background accounts for the whole cutter abrasion imageThe average gray of the pixels belonging to the object isThe average gray of the pixels belonging to the background isTotal average gray scale [ mu ] omega of tool wear image0×μ0+ω1×μ1(ii) a The inter-class variance of the tool wear image is g-omega0×(μ0-μ)2+ω1×(μ1-μ)2Equivalence is g ═ ω0ω1(μ0-μ1)2(ii) a At the moment, a segmentation threshold T which enables g to be maximum is found in a traversing mode, the cutter abrasion image is divided into a target part and a background part through the segmentation threshold T, and the target part is an extracted abrasion area of the cutter.
Preferably, in step (1), the binarization processing procedure is as follows:
according to the gray thresholdT∈[0,255]And (4) binarizing the tool wear image after the segmentation processing, and extracting a target area, namely the wear area of the tool.
According to the present invention, in the step (1), the open operation processing procedure is as follows:
after the divided wear area image is obtained, the wear area image X is subjected to erosion calculation using a structure B, where X-B is { a | B ═ BaBelongs to X, eliminates small meaningless areas, carries out expansion operation, fills damaged areas,
preferably, in step (1), a mapping between the number of pixels in the image wear region and the size of the wear value is established, specifically:
calculating the pixel number of the wear area of the cutter according to the cutter wear image obtained after binarization processing, and establishing a mapping relation corresponding to the wear area obtained by measurement;
the number of pixels in the cutter abrasion image is positively correlated with the abrasion area of the cutter, and a linear fitting relation is established;
setting the abrasion area value of the label as y ═ y1,y2,…,yn]TThe number of pixels is x ═ x1,x2,…,xn]TWhere n is the number of samples, the relationship between label and number of pixels is defined as y ═ a0+a1x, constructing an equation system as shown in the formula (I):
and b, solving to obtain a mapping relation between the pixel value of the tool wear image and the wear area.
According to the present invention, preferably, in the step (2-1), the vibration signal is preprocessed by:
and (3) removing obvious abnormity by using interception operation, wherein the obvious abnormity refers to a vibration signal of which the vibration amplitude gradually decreases from 5, and filtering and denoising are carried out by using a wiener filter.
Preferably, in the step (2-2), the extracting of the feature of the vibration signal preprocessed in the step (2-1) includes:
extracting 30 features of a time domain, a frequency domain and a time-frequency domain in total; the 30 features include a maximum, a minimum, a crest factor, a skew, a margin, a form factor, a pulse factor, a skewness, a mean, a root mean square, a variance, a kurtosis, a frequency domain mean, a frequency domain variance, and 16 4-layer wavelet packet decomposition energy features.
Preferably, in step (2-3), the feature selection is performed by:
calculating Pearson correlation coefficients of the extracted 30 features, selecting 10 features with the maximum correlation from high to low according to the value of the Pearson correlation coefficients as training sample data and test sample data, wherein X represents the extracted features, and Y represents the label of the sample, and X ═ X [ X represents the label of the sample ═ X1,X2,…,Xn],Y=[Y1,Y2,…,Yn]Then the feature mean and the tag mean are expressed asThe standard deviation is expressed asCovariance ofPearson's correlation coefficient of
According to the invention, preferably, in the step (2-4), building a network model means:
the network model comprises a BP network and a Softmax classifier, the BP network comprises an input layer, a hidden layer and an output layer, the dimension of the input layer is set to be 10, the dimension of the hidden layer is set to be 20, the dimension of the output layer is set to be 5, and the Softmax classifier is used for classifying the BP network into 3 classes.
According to the invention, preferably, the tool wear image prediction result is obtained through the step (1), the vibration signal prediction result is obtained through the step (2), and the final result is output after fusion, wherein the specific implementation process comprises the following steps:
(3-1) obtaining the result of the tool wear image detection
Acquiring a wear image of a machine tool cutter according to the step (1), sequentially carrying out graying, denoising, segmentation, binarization and opening operation processing on the wear image of the machine tool cutter to obtain an accurate wear region, and outputting a wear value of the machine tool cutter according to the established mapping relation between the pixel number of the wear region of the image and the wear value;
(3-2) obtaining vibration signal prediction results
According to the step (2), preprocessing and feature extraction are carried out on the vibration signals in sequence, the vibration signals are input into a trained network model, and the wear value of the machine tool cutter obtained through the vibration signals is output;
(3-3) fusing the two decisions to output the final result
Establishing a decision fusion method according to the judgment decision obtained in the step (3-1) and the step (3-2), outputting the tool wear value after fusion, and setting the result obtained after the tool wear image in the step (3-1) is mapped as R1And (3) outputting a vibration signal and a network model in the step (3-2) to obtain a result R2The final output result is expressed as a weighted sum of the two, the weight is set as alpha, and the value range of the alpha is [0,1 ]]Then the final output result is R ═ α R1+(1-α)R2。
A computer device comprising a memory storing a computer program and a processor implementing the steps of a method for monitoring the state of wear of a machine tool based on vibration signals and image characteristics of semi-supervised learning when executing said computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method for monitoring the state of wear of a machine tool on the basis of vibration signals and image characteristics of semi-supervised learning.
The invention has the beneficial effects that:
when the wear state of the machine tool machining cutter is detected, the semi-supervised learning vibration signal and the image characteristic machine tool cutter wear state monitoring method are used, the mapping relation from the wear image of the cutter to the wear area detection is established, the vibration signal is used for simultaneously extracting the characteristic to predict the wear value, two decision results are fused, the error of a single judgment method is compensated, the wear state of the cutter can be more accurately judged, the service life utilization rate of the cutter is improved, and meanwhile, the method has good adaptability and robustness to different machine tools and cutters.
Drawings
FIG. 1 is a schematic flow chart of a vibration signal and image characteristic machine tool wear state monitoring method based on semi-supervised learning according to the present invention;
FIG. 2 is a block diagram of the architecture of the network model of the present invention;
FIG. 3 is a schematic diagram of a network model according to the present invention.
Detailed Description
The invention is further defined in the following, but not limited to, the figures and examples in the description.
Example 1
A vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning is disclosed, as shown in figure 1, and comprises the following steps:
(1) establishing a mapping of a tool wear image and a wear area: sequentially carrying out graying, denoising, segmentation, binarization and opening operation processing on the tool wear image, and establishing mapping of the pixel number and the wear value of an image wear region;
(2) judging the wear state of the tool by using the vibration signal: the method comprises the following steps:
(2-1) preprocessing the vibration signal;
(2-2) extracting the characteristics of the vibration signal preprocessed in the step (2-1);
(2-3) performing feature selection;
(2-4) building a network model;
(3) and (3) obtaining a tool wear image prediction result through the step (1), obtaining a vibration signal prediction result through the step (2), and outputting a final result after fusion.
Example 2
The vibration signal and image characteristic machine tool wear state monitoring method based on semi-supervised learning is different from the vibration signal and image characteristic machine tool wear state monitoring method based on embodiment 1 in that:
in the step (1), the graying process is as follows: setting the resolution of the tool wear image to be M multiplied by N, wherein three-channel pixel values at any position (i, j) are respectively R (i, j), G (i, j) and B (i, j), and after the tool wear image is subjected to graying operation, the single-channel pixel values are as follows:the gray scale k of the cutter abrasion image belongs to [0,255 ]]The number of pixel points per gray level is hk。
In the step (1), the denoising pretreatment process is as follows: for any pixel I (I, j), the gray values of 8 pixels in the neighborhood are ordered from large to small as { I0,I1,…,I7Is then the pixel I (I, j) is represented as
In the step (1), the segmentation process is as follows: the 8-bit represented grayscale image level k ∈ [0,255 ]]Determining a segmentation threshold T by adopting a threshold segmentation method, dividing the tool wear image subjected to denoising pretreatment into a target part and a background part by using the segmentation threshold T, and recording the number of pixel points belonging to the target as N0The number of pixels belonging to the background is recorded as N1The number of pixels belonging to the target being proportional to the whole wear image of the toolThe pixel number belonging to the background accounts for the whole cutter abrasion imageImages belonging to an objectThe average gray scale of the pixels isThe average gray of the pixels belonging to the background isTotal average gray scale [ mu ] omega of tool wear image0×μ0+ω1×μ1(ii) a The inter-class variance of the tool wear image is g-omega0×(μ0-μ)2+ω1×(μ1-μ)2Equivalence is g ═ ω0ω1(μ0-μ1)2(ii) a At the moment, a segmentation threshold T which enables g to be maximum is found in a traversing mode, the cutter abrasion image is divided into a target part and a background part through the segmentation threshold T, and the target part is an extracted abrasion area of the cutter.
In the step (1), the binarization processing process is as follows: according to the gray thresholdT∈[0,255]And (4) binarizing the tool wear image after the segmentation processing, and extracting a target area, namely the wear area of the tool.
In the step (1), the opening operation processing process comprises the following steps: after the segmented wear area image is obtained, further morphological processing is performed to obtain a more accurate segmented area. The erosion calculation is performed on the wear area image X using a structure B, where X-B is { a | B }aE.g., X, small meaningless regions are removed, and an expansion operation is performed to fill the damaged area, where X ^ B ^ a | B ^ Ba↑X}。
In the step (1), a mapping of the number of pixels in the image wear area and the size of the wear value is established, specifically:
calculating the pixel number of the wear area of the cutter according to the cutter wear image obtained after binarization processing, and establishing a mapping relation corresponding to the wear area obtained by measurement;
the number of pixels in the cutter abrasion image is positively correlated with the abrasion area of the cutter, and a linear fitting relation is established;
setting the abrasion area value of the label as y ═ y1,y2,…,yn]TThe number of pixels is x ═ x1,x2,…,xn]TWhere n is the number of samples, the relationship between label and number of pixels is defined as y ═ a0+a1x, constructing an equation system as shown in the formula (I):
and b, solving to obtain a mapping relation between the pixel value of the tool wear image and the wear area.
Using sum of squares of errorsRoot mean square errorDetermining coefficientsThe effect of the fit was evaluated.
In the step (2-1), the vibration signal is preprocessed, which means that:
the collected vibration signals generated during machine tool machining cannot be directly used, and pretreatment is needed. The method comprises the steps that abnormal data can be generated due to the self problem of a sensor in collected vibration signals during machine tool machining, the vibration signals under normal conditions are collected by an acceleration sensor, the amplitude of vibration is recorded and fluctuates up and down at a value of 0, the abnormal signals can gradually decline from the amplitude of 5 due to the power-on self-detection of the sensor, and normal signals are collected after calibration is completed. Therefore, the interception operation is used for removing obvious abnormity, wherein the obvious abnormity refers to a vibration signal of which the amplitude of vibration gradually decreases from 5, and a wiener filter is used for filtering and denoising. The method has the advantages of filtering noise or false components in the vibration signal, improving the signal-to-noise ratio, smoothing analysis data, suppressing interference signals, sharing frequency components and the like.
In the step (2-2), extracting the characteristics of the vibration signal preprocessed in the step (2-1) refers to:
and in order to facilitate subsequent processing, the data volume is reduced, and feature extraction is carried out on the vibration signal.
Extracting 30 features of a time domain, a frequency domain and a time-frequency domain in total; the 30 features include a maximum, a minimum, a crest factor, a skew, a margin, a form factor, a pulse factor, a skewness, a mean, a root mean square, a variance, a kurtosis, a frequency domain mean, a frequency domain variance, and 16 4-layer wavelet packet decomposition energy features. The 16 wavelet packet decomposition energy characteristics of 4 layers refer to that: wavelet packet decomposition is carried out on the signal, the frequency band is divided into multiple layers, and the 4 th layer is decomposed into 16 energy characteristics.
Maximum max ═ max { s1,s2,...sN};
Min is min { s ═ minimum value1,s2,...sN};
In the formula, N represents the number of signal points, siRepresenting the amplitude of the time-domain signal, fiIndicating the amplitude of the frequency domain signal and i indicating the sequence number of the ith signal in the signal sequence.
In the step (2-3), the feature selection is performed, which means that:
calculating Pearson correlation coefficients of the extracted 30 features, selecting 10 features with the maximum correlation from high to low according to the value of the Pearson correlation coefficients as training sample data and test sample data, wherein X represents the extracted features, and Y represents the label of the sample, and X ═ X [ X represents the label of the sample ═ X1,X2,…,Xn],Y=[Y1,Y2,…,Yn]Then the feature mean and the tag mean are expressed asThe standard deviation is expressed asCovariance ofPearson's correlation coefficient of
In the step (2-4), a network model is built, which means that:
and predicting the wear state of the cutter by using a neural network building model according to the features extracted and selected by the vibration signals, wherein the complexity of the network can be greatly reduced because the features are selected and the input of the network is the feature dimension with higher correlation, and a single hidden layer network is selected for training to output the wear value of the cutter.
As shown in fig. 2 and 3, the network model includes a BP network and a Softmax classifier, the BP network includes an input layer, an implied layer, and an output layer, the dimension of the input layer is set to 10, the dimension of the implied layer is set to 20, the dimension of the output layer is set to 5, and the network model is classified into 3 classes using the Softmax classifier.
The samples in the data set are divided into three parts of initial wear, normal wear and rapid wear according to the wear value of the label.
Obtaining a tool wear image prediction result through the step (1), obtaining a vibration signal prediction result through the step (2), and outputting a final result after fusion, wherein the specific implementation process comprises the following steps:
(3-1) obtaining the result of the tool wear image detection
Acquiring a wear image of a machine tool cutter according to the step (1), sequentially carrying out graying, denoising, segmentation, binarization and opening operation processing on the wear image of the machine tool cutter to obtain an accurate wear region, and outputting a wear value of the machine tool cutter according to the established mapping relation between the pixel number of the wear region of the image and the wear value;
(3-2) obtaining vibration signal prediction results
According to the step (2), preprocessing and feature extraction are carried out on the vibration signals in sequence, at the moment, only the highly relevant features reserved after feature selection need to be extracted and input into a trained network model, and the wear value of the machine tool cutter obtained through the vibration signals is output;
(3-3) fusing the two decisions to output the final result
Establishing a decision fusion method according to the judgment decision obtained in the step (3-1) and the step (3-2), outputting the tool wear value after fusion, and setting the result obtained after the tool wear image in the step (3-1) is mapped as R1And (3) outputting a vibration signal and a network model in the step (3-2) to obtain a result R2The final output result is expressed as a weighted sum of the two, the weight is set as alpha, and the value range of the alpha is [0,1 ]]Then the final output result is R ═ α R1+(1-α)R2。
Table 1 shows the comparison result of the single-decision tool wear state recognition rate and the recognition rate by the method of the present invention on the tool wear data sets of the domestic and foreign sources.
TABLE 1
Rate of accuracy | PHM2010 | SDU-QIT | Milling cutter data of the company Wida |
Single decision | 90.37% | 86.19% | 82.03% |
The method of the invention | 98.44% | 95.86% | 93.66% |
As can be seen from Table 1, on these data sets, the method of the present invention has greater superiority in accuracy of item wear state identification compared with a single decision result, and in addition, the method is also better in universality and robustness for different types of tools.
Claims (10)
1. A vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning is characterized by comprising the following steps:
(1) establishing a mapping of a tool wear image and a wear area: sequentially carrying out graying, denoising, segmentation, binarization and opening operation processing on the tool wear image, and establishing mapping of the pixel number and the wear value of an image wear region;
(2) judging the wear state of the tool by using the vibration signal: the method comprises the following steps:
(2-1) preprocessing the vibration signal;
(2-2) extracting the characteristics of the vibration signal preprocessed in the step (2-1);
(2-3) performing feature selection;
(2-4) building a network model;
(3) and (3) obtaining a tool wear image prediction result through the step (1), obtaining a vibration signal prediction result through the step (2), and outputting a final result after fusion.
2. The vibration signal and image feature machine tool wear state monitoring method based on semi-supervised learning according to claim 1, wherein in the step (2-2), the feature of the vibration signal preprocessed in the step (2-1) is extracted, and is characterized in that:
extracting 30 features of a time domain, a frequency domain and a time-frequency domain in total; the 30 features include a maximum, a minimum, a crest factor, a skew, a margin, a form factor, a pulse factor, a skewness, a mean, a root mean square, a variance, a kurtosis, a frequency domain mean, a frequency domain variance, and 16 4-layer wavelet packet decomposition energy features.
3. The vibration signal and image feature machine tool wear state monitoring method based on semi-supervised learning according to claim 2, wherein in the step (2-3), feature selection is performed, and is characterized in that:
calculating Pearson correlation coefficients of the extracted 30 features, selecting 10 features with the maximum correlation from high to low according to the value of the Pearson correlation coefficients as training sample data and test sample data, wherein X represents the extracted features, and Y represents the label of the sample, and X ═ X [ X represents the label of the sample ═ X1,X2,…,Xn],Y=[Y1,Y2,…,Yn]Then the feature mean and the tag mean are expressed asThe standard deviation is expressed asCovariance ofPearson's correlation coefficient of
4. The vibration signal and image feature machine tool cutter wear state monitoring method based on semi-supervised learning according to claim 1, wherein in the step (2-4), building a network model refers to:
the network model comprises a BP network and a Softmax classifier, the BP network comprises an input layer, a hidden layer and an output layer, the dimension of the input layer is set to be 10, the dimension of the hidden layer is set to be 20, the dimension of the output layer is set to be 5, and the Softmax classifier is used for classifying the BP network into 3 classes.
5. The method for monitoring the wear state of the tool of the machine tool based on the vibration signal and the image characteristic of the semi-supervised learning according to the claim 1, is characterized in that a tool wear image prediction result is obtained through the step (1), a vibration signal prediction result is obtained through the step (2), and a final result is output after fusion, and the specific implementation process comprises the following steps:
(3-1) obtaining the result of the tool wear image detection
Acquiring a wear image of a machine tool cutter according to the step (1), sequentially carrying out graying, denoising, segmentation, binarization and opening operation processing on the wear image of the machine tool cutter to obtain an accurate wear region, and outputting a wear value of the machine tool cutter according to the established mapping relation between the pixel number of the wear region of the image and the wear value;
(3-2) obtaining vibration signal prediction results
According to the step (2), preprocessing and feature extraction are carried out on the vibration signals in sequence, the vibration signals are input into a trained network model, and the wear value of the machine tool cutter obtained through the vibration signals is output;
(3-3) fusing the two decisions to output the final result
Establishing a decision fusion method according to the judgment decision obtained in the step (3-1) and the step (3-2), outputting the tool wear value after fusion, and setting the result obtained after the tool wear image in the step (3-1) is mapped as R1And (3) outputting a vibration signal and a network model in the step (3-2) to obtain a result R2The final output result is expressed as a weighted sum of the two, the weight is set as alpha, and the value range of the alpha is [0,1 ]]Then the final output result is R ═ α R1+(1-α)R2。
6. The method for monitoring the wear state of the tool of the machine tool based on the vibration signal and the image characteristic learned through semi-supervised learning according to the claim 1, wherein in the step (1), a mapping of the pixel number and the wear value size of an image wear region is established, specifically:
calculating the pixel number of the wear area of the cutter according to the cutter wear image obtained after binarization processing, and establishing a mapping relation corresponding to the wear area obtained by measurement;
the number of pixels in the cutter abrasion image is positively correlated with the abrasion area of the cutter, and a linear fitting relation is established;
setting the abrasion area value of the label as y ═ y1,y2,…,yn]TThe number of pixels is x ═ x1,x2,…,xn]TWhere n is the number of samples, the relationship between label and number of pixels is defined as y ═ a0+a1x, constructing an equation system as shown in the formula (I):
and b, solving to obtain a mapping relation between the pixel value of the tool wear image and the wear area.
7. The vibration signal and image characteristic machine tool wear state monitoring method based on semi-supervised learning according to claim 1, wherein in the step (1), the graying processing procedure is as follows:
setting the resolution of the tool wear image to be M multiplied by N, wherein three-channel pixel values at any position (i, j) are respectively R (i, j), G (i, j) and B (i, j), and after the tool wear image is subjected to graying operation, the single-channel pixel values are as follows:the gray scale k of the cutter abrasion image belongs to [0,255 ]]The number of pixel points per gray level is hk。
8. The method for monitoring the wear state of the tool of the machine tool based on the vibration signal and the image characteristic of the semi-supervised learning as recited in claim 1, wherein in the step (1), the denoising preprocessing process comprises the following steps:
9. The vibration signal and image feature machine tool wear state monitoring method based on semi-supervised learning according to claim 1, wherein in the step (1), the segmentation processing procedure is as follows:
dividing the tool wear image subjected to denoising pretreatment into a target part and a background part by a segmentation threshold T, and recording the number of pixel points belonging to the target as N0The number of pixels belonging to the background is recorded as N1The number of pixels belonging to the target being proportional to the whole wear image of the toolThe pixel number belonging to the background accounts for the whole cutter abrasion imageThe average gray of the pixels belonging to the object isThe average gray of the pixels belonging to the background isTotal average gray scale [ mu ] omega of tool wear image0×μ0+ω1×μ1(ii) a The inter-class variance of the tool wear image is g-omega0×(μ0-μ)2+ω1×(μ1-μ)2Equivalence is g ═ ω0ω1(μ0-μ1)2(ii) a At the moment, a segmentation threshold T which enables g to be maximum is found in a traversing mode, the cutter abrasion image is divided into a target part and a background part through the segmentation threshold T, and the target part is an extracted abrasion area of the cutter.
10. The method for monitoring the wear state of the tool of the machine tool based on the vibration signal and the image characteristic of the semi-supervised learning as claimed in claim 1, wherein in the step (1), the binarization processing procedure is as follows:
according to the gray thresholdCarrying out binarization on the tool wear image subjected to the segmentation processing, and extracting a target area, namely the wear area of the tool;
in the step (1), the opening operation processing process comprises the following steps:
after the divided wear area image is obtained, the wear area image X is subjected to erosion calculation using a structure B, where X-B is { a | B ═ BaE.g., X, small meaningless regions are removed, and an expansion operation is performed to fill the damaged area, where X ^ B ^ a | B ^ Ba↑X};
In the step (2-1), the vibration signal is preprocessed, which means that:
and (3) removing obvious abnormity by using interception operation, wherein the obvious abnormity refers to a vibration signal of which the vibration amplitude gradually decreases from 5, and filtering and denoising are carried out by using a wiener filter.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111191193.XA CN113919396A (en) | 2021-10-13 | 2021-10-13 | Vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111191193.XA CN113919396A (en) | 2021-10-13 | 2021-10-13 | Vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113919396A true CN113919396A (en) | 2022-01-11 |
Family
ID=79240058
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111191193.XA Pending CN113919396A (en) | 2021-10-13 | 2021-10-13 | Vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113919396A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116952765A (en) * | 2023-09-19 | 2023-10-27 | 西南交通大学 | Method for inhibiting and quantitatively evaluating polygonal abrasion of wagon wheels |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160282098A1 (en) * | 2013-09-17 | 2016-09-29 | Bae Systems Plc | Method and apparatus for characterising fragmentation of an explosive device |
CN108882896A (en) * | 2015-09-23 | 2018-11-23 | 诺瓦达克技术公司 | For evaluating the method and system of the healing of tissue |
CN109514349A (en) * | 2018-11-12 | 2019-03-26 | 西安交通大学 | Monitoring Tool Wear States in Turning based on vibration signal and Stacking integrated model |
CN109571141A (en) * | 2018-11-01 | 2019-04-05 | 北京理工大学 | A kind of Monitoring Tool Wear States in Turning based on machine learning |
CN113111703A (en) * | 2021-03-02 | 2021-07-13 | 郑州大学 | Airport pavement disease foreign matter detection method based on fusion of multiple convolutional neural networks |
US20210287098A1 (en) * | 2019-12-24 | 2021-09-16 | Dalian University Of Technology | On-line prediction method of surface roughness of parts based on sdae-dbn algorithm |
CN113485244A (en) * | 2021-07-27 | 2021-10-08 | 同济大学 | Numerical control machine tool control system and method based on cutter wear prediction |
CN113496312A (en) * | 2021-07-05 | 2021-10-12 | 山东大学 | Cutter residual life prediction method and equipment based on multi-dimensional feature extraction fusion and long-term and short-term memory network and storage medium |
-
2021
- 2021-10-13 CN CN202111191193.XA patent/CN113919396A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160282098A1 (en) * | 2013-09-17 | 2016-09-29 | Bae Systems Plc | Method and apparatus for characterising fragmentation of an explosive device |
CN108882896A (en) * | 2015-09-23 | 2018-11-23 | 诺瓦达克技术公司 | For evaluating the method and system of the healing of tissue |
CN109571141A (en) * | 2018-11-01 | 2019-04-05 | 北京理工大学 | A kind of Monitoring Tool Wear States in Turning based on machine learning |
CN109514349A (en) * | 2018-11-12 | 2019-03-26 | 西安交通大学 | Monitoring Tool Wear States in Turning based on vibration signal and Stacking integrated model |
US20210287098A1 (en) * | 2019-12-24 | 2021-09-16 | Dalian University Of Technology | On-line prediction method of surface roughness of parts based on sdae-dbn algorithm |
CN113111703A (en) * | 2021-03-02 | 2021-07-13 | 郑州大学 | Airport pavement disease foreign matter detection method based on fusion of multiple convolutional neural networks |
CN113496312A (en) * | 2021-07-05 | 2021-10-12 | 山东大学 | Cutter residual life prediction method and equipment based on multi-dimensional feature extraction fusion and long-term and short-term memory network and storage medium |
CN113485244A (en) * | 2021-07-27 | 2021-10-08 | 同济大学 | Numerical control machine tool control system and method based on cutter wear prediction |
Non-Patent Citations (3)
Title |
---|
林杨;高思煜;刘同舜;朱锟鹏;: "基于深度学习的高速铣削刀具磨损状态预测方法", 机械与电子, no. 07, 24 July 2017 (2017-07-24) * |
陈刚;焦黎;颜培;王西彬;史雪春;彭振新;: "基于多传感器数据融合的刀具磨损状态监测研究", 新技术新工艺, no. 11, 25 November 2017 (2017-11-25) * |
陈启鹏;谢庆生;袁庆霓;黄海松;魏琴;李宜汀;: "基于深度门控循环单元神经网络的刀具磨损状态实时监测方法", 计算机集成制造***, no. 07, 15 July 2020 (2020-07-15) * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116952765A (en) * | 2023-09-19 | 2023-10-27 | 西南交通大学 | Method for inhibiting and quantitatively evaluating polygonal abrasion of wagon wheels |
CN116952765B (en) * | 2023-09-19 | 2023-12-12 | 西南交通大学 | Method for inhibiting and quantitatively evaluating polygonal abrasion of wagon wheels |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111340791A (en) | Photovoltaic module unsupervised defect detection method based on GAN improved algorithm | |
CN102928435A (en) | Aircraft skin damage identification method and device based on image and ultrasound information fusion | |
CN108711148B (en) | Tire defect intelligent detection method based on deep learning | |
CN115797354B (en) | Method for detecting appearance defects of laser welding seam | |
CN116758491B (en) | Printing monitoring image analysis method and system applied to 3D printing | |
CN111815555A (en) | Metal additive manufacturing image detection method and device combining anti-neural network with local binary | |
CN113469951B (en) | Hub defect detection method based on cascade region convolutional neural network | |
CN106198749B (en) | A kind of data fusion method of the multiple sensor based on Metal Crack monitoring | |
CN117152161B (en) | Shaving board quality detection method and system based on image recognition | |
CN114619292A (en) | Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network | |
CN112200225A (en) | Steel rail damage B display image identification method based on deep convolutional neural network | |
CN111783616B (en) | Nondestructive testing method based on data-driven self-learning | |
WO2021109011A1 (en) | Intelligent capacitor internal defect detection method based on ultrasound image | |
CN116563277B (en) | Machine vision-based copper-clad aluminum surface defect detection method and system | |
CN115294033A (en) | Tire belt layer difference level and misalignment defect detection method based on semantic segmentation network | |
CN116740044B (en) | Copper pipe milling surface processing method and system based on visual detection and control | |
CN117173461A (en) | Multi-visual task filling container defect detection method, system and medium | |
CN111753877B (en) | Product quality detection method based on deep neural network migration learning | |
CN115861190A (en) | Comparison learning-based unsupervised defect detection method for photovoltaic module | |
CN113919396A (en) | Vibration signal and image characteristic machine tool cutter wear state monitoring method based on semi-supervised learning | |
CN111695582A (en) | Flutter texture detection method and device | |
CN115294009A (en) | Method and equipment for detecting welding defects of battery tabs based on machine learning and storage medium | |
CN117269168B (en) | New energy automobile precision part surface defect detection device and detection method | |
CN116777836A (en) | Multi-mode data driving-based injection molding process product quality detection method | |
CN114119562B (en) | Brake disc outer surface defect detection method and system based on deep learning |
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 |