CN111421386A - Method for monitoring wear state of micro-milling cutter based on manifold learning method - Google Patents
Method for monitoring wear state of micro-milling cutter based on manifold learning method Download PDFInfo
- Publication number
- CN111421386A CN111421386A CN202010406461.4A CN202010406461A CN111421386A CN 111421386 A CN111421386 A CN 111421386A CN 202010406461 A CN202010406461 A CN 202010406461A CN 111421386 A CN111421386 A CN 111421386A
- Authority
- CN
- China
- Prior art keywords
- particle
- neural network
- signal
- time domain
- micro
- 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
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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a wear state monitoring method of a micro-milling cutter based on a manifold learning method, wherein acceleration signals acquired by an acceleration sensor arranged on a main shaft of a vertical milling machine in the vertical milling cutting process are represented as X ═ X (1), X (2), …, X (n), n represents the signal length, and the time domain characteristics of cutting time signals are extracted; performing dimension reduction and reduction processing on the extracted time domain features, and processing the time domain features by using a manifold learning method to obtain 2 dimension-reduced time domain feature quantities with strong correlation; inputting the characteristic quantity after the dimensionality reduction into a BP neural network for training, optimizing the neural network by a particle swarm algorithm, classifying signals generated by cutters with different wear degrees by the neural network according to the nonlinear mapping capability of the neural network, and accurately judging the category of an unknown signal. The invention has higher recognition rate to different abrasion states and can find the violent abrasion of the cutter earlier.
Description
Technical Field
The invention relates to the technical field of cutter wear monitoring, in particular to a method for monitoring the wear state of a micro-milling cutter based on a manifold learning method.
Background
The micro-milling technology has the advantages of diversity of processing materials and unique advantage of three-dimensional curved surface processing, and has wide application in the aspect of processing parts in a microscale. However, micro milling has the characteristics of small tool size, high cutting speed and discontinuous cutting, so that the micro milling cutter is worn quickly, and the precision and the surface quality of a product are influenced. Severe tool wear can also cause tool breakage, breakage and chatter, resulting in damage to the machine tool. Therefore, it is highly desirable to provide an effective tool wear monitoring method. The early research work mainly includes two aspects: 1) the direct method comprises the following steps: processing the milling cutter abrasion image by a digital image processing technology to realize the abrasion monitoring of the cutter; 2) an indirect method: tool wear is predicted by signal processing, feature extraction and classification of cutting force, vibration and acoustic emission signals.
The direct method mainly obtains a milling cutter abrasion image by means of a high-precision high-speed camera, obtains cutter abrasion characteristics by means of a digital image processing technology, and classifies cutter abrasion degrees by means of algorithms such as a support vector machine and an artificial neural network. It can be seen that the direct method can obtain the real state of the tool abrasion through an advanced high-speed camera, and has higher precision. However, the coolant and chips are not conducive to obtaining a high quality image of tool wear, and in addition, the tool is in continuous contact with the workpiece during machining, and machine vision is difficult to obtain an image of the cutting area in real time, which requires the tool to be detached from the spindle, possibly causing tool misalignment in the next operation.
To overcome the problems of the direct method, some researchers have proposed an indirect method based on process sensor signals. The indirect method mainly comprises sensor selection, feature extraction and feature classification, wherein the sensing signals used for monitoring the cutter wear comprise cutting force, vibration, acoustic emission and a multi-sensing system. Due to the characteristics of close connection between the vibration signal and the machining process and convenience in installation, the vibration signal is selected for micro-milling wear monitoring.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides the method for monitoring the wear state of the micro-milling cutter based on the manifold learning method, which is convenient for feature classification, improves the classification precision of a BP neural network, is closely connected with the machining process and is convenient to install.
The purpose of the invention is realized by the following technical scheme.
A method for monitoring the wear state of a micro-milling cutter based on a manifold learning method comprises the following steps:
s1, acquiring an acceleration signal in the vertical milling and cutting process through an acceleration sensor mounted on a main shaft of the vertical milling machine, and expressing the acceleration signal as X ═ X (1), X (2), …, X (n), wherein n represents the signal length;
s2, performing time domain analysis on the signal in the cutting state, and extracting time domain signal characteristic quantity:
wherein n is the signal length; i is a scale parameter of signal continuity; x is a certain section of signal extreme value; the function max (| x |) is the maximum absolute value in a certain section of signal;
s3, performing feature space reduction on the extracted multiple time domain feature quantities to obtain a more concise time domain feature space with better correlation;
s4, training the neural network by using the reduced feature space for classification, determining the structure of the BP neural network to be 2-3-4 according to the characteristics of the cutter abrasion signal, and initializing the weight and the threshold of the BP neural network by using a particle swarm algorithm; next, training a BP neural network by using training data, and adjusting the weight and the threshold of the network according to errors in the training process; analyzing the BP neural network classification capability according to the classification result by using the trained BP neural network classification tool wear state characteristic signal;
and S5, classifying the characteristic signal test data of different tool wear states by using the trained PSO-BP neural network to obtain a BP neural network classification error map and BP neural network classification accuracy.
In step S3, a laplacian feature mapping in the manifold learning method is used to perform dimensionality reduction, so as to obtain 2 time domain feature quantities with better correlation.
The Laplace feature mapping utilizes L aplarian-Beltrami operators in differential geometry to measure the functional smoothness on the manifold, namely the smoothness degree is a standard for measuring the good and bad dimension reduction effect, a non-directional weighted adjacent graph is constructed to represent the manifold, and the graph is displayed in a space with small dimension.
The dimensionality reduction step of the Laplace feature map is as follows:
1) constructing an undirected weighted graph: constructing a graph from all points using a method that sets an edge between two data points if the distance between the two data points meets the proximity threshold requirement, otherwise, no edge exists between the two points;
2) determining edge weights of the laplacian graph: the edge weight is calculated in a simplified mode, and when the node a and the node b have edges, w isabIs 1, otherwise wabIs 0, wherein wabIs the edge weight;
3) and (4) solving the characteristic mapping, namely firstly, carrying out generalized eigenvalue decomposition on the Laplace undirected graph as follows, wherein L Y is lambda DY, Y is a matrix after dimensionality reduction, and D isab=∑bWabL is the Laplace matrix corresponding to Laplace undirected graph, L is D-W, and the mapping result Y after dimensionality reduction is the eigenvector V corresponding to the minimum m non-zero eigenvalues1,V2,...,VmForm, then original high-dimensional data tuple xiIn the low dimension the manifold can be represented as: y isi=[V1,V2,…,Vm]T。
5. The method for monitoring the wear state of the micro-milling cutter based on the manifold learning method as claimed in claim 1, wherein the particle group algorithm step in the step S4 comprises: initializing a particle swarm; calculating the fitness of each particle; updating the individual extreme value and the global extreme value according to the fitness, and updating the particle speed and the particle position according to the fitness; and when the maximum iteration times are reached or the adaptive value corresponding to the global optimal particle is smaller than a specified threshold value, ending, otherwise, recalculating the fitness of each particle.
The particle swarm algorithm seeks an optimal solution according to two steps of initialization and iteration, and two extreme values are followed in the iteration to change the particle swarm algorithm, wherein the two extreme values are respectively called as individual extreme values pbesti(i ═ 1, 2.. N) and a global extremum gbest; individual extremum pbesti(i ═ 1, 2.. N) is the optimal solution explored by a certain particle, and N is the number of particles; all-purposeLocal extreme value gbest is the optimal solution searched out at the moment of all the populations; in D-dimensional space, the velocity and position of each particle i can be represented as Vi=(Vi1,Vi2,...,ViD) And Xi=(Xi1,Xi2,...,XiD) Similarly, the individual extremum can be expressed as pbesti=(pi1,pi2,...,piD) The global extremum is expressed as gbest ═ g1,g2,...,gD) So far, we can give the particle update formulas in two iterative processes as follows:
is the d-dimension velocity update formula of the particle i in k rounds, soIs the component of a certain particle velocity vector in a certain iteration, c1And c2Is an acceleration constant, also called acceleration particle, used to adjust the maximum step length of learning; r is1And r2Is two [0, 1 ]]A random function of (a) to increase search randomness; w is used to change the probed area for the solution space;is the d-th dimension position update formula of the particle i,representing the d-dimension component of the location vector of the particle i at the k-th iteration.
Compared with the prior art, the invention has the advantages that: 1) the micro-milling vibration signal feature extraction method based on manifold learning can extract essential features closely related to the wear state of the cutter, and facilitates feature classification;
2) and each weight and threshold of the BP neural network model are optimized based on a particle swarm optimization algorithm, so that the classification precision of the BP neural network is greatly improved.
Drawings
FIG. 1 is a topology structure diagram of a BP neural network according to the present invention;
FIG. 2 is a flow chart of the PSO algorithm;
FIG. 3 is a PSO-BP algorithm optimization flow chart.
Detailed Description
The invention is described in detail below with reference to the drawings and specific examples.
A wear state monitoring method of a micro-milling cutter based on a manifold learning method comprises the following steps:
s1, acquiring an acceleration signal in the vertical milling and cutting process through an acceleration sensor mounted on a main shaft of the vertical milling machine, and expressing the acceleration signal as X ═ X (1), X (2), …, X (n), wherein n represents the signal length;
s2, performing time domain analysis on the signal in the cutting state, and extracting time domain signal characteristic quantity:
the method comprises the following steps of S3, performing feature space reduction on a plurality of extracted time domain feature quantities to obtain a more concise time domain feature space with better correlation, and performing dimension reduction on Laplacian feature mapping (L E) in a manifold learning method to obtain 2 time domain feature quantities with better correlation, wherein the Laplacian feature mapping measures the function smoothness on the manifold by using L aplarian-Beltrami operators in differential geometry, namely the smoothness is a standard for measuring the good and bad dimension reduction effect, constructing an undirected weighted adjacent graph to represent the manifold, and implementing the graph in a space with small dimension.
(1) Constructing an undirected weighted graph: using some method to construct a graph of all points, if twoAnd if the distance between the data points reaches the requirement of the adjacent threshold value, setting an edge between the two points, otherwise, not having an edge between the two points. The method adopts a neighbor method, for example, the distance between a node a and a node b satisfies | | da-dbIf | l <, then an edge is considered to exist between nodes a and b, otherwise it does not exist.
(2) Determining edge weights of the laplacian graph: the edge weight is calculated in a simplified mode, and when the node a and the node b have edges, w isabIs 1, otherwise wabIs 0.
Wherein wabIs the edge weight.
(3) Solving the characteristic mapping: firstly, the generalized eigenvalue decomposition is carried out on the Laplace undirected graph as follows:
LY=λDY
wherein Y is a reduced-dimension matrix, Dab=∑bWabL is the Laplace matrix corresponding to Laplace undirected graph, L is D-W, and the mapping result Y after dimensionality reduction is the eigenvector V corresponding to the minimum m non-zero eigenvalues1,V2,...,VmAnd (4) forming.
Then the original high-dimensional data tuple xiIn the low dimension the manifold can be represented as:
yi=[V1,V2,…,Vm]T
and S4, training the neural network by the reduced feature space for classification. The structure of the BP neural network is determined to be 2-3-4 according to the characteristics of the cutter wear signal, and the topological structure diagram of the BP neural network is shown in figure 1.
The four different wear state characteristic signals of the cutter extracted by the manifold learning method are respectively marked by 1, 2, 3 and 4, the 1 st dimension of each group of data is a category mark, and the subsequent dimension is a cutter wear characteristic signal. Combining the four types of tool wear characteristic signals into a group, randomly selecting 1500 groups of data as training data, randomly selecting 500 groups of data as test data, and carrying out normalization processing on the training data. And setting an expected output value of each group of tool wear state signals according to the tool wear state category identification, wherein if the identification category is 1, an expected output vector is [ 1000 ].
Determining the structure of the BP neural network to be 2-3-4 according to the characteristic signal characteristics of the wear state of the cutter, and initializing the weight and the threshold of the BP neural network by utilizing a Particle Swarm Optimization (PSO).
The Particle Swarm algorithm, also known as Particle Swarm Optimization (Particle Swarm Optimization PSO) algorithm, was proposed by J.Kennedy et al in 1995, PSO mimicking bird Swarm predation behavior. A flock of birds randomly forages in a defined area, all birds having no knowledge of the specific location of the food, and the area having only one food, but they know the distance to the food, the optimal strategy for finding a food is to explore the area around the bird that is closest to the food at that time. The PSO derives from the above behavior and uses it to process the optimization problem, each solution of which is a single bird in the solution space that can be called a "particle", and each particle has its fitness value that can be determined by the optimization function, while all particles have-velocity, respectively, to determine their deviation and distance of flight, so that they can be probed in solution space based on the optimal particle at that time.
The PSO can seek the optimal solution according to two steps of initialization and iteration, and needs to follow two extreme values in the iteration to change the PSO, wherein the two extreme values are respectively called individual extreme values pbesti(i ═ 1, 2.. N) and a global extremum gbest. The former is the optimal solution explored by a certain particle, and N is the number of particles; the latter is the best solution searched out at this time for all populations. In D-dimensional space, the velocity and position of each particle i can be represented as Vi=(Vi1,Vi2,...,ViD) And Xi=(Xi1,Xi2,...,XiD) Similarly, the individual extremum can be expressed as pbesti=(pi1,pi2,...,piD) The global extremum is expressed as gbest ═ g1,g2,...,gD) So far, we can give the particle update formulas in two iterative processes as follows:
equation (1) is a d-dimension velocity update equation of particle i at k rounds, soIs the component of a certain particle velocity vector in a certain iteration, c1And c2Is an acceleration constant, also called acceleration particle, used to adjust the maximum step length of learning; r is1And r2Is two [0, 1 ]]A random function of (a) to increase search randomness; w is used to change the probed area of the solution space. Equation (2) is a d-th dimension position update equation of the particle i,expressing the d-dimension component of the position vector of the particle i in the k iteration, the formula (1) and the formula (2) form the necessary self-cognition and social experience in the particle swarm optimization. We can give a flow chart of the particle swarm optimization algorithm according to the principle of PSO and then combining equation (1) and equation (2) as shown in fig. 2 below.
The method comprises the steps of initializing particle swarm, setting the number, random speed and position of particles, solving the fitness of each particle according to a formula, comparing the current adaptive value of each particle with the adaptive value of the individual historical optimal position, and updating the historical optimal position by using the current position if the current adaptive value is higher, otherwise, keeping the historical optimal position unchanged. The global extremum may be compared to the global optimum based on the current fitness value for each particle, and the global optimum is updated with the current particle position if the current fitness value is higher. And then updating the speed and the position according to the formula (1) and the formula (2), and finally judging until the judgment condition is met, and finishing the algorithm.
And training the BP neural network by using training data, and adjusting the weight and the threshold of the network according to the error in the training process. And (4) classifying the wear state characteristic signals of the cutter by using the trained BP neural network, and analyzing the classifying capability of the BP neural network according to the classification result. The whole PSO-BP algorithm optimization flow chart is shown in the following figure 3.
And S5, classifying the characteristic signal test data of different tool wear states by using the trained PSO-BP neural network to obtain a BP neural network classification error map and BP neural network classification accuracy.
Acceleration signals in the cutting process of the vertical milling machine are acquired through an acceleration sensor arranged on a main shaft of the vertical milling machine and are represented as X ═ X (1), X (2), …, X (n), and n represents signal length, wherein the acceleration signals comprise four types including an initial wear state (1000), a light wear state (0100), a medium wear state (0010) and a heavy wear state (0001). The method for monitoring the wear state of the micro-milling cutter based on the manifold learning method is based on a neural network and the manifold learning method, and comprises the steps of firstly, extracting time domain characteristics of a cutting time signal; then, performing dimension reduction and reduction processing on the extracted time domain features, and processing the time domain features by using a manifold learning method to obtain 2 dimension-reduced time domain feature quantities with strong correlation; and finally, inputting the characteristic quantity subjected to the dimensionality reduction into a BP neural network for training, optimizing the neural network by a particle swarm algorithm, classifying signals generated by cutters with different wear degrees by the neural network according to the nonlinear mapping capability of the neural network, and accurately judging the category of the unknown signal. Compared with the conventional micro-milling cutter wear state monitoring method, the method has higher recognition rate of different wear states, can discover the severe wear of the cutter earlier, and prevents the irreversible damage of the workpiece and machine tool parts caused by the severe wear of the cutter.
Claims (6)
1. A method for monitoring the wear state of a micro-milling cutter based on a manifold learning method is characterized by comprising the following steps:
s1, acquiring an acceleration signal in the vertical milling and cutting process through an acceleration sensor mounted on a main shaft of the vertical milling machine, and expressing the acceleration signal as X ═ X (1), X (2), …, X (n), wherein n represents the signal length;
s2, performing time domain analysis on the signal in the cutting state, and extracting time domain signal characteristic quantity:
wherein n is the signal length; i is a scale parameter of signal continuity; x is a certain section of signal extreme value; the function max (| x |) is the maximum absolute value in a certain section of signal;
s3, performing feature space reduction on the extracted multiple time domain feature quantities to obtain a more concise time domain feature space with better correlation;
s4, training the neural network by using the reduced feature space for classification, determining the structure of the BP neural network to be 2-3-4 according to the characteristics of the cutter abrasion signal, and initializing the weight and the threshold of the BP neural network by using a particle swarm algorithm; next, training a BP neural network by using training data, and adjusting the weight and the threshold of the network according to errors in the training process; analyzing the BP neural network classification capability according to the classification result by using the trained BP neural network classification tool wear state characteristic signal;
and S5, classifying the characteristic signal test data of different tool wear states by using the trained PSO-BP neural network to obtain a BP neural network classification error map and BP neural network classification accuracy.
2. The method for monitoring the wear state of the micro-milling tool based on the manifold learning method as claimed in claim 1, wherein in step S3, the laplacian feature mapping in the manifold learning method is used to perform dimension reduction to obtain 2 time domain feature quantities with better correlation.
3. The method for monitoring the wear state of the micro-milling cutter based on the manifold learning method as claimed in claim 2, wherein the laplacian feature map measures the smoothness of the function on the manifold by using L aplarian-Beltrami operator in differential geometry, that is, the smoothness is a standard for measuring the performance of dimension reduction, a non-directional weighted neighbor graph is constructed to represent the manifold, and the graph is represented in a space with small dimension.
4. The method for monitoring the wear state of the micro-milling cutter based on the manifold learning method according to claim 3, wherein the dimension reduction step of the Laplace feature map is as follows:
1) constructing an undirected weighted graph: constructing a graph from all points using a method that sets an edge between two data points if the distance between the two data points meets the proximity threshold requirement, otherwise, no edge exists between the two points;
2) determining edge weights of the laplacian graph: the edge weight is calculated in a simplified mode, and when the node a and the node b have edges, w isabIs 1, otherwise wabIs 0, wherein wabIs the edge weight;
3) solving the characteristic mapping, namely firstly carrying out generalized eigenvalue decomposition on the Laplace undirected graph as follows, wherein L Y = lambda DY, Y is a matrix after dimensionality reduction, and D isab=∑bwabL is a Laplace matrix corresponding to the Laplace undirected graph, L = D-W, and the dimensionality reduced mapping result Y is the eigenvector V corresponding to the minimum m non-zero eigenvalues1,V2,…,VmForm, then original high-dimensional data tuple xiIn the low dimension the manifold can be represented as:
5. the method for monitoring the wear state of the micro-milling cutter based on the manifold learning method as claimed in claim 1, wherein the particle group algorithm step in the step S4 comprises: initializing a particle swarm; calculating the fitness of each particle; updating the individual extreme value and the global extreme value according to the fitness, and updating the particle speed and the particle position according to the fitness; and when the maximum iteration times are reached or the adaptive value corresponding to the global optimal particle is smaller than a specified threshold value, ending, otherwise, recalculating the fitness of each particle.
6. According to claim 5The method for monitoring the wear state of the micro-milling cutter based on the manifold learning method is characterized in that the particle swarm algorithm seeks an optimal solution according to two steps of initialization and iteration, two extreme values are required to be followed in the iteration to change the particle swarm algorithm, and the two extreme values are respectively called as an individual extreme value pbedti(i ═ 1, 2.. N) and a global extremum gbest; individual extremum pbesti(i ═ 1, 2.. N) is the optimal solution explored by a certain particle, and N is the number of particles; the global extreme value gbest is the optimal solution searched by all the populations at the moment; in D-dimensional space, the velocity and position of each particle i can be represented as Vi=(Vi1,Vi2,…,ViD) And Xi=(Xi1,Xi2,…,XiD) Similarly, the individual extremum can be expressed as pbesti=(pi1,pi2,…,piD) The global extremum is denoted as gbest = (g)1,g2,…,gD) So far, we can give the particle update formulas in two iterative processes as follows: is the d-dimension velocity update formula of the particle i in k rounds, soIs the component of a certain particle velocity vector in a certain iteration, c1And c2Is an acceleration constant, also called acceleration particle, used to adjust the maximum step length of learning; r is1And r2Is two [0, 1 ]]A random function of (a) to increase search randomness; w is used to change the probed area for the solution space;is the d-th dimension position update formula of the particle i,representing the d-dimension component of the location vector of the particle i at the k-th iteration.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010406461.4A CN111421386A (en) | 2020-05-14 | 2020-05-14 | Method for monitoring wear state of micro-milling cutter based on manifold learning method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010406461.4A CN111421386A (en) | 2020-05-14 | 2020-05-14 | Method for monitoring wear state of micro-milling cutter based on manifold learning method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111421386A true CN111421386A (en) | 2020-07-17 |
Family
ID=71552828
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010406461.4A Pending CN111421386A (en) | 2020-05-14 | 2020-05-14 | Method for monitoring wear state of micro-milling cutter based on manifold learning method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111421386A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111881860A (en) * | 2020-07-31 | 2020-11-03 | 重庆理工大学 | Modeling method of hob abrasion in-situ recognition model and hob abrasion in-situ recognition method |
CN111993158A (en) * | 2020-08-28 | 2020-11-27 | 深圳市玄羽科技有限公司 | Method for generating tool detection model, method, system, device and medium for detecting tool detection model |
CN112818793A (en) * | 2021-01-25 | 2021-05-18 | 常州大学 | Micro-milling cutter wear state monitoring method based on local linear embedding method |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3076548B2 (en) * | 1998-03-20 | 2000-08-14 | 川崎重工業株式会社 | Tool wear detecting device and method |
CN102689230A (en) * | 2012-05-09 | 2012-09-26 | 天津大学 | Tool wear condition monitoring method based on conditional random field model |
CN105740893A (en) * | 2016-01-27 | 2016-07-06 | 大连楼兰科技股份有限公司 | Vehicle characteristic extraction and identification algorithm based on manifold |
CN105834834A (en) * | 2016-04-21 | 2016-08-10 | 四川大学 | Milling cutter abrasion state monitoring method based on fruit fly optimization algorithm |
CN108747590A (en) * | 2018-06-28 | 2018-11-06 | 哈尔滨理工大学 | A kind of tool wear measurement method based on rumble spectrum and neural network |
CN109033730A (en) * | 2018-09-30 | 2018-12-18 | 北京工业大学 | A kind of tool wear prediction technique based on improved particle swarm optimization algorithm |
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 |
-
2020
- 2020-05-14 CN CN202010406461.4A patent/CN111421386A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3076548B2 (en) * | 1998-03-20 | 2000-08-14 | 川崎重工業株式会社 | Tool wear detecting device and method |
CN102689230A (en) * | 2012-05-09 | 2012-09-26 | 天津大学 | Tool wear condition monitoring method based on conditional random field model |
CN105740893A (en) * | 2016-01-27 | 2016-07-06 | 大连楼兰科技股份有限公司 | Vehicle characteristic extraction and identification algorithm based on manifold |
CN105834834A (en) * | 2016-04-21 | 2016-08-10 | 四川大学 | Milling cutter abrasion state monitoring method based on fruit fly optimization algorithm |
CN108747590A (en) * | 2018-06-28 | 2018-11-06 | 哈尔滨理工大学 | A kind of tool wear measurement method based on rumble spectrum and neural network |
CN109033730A (en) * | 2018-09-30 | 2018-12-18 | 北京工业大学 | A kind of tool wear prediction technique based on improved particle swarm optimization algorithm |
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 |
Non-Patent Citations (1)
Title |
---|
姚舜才 孙传猛: "《机器学习基础教程》", 31 March 2020, 西安电子科技大学出版社 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111881860A (en) * | 2020-07-31 | 2020-11-03 | 重庆理工大学 | Modeling method of hob abrasion in-situ recognition model and hob abrasion in-situ recognition method |
CN111881860B (en) * | 2020-07-31 | 2022-05-03 | 重庆理工大学 | Modeling method of hob abrasion in-situ recognition model and hob abrasion in-situ recognition method |
CN111993158A (en) * | 2020-08-28 | 2020-11-27 | 深圳市玄羽科技有限公司 | Method for generating tool detection model, method, system, device and medium for detecting tool detection model |
CN112818793A (en) * | 2021-01-25 | 2021-05-18 | 常州大学 | Micro-milling cutter wear state monitoring method based on local linear embedding method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111421386A (en) | Method for monitoring wear state of micro-milling cutter based on manifold learning method | |
CN110647943B (en) | Cutting tool wear monitoring method based on evolution data cluster analysis | |
CN110153802B (en) | Tool wear state identification method based on convolution neural network and long-term and short-term memory neural network combined model | |
CN109571141A (en) | A kind of Monitoring Tool Wear States in Turning based on machine learning | |
CN111680542A (en) | Steel coil point cloud identification and classification method based on multi-scale feature extraction and Pointernet neural network | |
CN105834834B (en) | Cutter wear state monitoring method based on drosophila optimized algorithm | |
Shi et al. | A reinforced k-nearest neighbors method with application to chatter identification in high-speed milling | |
CN111444769B (en) | Laser radar human leg detection method based on multi-scale self-adaptive random forest | |
CN112347894A (en) | Single-plant vegetation extraction method based on transfer learning and Gaussian mixture model separation | |
CN111730412A (en) | Ant colony optimization algorithm-based micro milling cutter wear state monitoring method of support vector machine | |
CN111160405B (en) | Engine lubricating oil abrasive particle identification method based on deep learning | |
CN110503148B (en) | Point cloud object identification method with scale invariance | |
CN110427019B (en) | Industrial process fault classification method and control device based on multivariate discriminant analysis | |
Chaowen et al. | Research on tool wear monitoring based on GRU-CNN | |
CN116721144A (en) | Cone hole size measurement method based on point cloud slicing | |
CN113344039B (en) | Multi-extension target tracking method based on space-time correlation | |
CN108537249B (en) | Industrial process data clustering method for density peak clustering | |
CN116933196B (en) | Method and system for intelligently eliminating abnormal values of multidimensional logging data | |
CN114639238A (en) | Urban expressway traffic state estimation method based on normalized spectral clustering algorithm | |
CN116362005A (en) | Reuse design method for similar instance processing technology of injection mold part based on instance reasoning | |
CN106203469A (en) | A kind of figure sorting technique based on orderly pattern | |
CN113850304B (en) | High-accuracy point cloud data classification segmentation improvement method | |
Li et al. | Tool wear monitoring using machine learning | |
CN112818793A (en) | Micro-milling cutter wear state monitoring method based on local linear embedding method | |
CN112884027B (en) | Cutting process real-time state monitoring method and device based on pattern recognition |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200717 |
|
RJ01 | Rejection of invention patent application after publication |