CN109434562A - Milling cutter state of wear recognition methods based on partition clustering - Google Patents
Milling cutter state of wear recognition methods based on partition clustering Download PDFInfo
- Publication number
- CN109434562A CN109434562A CN201811163321.8A CN201811163321A CN109434562A CN 109434562 A CN109434562 A CN 109434562A CN 201811163321 A CN201811163321 A CN 201811163321A CN 109434562 A CN109434562 A CN 109434562A
- Authority
- CN
- China
- Prior art keywords
- wear
- state
- sample
- cutter
- cutting
- 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
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
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Length Measuring Devices With Unspecified Measuring Means (AREA)
Abstract
The invention discloses the milling cutter state of wear recognition methods based on partition clustering, steps are as follows: the three-dimensional Cutting Force Signal sample under acquisition cutting-tool wear state not of the same race；12 temporal signatures and 7 frequency domain characters are extracted in each direction, put into eigenmatrix X；LPP method is selected to obtain matrix Y as the dimension reduction method of high dimensional feature data set X；It is clustered using the clustering algorithm based on segmentation central point (PAM), exports cluster result.The cluster result of PAM output is assessed, as found from the results, PAM ratio k-means and FCM is more acurrate, improve the accuracy rate of the online recognition of tool wear in process, especially there is strong adaptability there are complex nonlinear signal to difficult-to-machine material, Cutter wear identification is of great significance, and is also of great significance in terms of improving machined surface quality.
Description
Technical field
The invention belongs to change system malfunction monitorings, diagnostic field, and in particular to one kind is analyzed based on partition clustering (PAM)
Tool condition monitoring and identification technique field.
Background technique
Milling cutter abrasion can seriously affect processing performance, it is therefore necessary to Cutter wear state monitored on-line with
Improve service life and the surface quality of workpiece.In tool condition monitoring (TCM) system, signal acquisition method can be divided into direct method and
Indirect method.Direct method has the advantages that high-precision, but is difficult to monitor cutting-tool wear state on-line.Therefore, nearest researcher collection
In in research indirect method.The supervised learnings method such as neural network (NN) and support vector machines (SVM) is in Tool Wear Monitoring
It is very effective.However supervised learning needs to be cost raising diagnostic accuracy by sample training before actual monitoring, and nothing
Supervised learning does not need then.Therefore in order to find new or unknown failure during diagnosis, select unsupervised learning come into
Row cutting-tool wear state on-line monitoring.Cluster is a kind of typical unsupervised learning method and obtains in fault diagnosis field
It is widely applied.But traditional clustering method such as k-means and FCM are due to very sensitive to noise spot, so cluster is accurate
It spends lower, and the stability of system may be made to change.
Summary of the invention
It is an object of the invention to overcome the defect of prior art, a kind of high-precision for realizing cutting-tool wear state is provided
The milling cutter state of wear recognition methods based on partition clustering of identification.
The invention proposes the milling cutter state of wear recognition methods based on partition clustering, this method includes following step
It is rapid:
Step 1: installing force snesor on workpiece, pass through state of wear classes different in force sensor measuring cutting process
Respectively in three-dimensional space x, y, the Cutting Force Signal in the direction z between cutter and workpiece under not；
Step 2: acquiring the three-dimensional Cutting Force Signal between the cutter and workpiece under different state of wear classifications, then will
Cutting Force Signal data under each state of wear form multiple samples, each abrasion shape according to the sampling period even partition of setting
Two swing circles of cutter are taken to form a sample as a sampling period and by the signal in two swing circles of cutter under state
This, then every kind of cutting-tool wear state has the sample set formed by Num sample, and total sample number T=Num*C, C are tool wears
Classification number；Finally according to the degree of wear from the sample set carry out sequence row gently to the sequence of weight to cutter under different state of wear
Column, and to sample set according to successively marked from first trip to the sequence of tail row ranks value be 1,2 ... C, procession value mark
Cutting-tool wear state classification corresponding with the ranks value is labeled simultaneously, so that building forms cluster labels collection R=T × 1,
Wherein line number with to answer total sample number be mutually all T, columns is 1 column；
Step 3: all samples are carried out with temporal signatures respectively and frequency domain character extracts to obtain characteristic B, and pass through public affairs
Feature sum L is calculated in formula L=B*C；Then using total sample number T as the line number of higher-dimension initial characteristic data collection, using L as
The columns of higher-dimension initial characteristic data collection obtains higher-dimension initial characteristic data collection matrix T × L；
Step 4: using LPP method, by extracting Z preferred feature of most identification to higher-dimension primitive character number
According to collection matrix dimensionality reduction, sample characteristics matrix is T × Z at this time；
Step 5: inputting the sample characteristics matrix after dimensionality reduction in a computer, imports PAM algorithm and clustered, obtain PAM
The input sample classification of algorithm output.
Compared with prior art, the invention proposes one kind based on segmentation central point (Partitioning Around
Medoids, PAM) new method identify milling cutter state of wear.LPP dimension-reduction treatment is carried out to high dimensional feature data set, is carried out
Five kinds of Cluster Evaluation method objective evaluation clustering performances are used after clustering processing.As found from the results, PAM ratio k-means and FCM
It is more acurrate, the accuracy rate of the online recognition of tool wear in process is improved, is especially existed to difficult-to-machine material complicated
Nonlinear properties have strong adaptability, and Cutter wear identification is of great significance, in terms of improving machined surface quality
Also it is of great significance.
Detailed description of the invention
Fig. 1 is system construction drawing of the invention；
Fig. 2-1 is the x direction signal sample in second step of the embodiment of the present invention under collected new knife-like state；
Fig. 2-2 is the y direction signal sample in second step of the embodiment of the present invention under collected new knife-like state；
Fig. 2-3 is the x direction signal sample in second step of the embodiment of the present invention under collected mild wear state；
Fig. 2-4 is the y direction signal sample in second step of the embodiment of the present invention under collected mild wear state；
Fig. 2-5 is the x direction signal sample in second step of the embodiment of the present invention under collected moderate state of wear；
Fig. 2-6 is the y direction signal sample in second step of the embodiment of the present invention under collected moderate state of wear；
Fig. 2-7 is the x direction signal sample in second step of the embodiment of the present invention under collected severe state of wear；
Fig. 2-8 is the y direction signal sample in second step of the embodiment of the present invention under collected severe state of wear.
Specific embodiment
PAM is a kind of clustering algorithm based on division methods, it overcomes in traditional clustering method for the quick of exceptional value
The defects, more robustness such as cause clustering precision low that sensitivity is excessively high, are of great significance for tool condition monitoring.
Specific embodiments of the present invention are illustrated below in conjunction with attached drawing.
The present invention is based on the milling cutter state of wear recognition methods of partition clustering as shown in Figure 1, comprising the following steps:
Step 1: installing force snesor on workpiece, pass through state of wear classes different in force sensor measuring cutting process
Respectively in three-dimensional space x, y, the Cutting Force Signal in the direction z between cutter and workpiece under not.
As one embodiment of the present invention, cutting-tool wear state classification include new knife, mild wear, moderate abrasion and
Severe wears four classifications.
The force snesor can be one of piezoelectric force transducer or strain force sensor.
Step 2: acquiring under different state of wear classifications (new knife, mild wear, moderate abrasion and severe wear four kinds)
Three-dimensional Cutting Force Signal between cutter and workpiece, then by Cutting Force Signal data the adopting according to setting under each state of wear
Sample period even partition forms multiple samples, taken under each state of wear two swing circles of cutter as a sampling period simultaneously
Signal in two swing circles of cutter is formed into a sample, then every kind of cutting-tool wear state has is formed by Num sample
Sample set, total sample number T=Num*C, C are tool wear classification numbers, if cutting-tool wear state classification has 4, C=4；Most
Afterwards according to the degree of wear from the sample set carry out sequence arrangement gently to the sequence of weight to cutter under different state of wear, and to sample
This collection according to successively marked from first trip to the sequence of tail row ranks value be 1,2 ... C, procession value mark while pair with
The corresponding cutting-tool wear state classification of the ranks value is labeled, and forms cluster labels collection R=T × 1 to construct, wherein line number
It is mutually all T with total sample number, columns is 1 column；
Step 3: all samples are carried out with temporal signatures respectively and frequency domain character extracts to obtain characteristic B, and pass through public affairs
Feature sum L is calculated in formula L=B*C；Then using total sample number T as the line number of higher-dimension initial characteristic data collection, using L as
The columns of higher-dimension initial characteristic data collection obtains higher-dimension initial characteristic data collection matrix T × L；
As one embodiment of the present invention, wherein temporal signatures include: mean value, root-mean-square value, peak value, variance, peak
Peak value, signal energy, crest factor, kurtosis, pulse index, the degree of bias and nargin；Frequency domain character includes: power spectrum and power spectrum
Mean value, power spectrum variance, the power spectrum degree of bias, power spectrum kurtosis and spectrum peak.
If C=4, L=57, the huge feature set for 57 features being made of three-dimensional cutting force is obtained.
Step 4: the dimension reduction method using LPP method as higher-dimension initial characteristic data collection matrix, LPP is by extracting most
Feature with identification carries out dimensionality reduction, retains local message, reduces the factors for influencing clustering recognition.It is obtained by dimensionality reduction
Obtain optimal Z preferred feature.Sample characteristics matrix is T × Z at this time；
LPP algorithm is referring specifically to periodical " Advances in Neural Information Processing
Systems " page 186 to page 197 of volume 16 upper article " Locality published in (development of neural information processing systems)
Preserving projections " (partial projection).
Step 5: inputting the sample characteristics matrix after dimensionality reduction in a computer, PAM algorithm is imported (referring specifically to periodical " meter
Calculation machine and modernization " volume 9 page 1 to page 3 publication article " analysis and realization of PAM Algorithm ") clustered,
Obtain the input sample classification of PAM algorithm output.
Embodiment 1
Step 1: piezoelectric force transducer is mounted on workpiece, experiment workpiece selection titanium (Ti-6Al-4V), having a size of
150 × 100 × 30mm, speed of mainshaft 30m/min, feed engagement 0.1mm.It was cut by piezoelectric force transducer measurement
In three direction x of three-dimensional space, the Cutting Force Signal of y, z between cutter and workpiece in journey under different conditions.Acquire cutter letter
Number include new knife, mild wear, moderate abrasion and severe wear four kinds of states standard force signal.Tool wear value range difference
Correspond to 0-0.06mm, 0.07mm-0.14mm, 0.26mm-0.36mm and 0.51mm-0.61mm；
Step 2: acquiring the three-dimensional cutting that new knife, mild wear, moderate abrasion and severe wear four kinds of cutting-tool wear states
Force signal.C=4 is tool wear classification number.After obtaining each Cutting Force Signal, the signal conduct in two swing circles of cutter is taken
One sample, to the original signal data even partition of acquisition.The sample signal such as Fig. 2-1 and Fig. 2-in the new direction knife x and the direction y
Shown in 2；The sample signal in the new direction knife x and the direction y is as shown in Fig. 2-1 and Fig. 2-2；The sample in the direction mild wear x and the direction y
Signal is as shown in Fig. 2-3 and Fig. 2-4；The sample signal in the moderate abrasion direction x and the direction y is as shown in Fig. 2-5 and Fig. 2-6；Severe
The sample signal in the abrasion direction x and the direction y is as shown in Fig. 2-7 and Fig. 2-8；Every kind of cutting-tool wear state has 50 samples in this example
This, total sample number T=50 × 4=200.Cluster labels collection R=200 × 1 is constructed, it is 200 that wherein line number, which corresponds to total sample number, column
Number is 1 column, and train value is the corresponding mark value of sample, is demarcated to all sample generics.1st row is to the 50th ranks value
1, belong to new knife；51st row to the 100th ranks value is 2, belongs to mild wear；101st row to the 150th ranks value is 3, is belonged to
Degree abrasion；151st row to the 200th ranks value is 4, belongs to severe abrasion.
Step 3: carrying out feature extraction to all samples.Extract following 12 temporal signatures respectively: mean value, root-mean-square value,
Peak value, variance, peak-to-peak value, signal energy, crest factor, kurtosis, pulse index, the degree of bias, nargin.And following 7 are extracted respectively
A frequency domain character: power spectrum and power spectrum mean value, power spectrum variance, the power spectrum degree of bias, power spectrum kurtosis, spectrum peak.
To the above-mentioned time domain of each sample extraction and frequency domain character, the height for 57 features being made of three-dimensional cutting force is obtained
Dimensional feature collection.Sample of the present invention sum T=200 is the line number of above-mentioned primitive character collection；L is characteristic species number, is above-mentioned original spy
The columns of collection, here L=57.Sample characteristics matrix is 200 × 57 at this time；
4th step selects LPP method as the dimension reduction method of high dimensional feature data set, and LPP is most differentiated by extracting
Property feature carry out dimensionality reduction, retain local message, reduce the factors for influencing clustering recognition.Optimal 17 are obtained in this example
A preferred feature.Sample characteristics matrix is 200 × 17 at this time.In order to verify the validity of LPP, compare the spy with LPP dimensionality reduction
The clustering precision of collection and the feature set without LPP dimensionality reduction.
5th step inputs 200 × 17 sample characteristics matrixes, imports PAM algorithm and is clustered, and obtains the output of PAM algorithm
Input sample classification.
For objective measure clustering performance, five kinds of external Cluster Evaluation methods are proposed.Be respectively: NMI, CSM, ERR,
CluCE and ClaCE.Cluster result NMI and CSM is higher, and clustering result quality is better.Cluster result ERR and CluCE and ClaCE
Lower, clustering result quality is better.It is illustrated separately below:
(1) cluster result NMI is higher, and clustering result quality is better.NMI value reaches 1, this shows that actual result is complete with cluster result
U.S. matching.On the contrary, NMI value is 0 if data are random distributions.Algorithm is referring specifically to periodical " Journal of
Machine Learning Research " in page 583 to page 617 of volume 3, the entitled " Cluster ensembles-a of paper
A kind of knowledge reuse framework for combining multiple partitions " (use of Cluster-Fusion-
In the knowledge reorganization frame that combination divides), author: Strehl A and Ghosh J.
(2) ERR is by a kind of matching process by each cluster label mapping to a class label.In range (0,1), ERR
Lower, Clustering Effect is better.Algorithm is referring specifically in periodical " Advances in Neural Information
Processing System " page 1529 to page 1536 upper article " the A Local Learning Approach for published
Clustering " (a kind of local learning method for cluster), author: BJ Platt and T Hofmann.
(3) CSM value increases with the increase of cluster label and class tag match logarithm.Algorithm is referring specifically to periodical
Paper " Time-series clustering-the A of page 16 to page 38 publications of volume 53 in " Information Systems "
Decade review " (Time Series Clustering summary), author: S Aghabozorgi, AS Shirkhorshidi and TY Wah.
(4) range of CluCE and ClaCE is all from 0 to 1, but numerical value is lower, uncertain smaller.When in each cluster
Class when becoming more uniform, cross entropy is reduced.Difference between the two indexs is CluCE to point of the cluster in each class
Cloth is insensitive, and vice versa by ClaCE.CluCE algorithm and ClaCE algorithm are referring specifically to the " the in relation to data mining technology the 8th
Article " the Comparison of Cluster Representations from employed on secondary ieee international conference "
Partial Second-to Full Fourth-Order Cross Moments for Data Stream
Clustering ", author: Song MJ and Zhang L.
Correct cluster labels collection is compared with the cluster labels that PAM algorithm obtains by this five kinds of algorithms, to PAM
Clustering precision judged.Compare the clustering precision of the feature set with LPP dimensionality reduction and the feature set without LPP dimensionality reduction, ties
Fruit is as shown in table 1.Simultaneously in order to verify the advantage that PAM is identified in cutting-tool wear state, by PAM and other two kinds of typical cluster sides
Method FCM and Kmeans are compared.The results are shown in Table 2.
Table 1 is compared using LPP algorithm and without using the feature set PAM clustering precision of LPP algorithm
2 PAM, k-means and FCM clustering precision of table compares
From table 1 it follows that higher NMI and CSM can be obtained by carrying out dimensionality reduction to huge feature set using LPP algorithm
And lower ERR, CluCE and ClaCE, it can obtain higher clustering precision.
From Table 2, it can be seen that PAM can obtain higher NMI and CSM and lower ERR, CluCE and ClaCE,
Show that the clustering precision of PAM is apparently higher than k-means and FCM.
Claims (4)
1. the milling cutter state of wear recognition methods based on partition clustering, it is characterised in that the following steps are included:
Step 1: force snesor is installed on workpiece, by under state of wear classifications different in force sensor measuring cutting process
Cutter and workpiece between respectively in three-dimensional space x, y, the Cutting Force Signal in the direction z；
Step 2: the three-dimensional Cutting Force Signal between the cutter and workpiece under different state of wear classifications is acquired, then by each mill
Cutting Force Signal data under damage state form multiple samples according to the sampling period even partition of setting, under each state of wear
Two swing circles of cutter are taken to form a sample as a sampling period and by the signal in two swing circles of cutter, then
Every kind of cutting-tool wear state has the sample set formed by Num sample, and total sample number T=Num*C, C are tool wear classifications
Number；Finally according to the degree of wear from gently to sample set carry out sequence arrangement of the sequence to cutter under different state of wear of weight,
And to sample set according to successively marked from first trip to the sequence of tail row ranks value be 1,2 ... C, procession value mark it is same
When a pair cutting-tool wear state classification corresponding with the ranks value be labeled, thus building form cluster labels collection R=T × 1,
Middle line number with to answer total sample number be mutually all T, columns is 1 column；
Step 3: all samples are carried out with temporal signatures respectively and frequency domain character extracts to obtain characteristic B, and pass through formula L
Feature sum L is calculated in=B*C；Then using total sample number T as the line number of higher-dimension initial characteristic data collection, using L as height
The columns of Wei Yuanshitezhengshuojuji obtains higher-dimension initial characteristic data collection matrix T × L；
Step 4: using LPP method, by extracting Z preferred feature of most identification to higher-dimension initial characteristic data collection
Matrix dimensionality reduction, sample characteristics matrix is T × Z at this time；
Step 5: inputting the sample characteristics matrix after dimensionality reduction in a computer, imports PAM algorithm and clustered, obtain PAM algorithm
The input sample classification of output.
2. the milling cutter state of wear recognition methods according to claim 1 based on partition clustering, it is characterised in that: institute
The cutting-tool wear state classification stated includes that new knife, mild wear, moderate abrasion and severe wear four classifications.
3. the milling cutter state of wear recognition methods according to claim 1 or 2 based on partition clustering, feature exist
In: the force snesor is one of piezoelectric force transducer or strain force sensor.
4. the milling cutter state of wear recognition methods according to claim 3 based on partition clustering, it is characterised in that: institute
The temporal signatures stated include: that mean value, root-mean-square value, peak value, variance, peak-to-peak value, signal energy, crest factor, kurtosis, pulse refer to
Mark, the degree of bias and nargin；Frequency domain character includes: that power spectrum and power spectrum mean value, power spectrum variance, the power spectrum degree of bias, power spectrum are high and steep
Degree and spectrum peak.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811163321.8A CN109434562A (en) | 2018-09-30 | 2018-09-30 | Milling cutter state of wear recognition methods based on partition clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811163321.8A CN109434562A (en) | 2018-09-30 | 2018-09-30 | Milling cutter state of wear recognition methods based on partition clustering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109434562A true CN109434562A (en) | 2019-03-08 |
Family
ID=65544648
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811163321.8A Pending CN109434562A (en) | 2018-09-30 | 2018-09-30 | Milling cutter state of wear recognition methods based on partition clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109434562A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110576335A (en) * | 2019-09-09 | 2019-12-17 | 北京航空航天大学 | cutting force-based tool wear online monitoring method |
| CN110647943A (en) * | 2019-09-26 | 2020-01-03 | 西北工业大学 | Cutting tool wear monitoring method based on evolutionary data cluster analysis |
| CN112705766A (en) * | 2020-12-18 | 2021-04-27 | 成都航空职业技术学院 | Method for monitoring non-uniform wear state of cutter |
| CN114453630A (en) * | 2022-01-20 | 2022-05-10 | 湖北文理学院 | Method and device for controlling machine tool to mill non-stick tool, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103786069A (en) * | 2014-01-24 | 2014-05-14 | 华中科技大学 | Flutter online monitoring method for machining equipment |
| CN106271881A (en) * | 2016-08-04 | 2017-01-04 | 华中科技大学 | A kind of Condition Monitoring of Tool Breakage method based on SAEs and K means |
| CN107194427A (en) * | 2017-05-26 | 2017-09-22 | 温州大学 | A kind of milling cutter malfunction monitoring and recognition methods and system |
| CN108227633A (en) * | 2016-12-13 | 2018-06-29 | 发那科株式会社 | Numerical control device and machine learning device |
-
2018
- 2018-09-30 CN CN201811163321.8A patent/CN109434562A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103786069A (en) * | 2014-01-24 | 2014-05-14 | 华中科技大学 | Flutter online monitoring method for machining equipment |
| CN106271881A (en) * | 2016-08-04 | 2017-01-04 | 华中科技大学 | A kind of Condition Monitoring of Tool Breakage method based on SAEs and K means |
| CN108227633A (en) * | 2016-12-13 | 2018-06-29 | 发那科株式会社 | Numerical control device and machine learning device |
| CN107194427A (en) * | 2017-05-26 | 2017-09-22 | 温州大学 | A kind of milling cutter malfunction monitoring and recognition methods and system |
Non-Patent Citations (1)
| Title |
|---|
| ZHIMENG LI；GUOFENG WANG；GAIYUN HE: "Milling tool wear state recognition based on partitioning around medoids (PAM) clusteringMilling tool wear state recognition based on partitioning around medoids (PAM) clustering", 《THE INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110576335A (en) * | 2019-09-09 | 2019-12-17 | 北京航空航天大学 | cutting force-based tool wear online monitoring method |
| CN110576335B (en) * | 2019-09-09 | 2020-11-20 | 北京航空航天大学 | Cutting force-based tool wear online monitoring method |
| CN110647943A (en) * | 2019-09-26 | 2020-01-03 | 西北工业大学 | Cutting tool wear monitoring method based on evolutionary data cluster analysis |
| CN112705766A (en) * | 2020-12-18 | 2021-04-27 | 成都航空职业技术学院 | Method for monitoring non-uniform wear state of cutter |
| CN114453630A (en) * | 2022-01-20 | 2022-05-10 | 湖北文理学院 | Method and device for controlling machine tool to mill non-stick tool, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109514349B (en) | Tool wear state monitoring method based on vibration signal and Stacking integrated model | |
| Li et al. | Similarity-measured isolation forest: Anomaly detection method for machine monitoring data | |
| CN110132598B (en) | Fault noise diagnosis algorithm for rolling bearing of rotating equipment | |
| CN107584334B (en) | A kind of complex structural member numerical control machining cutter status real time monitor method based on deep learning | |
| CN109434562A (en) | Milling cutter state of wear recognition methods based on partition clustering | |
| Aydadenta et al. | A clustering approach for feature selection in microarray data classification using random forest | |
| CN112036301B (en) | Driving motor fault diagnosis model construction method based on intra-class feature transfer learning and multi-source information fusion | |
| CN110647943B (en) | Cutting tool wear monitoring method based on evolution data cluster analysis | |
| CN109158953B (en) | A kind of cutting-tool wear state on-line monitoring method and system | |
| Li et al. | A rolling bearing fault diagnosis strategy based on improved multiscale permutation entropy and least squares SVM | |
| CN112508105B (en) | Fault detection and retrieval method for oil extraction machine | |
| CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
| CN116070527B (en) | Milling cutter residual life prediction method based on degradation model | |
| Chen et al. | Data quality evaluation and improvement for prognostic modeling using visual assessment based data partitioning method | |
| CN112257767B (en) | Product key part state classification method aiming at class imbalance data | |
| CN104596780A (en) | Diagnosis method for sensor faults of motor train unit braking system | |
| CN114346761B (en) | Cutter abrasion condition detection method based on improved condition generation countermeasure network | |
| CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
| Du et al. | Intelligent turning tool monitoring with neural network adaptive learning | |
| CN114970643A (en) | High-speed electric spindle fault identification method based on UMAP dimension reduction algorithm | |
| CN114462475A (en) | Unsupervised machine abnormal sound detection method and unsupervised machine abnormal sound detection device based on single classification algorithm | |
| CN114714145A (en) | Method for enhancing, comparing, learning and monitoring tool wear state by using Gelam angular field | |
| Wang et al. | Self-adaptive fusion of local-temporal features for tool condition monitoring: A human experience free model | |
| CN113369993B (en) | Method for monitoring tool wear state under small sample | |
| Peng et al. | A multi-indicator fusion-based approach for fault feature selection and classification of rolling bearings |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190308 |