CN110682159A - Cutter wear state identification method and device - Google Patents
Cutter wear state identification method and device Download PDFInfo
- Publication number
- CN110682159A CN110682159A CN201910912879.XA CN201910912879A CN110682159A CN 110682159 A CN110682159 A CN 110682159A CN 201910912879 A CN201910912879 A CN 201910912879A CN 110682159 A CN110682159 A CN 110682159A
- Authority
- CN
- China
- Prior art keywords
- wear state
- svm
- acoustic emission
- cutter
- tool
- 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
- 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
- B23Q17/098—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 by measuring noise
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
The invention provides a cutter wear state identification method, which comprises the following steps: collecting acoustic emission signals in the working process of the cutter; extracting a frequency band energy characteristic value in the acoustic emission signal according to the harmonic wavelet packet; extracting training samples from the characteristic values; inputting the training samples into an LS-SVM model, and training to obtain an LS-SVM cutter wear state recognition model; and identifying the wear state of the cutter according to the LS-SVM cutter wear state identification model. According to the method, the frequency band energy characteristic value in the acoustic emission signal is extracted through the harmonic wave packet, for the complex non-stationary signal generated in the tool abrasion process, the harmonic wave packet can decompose the signal to different frequency bands without overlapping and omission, so that the characteristic information of the signal on different frequency bands is obtained, the tool abrasion characteristic signal can be better analyzed, the abrasion pattern recognition can be further performed, and the accuracy of the pattern recognition is improved.
Description
Technical Field
The invention relates to the technical field of cutter wear identification, in particular to a cutter wear state identification method and device.
Background
For modern machine tools, 20% of the downtime is due to tool failure, with 6.8% downtime due to tool breakage, directly resulting in low productivity and economic loss. If the abrasion state of the cutter can be monitored and identified on line in the cutter machining process, on one hand, the online error compensation can be carried out, the quality problem caused by cutter abrasion is greatly reduced, the machining precision is ensured, and the operation efficiency of the device is comprehensively improved; and secondly, personnel in a machining center can be reminded to change the tool in time before the tool is dull, so that the continuity and effectiveness of machining are guaranteed.
In the prior art, wear state related information is obtained by measuring parameters (such as cutting force, vibration signals, sound wave signals and the like) related to tool wear, so that tool wear amount or wear state is calculated. The method mainly extracts the characteristics in the parameters related to the cutter wear through a wavelet packet technology, and because the traditional wavelet basis functions have overlapping phenomena on frequency bands, energy overlapping exists under different frequency band signals, the signal characteristics of the cutter wear are difficult to accurately reflect, accurate analysis on the signals is influenced, and the identification accuracy of the traditional cutter wear detection method is low.
Disclosure of Invention
The invention solves the problems that: the traditional cutter wear detection method has low recognition accuracy rate for recognizing the wear state by extracting the characteristics in the relevant parameters of cutter wear through a wavelet packet technology.
In order to solve the above-mentioned problems,
in one aspect, the present invention provides a tool wear state identification method, including:
collecting acoustic emission signals in the working process of the cutter;
extracting a frequency band energy characteristic value in the acoustic emission signal according to the harmonic wavelet packet;
extracting training samples from the characteristic values;
inputting the training samples into an LS-SVM model, and training to obtain an LS-SVM cutter wear state recognition model;
and identifying the wear state of the cutter according to the LS-SVM cutter wear state identification model.
Optionally, after the acoustic emission signal in the working process of the tool is collected, before the frequency band energy characteristic value in the acoustic emission signal is extracted according to the harmonic wavelet packet, the method further includes:
and denoising the acoustic emission signal.
Optionally, the noise reduction of the acoustic emission signal includes:
acquiring a power spectrum of the acoustic emission signal;
smoothing the power spectrum of each frequency point of the acoustic emission signal;
carrying out nonlinear tracking on the minimum power value of each frequency point of the acoustic emission signal;
and denoising the acoustic emission signal by using a wiener filter.
Optionally, after the frequency band energy feature value in the acoustic emission signal is extracted according to the harmonic wavelet packet, before the training sample is extracted from the feature value, the method further includes:
and carrying out normalization processing on the characteristic values.
Optionally, inputting the training sample into an LS-SVM model, and training to obtain an LS-SVM tool wear state recognition model, including:
initializing a BSA (bovine serum albumin) population, setting training parameters and defining a fitness function;
performing LS-SVM training to obtain a fitness value, and recording an optimal individual and the optimal fitness value;
performing iterative optimization according to a BSA algorithm, and updating an optimal individual and an optimal fitness value;
if the end condition is met, outputting an optimized punishment factor and a kernel parameter square of the LS-SVM;
and establishing a wear state identification model of the LS-SVM cutter according to the penalty factor and the square of the nuclear parameter.
Optionally, after the training sample is input into the LS-SVM model and the LS-SVM tool wear state recognition model is obtained through training, before the wear state of the tool is recognized according to the LS-SVM tool wear state recognition model, the method further includes:
and extracting a test sample from the characteristic value, and inputting the test sample into the LS-SVM cutter wear state identification model for testing.
Optionally, after identifying the wear state of the tool according to the LS-SVM tool wear state identification model, the method further includes:
and performing function fitting on the continuous wear state values to obtain a derivative of the fitting function in the current wear state value, and discarding the current wear state value if the derivative is less than zero.
Compared with the prior art, the tool wear state identification method has the following advantages:
(1) the cutter wear state identification method can establish an LS-SVM cutter wear state identification model by extracting the energy characteristic value of an acoustic emission signal generated in the working process of the cutter, and further identify the cutter wear state according to the LS-SVM cutter wear state identification model; the frequency band energy characteristic value in the acoustic emission signal is extracted through the harmonic wavelet packet, for complex non-stationary signals generated in the tool abrasion process, the harmonic wavelet packet can decompose the signals to different frequency bands without overlapping and omission, and characteristic information of the signals on different frequency bands is obtained, so that the tool abrasion characteristic signal can be better analyzed, the abrasion pattern recognition can be further performed, and the accuracy of the pattern recognition is improved;
(2) the cutter wear state identification method disclosed by the invention automatically selects the punishment factors and the nuclear parameters of the LS-SVM by using the BSA (bovine serum albumin) optimization algorithm so as to reach the optimal combination value, and the accuracy of model wear identification is greatly improved;
(3) the tool wear state identification method corrects the identification result by calculating the change trend of the wear state value, and further improves the identification accuracy of the tool wear state.
On the other hand, the invention also provides a tool wear state identification device to solve the problem that the traditional tool wear detection method is low in identification accuracy rate by extracting the characteristics in the tool wear related parameters through a wavelet packet technology so as to identify the wear state.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
a tool wear state recognition device comprising:
the signal acquisition module is used for acquiring acoustic emission signals in the working process of the cutter;
the characteristic extraction module is used for extracting a frequency band energy characteristic value in the acoustic emission signal according to the harmonic wavelet packet;
the sample extraction module is used for extracting a training sample from the characteristic value;
the model construction module is used for inputting the training samples into an LS-SVM model and training to obtain an LS-SVM cutter wear state recognition model;
and the state identification module is used for identifying the wear state of the cutter according to the LS-SVM cutter wear state identification model.
Compared with the prior art, the tool wear state identification device and the tool wear state identification method have the same advantages, and are not described again.
Drawings
FIG. 1 is a flow chart of a tool wear status identification method according to an embodiment of the present invention;
FIG. 2 is another flow chart of a tool wear status identification method according to an embodiment of the present invention;
FIG. 3 is a flowchart of step S2 according to the embodiment of the present invention;
FIG. 4 is a flowchart of step S6 according to the embodiment of the present invention;
fig. 5 is a block diagram of a tool wear state recognition apparatus according to an embodiment of the present invention.
Description of reference numerals:
10-a signal acquisition module; 20-a feature extraction module; 30-a sample extraction module; 40-a model building module; 50-state identification module.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
Fig. 1 is a flow chart of a method for identifying a wear state of a cutting tool in the present embodiment; the tool wear state identification method comprises the following steps:
step S1, collecting acoustic emission signals in the working process of the cutter;
step S3, extracting a frequency band energy characteristic value in the acoustic emission signal according to the harmonic wavelet packet;
step S5, extracting training samples from the characteristic values;
step S6, inputting the training samples into an LS-SVM model, and training to obtain an LS-SVM cutter wear state recognition model;
and step S8, recognizing the wear state of the cutter according to the LS-SVM cutter wear state recognition model.
Wherein, the acoustic emission sensor can be used for collecting acoustic emission signals in the working process of the cutter. In machining manufacturing, the wear of the cutting tool will eventually propagate to a specific surface of the machining system in the form of elastic waves, these signals are directly derived from the cutting zone, containing information closely related to tool wear, and thus the state of tool wear can be further identified by detecting acoustic emission signals during tool operation. The frequency bandwidth of the acoustic emission sensor is 20 KHz-2 MHz, so that the interference of low-frequency mechanical noise which is seriously polluted in the processing process is avoided, and the acoustic emission sensor has the advantages of low requirement on the proximity degree of a measured component, sensitive linear defect, insensitive geometric dimension and the like.
In step S3, the frequency spectrum of the acoustic emission signal is first obtained, and the acoustic emission signal in the X direction is subjected to multi-layer (e.g., 5-layer) harmonic wavelet packet decomposition, so that the signal frequency can be decomposed into multiple (e.g., 32) frequency bands, the bandwidth of each frequency band is the same, the main frequency of the acoustic emission signal in the X direction is concentrated in the first several frequency bands, and the main frequency of the acoustic emission signal in the X direction is concentrated in the first 6 frequency bands in this embodiment. Energy values of signals of 6 frequency bands are respectively solved by adopting the following formula:
wherein i is 1,2.. 6, which represents the first 6 frequency bands; n represents the number of sampling points, xkRepresenting the sequence after harmonic wavelet transform, EiIs the energy value of the ith frequency band signal. The energy values respectively corresponding to the 6 frequency bands are taken as elements to form a feature vector: t ═ E1,E2,E3,E4,E5,E6]The tool wear state can be identified as an input vector of the LS-SVM.
In this embodiment, multiple feeding is set, one sample is taken in each feeding, and a certain number of feeding states are taken from multiple feeding samples as training samples. Thus, in step S5, the training sample includes feature vectors of a plurality of feed states.
In step S6, the Support Vector Machine (SVM) can effectively implement precision fitting of a high-dimensional nonlinear system based on small samples, and has good generalization by using the principle of minimum structural risk. The least square support vector machine (LS-SVM) is a deformation based on the SVM, when solving a quadratic programming problem, the SVM is solved by inequality, and the LS-SVM is converted into a linear equation set to be solved, so that the advantages of the SVM are achieved, an insensitive loss function is avoided, the complexity of the SVM is reduced, and the calculation speed is increased. The objective optimization function of the LS-SVM algorithm is as follows:
in the formula:
is a kernel space mapping function, w is a weight vector, b is an offset, eiIs an error variable, gamma is a penalty factor, xiFor a training sample input of dimension l, yiIs the training sample output of l dimension, and l is the number of samples.
The LS-SVM model for function estimation can be expressed as:
in the formula: k (x, x)i) For the kernel function, the present embodiment employs a radial basis kernel function; alpha is alphaiIs a lagrange multiplier. The radial basis kernel function is:
in the formula: σ is the width of the radial basis function. After two parameters of a penalty factor gamma and a nuclear parameter sigma are determined, an LS-SVM tool wear state identification model can be established according to the gamma and the sigma.
In this way, the LS-SVM tool wear state identification model can be established by extracting the energy characteristic value of the acoustic emission signal generated in the working process of the tool, and the tool wear state is identified according to the LS-SVM tool wear state identification model; and the frequency band energy characteristic value in the acoustic emission signal is extracted through the harmonic wave packet, for a complex non-stationary signal generated in the tool abrasion process, the harmonic wave packet can decompose the signal to different frequency bands without overlapping and omission, so that the characteristic information of the signal on different frequency bands is obtained, the tool abrasion characteristic signal can be better analyzed, the abrasion pattern recognition can be further carried out, and the accuracy of the pattern recognition is improved.
Optionally, as shown in fig. 2, after the step S1 and before the step S3, the method further includes:
and step S2, denoising the acoustic emission signal.
The tool can generate a great deal of noise in the cutting process, particularly when early wear and failure occur, the data measured by the sensor contains a great deal of noise signals, so that effective acoustic emission signals are submerged, and the change rule of the tool wear cannot be directly found from the effective acoustic emission signals. The observation signal is usually subjected to noise reduction first, and then to feature extraction.
Therefore, the noise reduction is carried out before the energy characteristic value of the acoustic emission signal is extracted, the distortion of the acoustic emission signal can be avoided, and the effectiveness of characteristic extraction is ensured.
Optionally, as shown in fig. 3, step S2 includes:
step S21, acquiring a power spectrum of the acoustic emission signal;
step S22, smoothing the power spectrum of each frequency point of the acoustic emission signal;
step S23, carrying out nonlinear tracking on the minimum power value of each frequency point of the acoustic emission signal;
and step S24, denoising the acoustic emission signal by using a filter.
In step S22, the power spectrum of each frequency point of the acoustic emission signal is smoothed by the following formula:
P(λ,k)＝βP(λ,k)+(1-β)|Y(λ,k)|2
in the formula: lambda is frequency point, k is frame number, P (lambda, k) is power spectrum of acoustic emission signal of k-th frame and lambda frequency point, beta is smoothing factor.
In step S23, if pmin(λ-1,k)<P (lambda, k), namely the minimum value of the power of the acoustic emission signal of the lambda frequency point of the previous frame is less than the power of the acoustic emission signal of the k frame and the lambda frequency point
In the formula: mu and eta are minimum calculation parameters; if p ismin(λ -1, k) ≥ P (λ, k), then
pmin(λ,k)＝P(λ,k)
Therefore, the power minimum value of each frequency point of the acoustic emission signal can be subjected to nonlinear tracking according to the formula.
In this embodiment, the nonlinear tracking in step S23 continuously estimates the noise power, and then determines the attenuation factor of the noise, and applies the attenuation factor to the acoustic emission signal and inputs the acoustic emission signal into the filter to reduce the noise.
Therefore, the acoustic emission signal can be denoised through the steps, and effective acoustic emission signal distortion can be avoided.
Optionally, as shown in fig. 2, after the step S3 and before the step S5, the method further includes:
step S4, normalization processing is performed on the feature value.
Because the extracted characteristic values have different physical meanings and different dimensions, and the magnitude of the characteristic values also has larger variation, the distribution characteristics of the characteristic values need to be unified, the characteristic values are normalized to [0,1], and the normalization formula is as follows:
therefore, the embodiment performs normalization processing on the characteristic values, simplifies the calculation process of the LS-SVM, and is beneficial to accelerating the training speed.
Alternatively, as shown in fig. 4, step S6 includes:
step S61, initializing a BSA (bovine serum albumin) population, setting training parameters and defining a fitness function;
step S62, LS-SVM training is carried out to obtain a fitness value, and an optimal individual and an optimal fitness value are recorded;
step S63, carrying out iterative optimization according to a BSA algorithm, and updating an optimal individual and an optimal fitness value;
step S64, if the end condition is satisfied, outputting an optimized penalty factor and a kernel parameter square of the LS-SVM;
and step S65, establishing the LS-SVM cutter wear state identification model according to the penalty factor and the kernel parameter square.
BSA (bird swarm optimization) is a new population evolution algorithm, a novel disturbance strategy and a novel mixing strategy are adopted, the optimization efficiency is greatly improved, and the algorithm only has one control parameter (mixing ratio parameter), so that the operation is simpler. The algorithm flow of the BSA has two selection operations, which may be referred to as selecting one and selecting two. One is selected to select the historical population and two are selected to update the population. The algorithm flow can be divided into two parts of initializing population, selecting one, mutating, crossing and selecting two 5. Selecting a new historical population oldPop for each iteration, and after the value of oldPop is determined, randomly ordering the individuals in oldPop to generate population oldPop'. After a new population is generated, the sizes of the elements in the population must be checked, and if the elements in the population exceed the extreme value boundaries, a new population is generated. And selecting two recorded current optimal solutions and corresponding solution vectors until the loop termination condition is met, and finally outputting the optimal solution.
In this embodiment, the penalty factor and the kernel parameter square of the LS-SVM are used as the two-dimensional coordinate value of each individual for model training, and the model training is performed according to the penalty factor and the kernel parameter square
Calculating the fitness of the individual; wherein, f fitness, ytIndicating the correct number of classifications and yy indicating the number of errors in classification. Aiming at each individual, comparing the fitness with the optimal value of the individual, and updating the optimal value of the individual; and after the individual optimal value is compared with the population optimal value, the population optimal value is updated. In step S64, the ending condition includes that the number of iterations of the BSA algorithm reaches the maximum or the accuracy reaches a preset requirement, and in this embodiment, the number of iterations of the BSA algorithm reaches the maximum. And after the iteration is finished, determining an optimal penalty factor and a kernel parameter square, and establishing an LS-SVM cutter wear state identification model according to the optimal penalty factor and the kernel parameter square.
In this way, the LS-SVM can be trained through the training samples to obtain an LS-SVM cutter wear state recognition model; and the penalty factor and the kernel parameter of the LS-SVM are automatically selected by applying the BSA optimization algorithm so as to reach the optimal combination value, and the accuracy of model wear identification is greatly improved.
Optionally, as shown in fig. 2, after the step S6 and before the step S8, the method further includes:
and step S7, extracting a test sample from the characteristic value, and inputting the test sample into the LS-SVM cutter wear state identification model for testing.
After the LS-SVM tool wear state identification model is built, the identification accuracy of the built LS-SVM tool wear state identification model needs to be tested, so that the built LS-SVM tool wear state identification model can be confirmed to be used for identifying the tool wear state, and identification errors are avoided.
Optionally, as shown in fig. 2, after step S8, the method further includes:
and step S9, performing function fitting on the continuous wear state values to obtain the derivative of the fitting function in the current wear state value, and discarding the current wear state value if the derivative is less than zero.
Since the tool wear magnitude cannot be reduced with increasing cutting time, this feature can be used for the correction of the recognition result. In this embodiment, function fitting is performed on the continuous wear state values, a fitting function is constructed, a derivative of the fitting function in the current wear state value is obtained, and if the derivative is smaller than zero, it is proved that the current wear state value is smaller than the adjacent historical wear state value. Since the tool wear magnitude cannot be reduced as the cutting time increases, the current recognition result may be considered inaccurate, and the current wear state value may be discarded to correct the recognition result. In step S9, a function fitting may be performed on the discrete continuous wear state values by a least square method.
Therefore, the embodiment can correct the recognition result by calculating the change trend of the wear state value, and further improves the recognition accuracy of the wear state of the cutter.
As shown in fig. 5, the present embodiment also provides a tool wear state recognition apparatus, including:
the signal acquisition module 10 is used for acquiring acoustic emission signals in the working process of the cutter;
the characteristic extraction module 20 is configured to extract a frequency band energy characteristic value in the acoustic emission signal according to the harmonic wavelet packet;
a sample extraction module 30, configured to extract a training sample from the feature values;
the model construction module 40 is used for inputting the training samples into an LS-SVM model and training to obtain an LS-SVM cutter wear state recognition model;
and the state identification module 50 is used for identifying the wear state of the cutter according to the LS-SVM cutter wear state identification model.
In this way, the LS-SVM tool wear state identification model can be established by extracting the energy characteristic value of the acoustic emission signal generated in the working process of the tool, and the tool wear state is identified according to the LS-SVM tool wear state identification model; and the frequency band energy characteristic value in the acoustic emission signal is extracted through the harmonic wave packet, for a complex non-stationary signal generated in the tool abrasion process, the harmonic wave packet can decompose the signal to different frequency bands without overlapping and omission, so that the characteristic information of the signal on different frequency bands is obtained, the tool abrasion characteristic signal can be better analyzed, the abrasion pattern recognition can be further carried out, and the accuracy of the pattern recognition is improved.
Optionally, the signal acquisition module 10 comprises an acoustic emission sensor and a preamplifier. The voltage signal output by the acoustic emission sensor is weak, and in order to improve the signal-to-noise ratio, the signal voltage needs to be amplified and improved to a certain degree through a preamplifier.
Although the present disclosure has been described above, the scope of the present disclosure is not limited thereto. Various changes and modifications may be effected therein by one of ordinary skill in the pertinent art without departing from the spirit and scope of the present disclosure, and these changes and modifications are intended to be within the scope of the present disclosure.
Claims (8)
1. A tool wear state recognition method, comprising:
collecting acoustic emission signals in the working process of the cutter;
extracting a frequency band energy characteristic value in the acoustic emission signal according to the harmonic wavelet packet;
extracting training samples from the characteristic values;
inputting the training samples into an LS-SVM model, and training to obtain an LS-SVM cutter wear state recognition model;
and identifying the wear state of the cutter according to the LS-SVM cutter wear state identification model.
2. The tool wear state identification method according to claim 1, wherein after the acoustic emission signal is collected during the tool operation process and before the frequency band energy characteristic value in the acoustic emission signal is extracted according to the harmonic wavelet packet, the method further comprises:
and denoising the acoustic emission signal.
3. The tool wear state identification method of claim 2, wherein de-noising the acoustic emission signal comprises:
acquiring a power spectrum of the acoustic emission signal;
smoothing the power spectrum of each frequency point of the acoustic emission signal;
carrying out nonlinear tracking on the minimum power value of each frequency point of the acoustic emission signal;
and denoising the acoustic emission signal by using a filter.
4. The tool wear state identification method according to claim 2, wherein after the extracting of the frequency band energy feature value in the acoustic emission signal from the harmonic wavelet packet and before the extracting of the training sample from the feature value, the method further comprises:
and carrying out normalization processing on the characteristic values.
5. The tool wear state recognition method according to claim 1, wherein the training samples are input into an LS-SVM model, and the LS-SVM tool wear state recognition model is obtained through training, and the method comprises the following steps:
initializing a BSA (bovine serum albumin) population, setting training parameters and defining a fitness function;
performing LS-SVM training to obtain a fitness value, and recording an optimal individual and the optimal fitness value;
performing iterative optimization according to a BSA algorithm, and updating an optimal individual and an optimal fitness value;
if the end condition is met, outputting an optimized punishment factor and a kernel parameter square of the LS-SVM;
and establishing a wear state identification model of the LS-SVM cutter according to the penalty factor and the square of the nuclear parameter.
6. The tool wear state recognition method according to claim 5, wherein after the training samples are input into the LS-SVM model and the LS-SVM tool wear state recognition model is obtained through training, before the wear state of the tool is recognized according to the LS-SVM tool wear state recognition model, the method further comprises:
and extracting a test sample from the characteristic value, and inputting the test sample into the LS-SVM cutter wear state identification model for testing.
7. The tool wear state recognition method according to claim 1, wherein after recognizing the wear state of the tool according to the LS-SVM tool wear state recognition model, the method further comprises:
and performing function fitting on the continuous wear state values to obtain a derivative of the fitting function in the current wear state value, and discarding the current wear state value if the derivative is less than zero.
8. A tool wear state recognition device, comprising:
the signal acquisition module (10) is used for acquiring acoustic emission signals in the working process of the cutter;
the characteristic extraction module (20) is used for extracting a frequency band energy characteristic value in the acoustic emission signal according to the harmonic wavelet packet;
a sample extraction module (30) for extracting training samples from the feature values;
the model construction module (40) is used for inputting the training samples into an LS-SVM model and training to obtain an LS-SVM cutter wear state recognition model;
and the state recognition module (50) is used for recognizing the wear state of the cutter according to the LS-SVM cutter wear state recognition model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910912879.XA CN110682159A (en) | 2019-09-25 | 2019-09-25 | Cutter wear state identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910912879.XA CN110682159A (en) | 2019-09-25 | 2019-09-25 | Cutter wear state identification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110682159A true CN110682159A (en) | 2020-01-14 |
Family
ID=69110276
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910912879.XA Pending CN110682159A (en) | 2019-09-25 | 2019-09-25 | Cutter wear state identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110682159A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111558849A (en) * | 2020-05-11 | 2020-08-21 | 内蒙古工业大学 | Disc milling cutter machining parameter optimization method and device, electronic equipment and storage medium |
| CN111590391A (en) * | 2020-06-04 | 2020-08-28 | 浙江中德自控科技股份有限公司 | Tool wear identification and prediction method based on high-low frequency acoustic emission sensor fusion |
| CN111721835A (en) * | 2020-06-28 | 2020-09-29 | 上海理工大学 | Intelligent monitoring method for grinding wheel state of hollow drill |
| CN112207631A (en) * | 2020-09-21 | 2021-01-12 | 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 |
| CN112884027A (en) * | 2021-02-02 | 2021-06-01 | 北京航空航天大学 | Cutting process real-time state monitoring method and device based on pattern recognition |
| CN113624848A (en) * | 2021-08-10 | 2021-11-09 | 北京理工大学 | Cutting state identification method and system based on acoustic emission |
| CN115008254A (en) * | 2022-05-30 | 2022-09-06 | 河源富马硬质合金股份有限公司 | Cutter state monitoring method in high-speed milling process |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4707687A (en) * | 1986-05-05 | 1987-11-17 | General Electric Company | Detector to discriminate between tool break acoustic signatures and spiky noise |
| CN103105820A (en) * | 2012-05-22 | 2013-05-15 | 华中科技大学 | Machining cutter abrasion state identification method of numerical control machine tool |
| CN109605127A (en) * | 2019-01-21 | 2019-04-12 | 南京航空航天大学 | A kind of cutting-tool wear state recognition methods and system |
| CN109635847A (en) * | 2018-11-19 | 2019-04-16 | 昆明理工大学 | A kind of cutting-tool wear state recognition methods based on vibration and sound emission |
| CN110060695A (en) * | 2019-04-24 | 2019-07-26 | 百度在线网络技术（北京）有限公司 | Information interacting method, device, server and computer-readable medium |
-
2019
- 2019-09-25 CN CN201910912879.XA patent/CN110682159A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4707687A (en) * | 1986-05-05 | 1987-11-17 | General Electric Company | Detector to discriminate between tool break acoustic signatures and spiky noise |
| CN103105820A (en) * | 2012-05-22 | 2013-05-15 | 华中科技大学 | Machining cutter abrasion state identification method of numerical control machine tool |
| CN109635847A (en) * | 2018-11-19 | 2019-04-16 | 昆明理工大学 | A kind of cutting-tool wear state recognition methods based on vibration and sound emission |
| CN109605127A (en) * | 2019-01-21 | 2019-04-12 | 南京航空航天大学 | A kind of cutting-tool wear state recognition methods and system |
| CN110060695A (en) * | 2019-04-24 | 2019-07-26 | 百度在线网络技术（北京）有限公司 | Information interacting method, device, server and computer-readable medium |
Non-Patent Citations (1)
| Title |
|---|
| 董彩云等: ""基于谐波小波包和BSA优化LS-SVM的铣刀磨损状态识别研究"", 《中国机械工程》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111558849A (en) * | 2020-05-11 | 2020-08-21 | 内蒙古工业大学 | Disc milling cutter machining parameter optimization method and device, electronic equipment and storage medium |
| CN111558849B (en) * | 2020-05-11 | 2021-10-22 | 内蒙古工业大学 | Disc milling cutter machining parameter optimization method and device, electronic equipment and storage medium |
| CN111590391A (en) * | 2020-06-04 | 2020-08-28 | 浙江中德自控科技股份有限公司 | Tool wear identification and prediction method based on high-low frequency acoustic emission sensor fusion |
| CN111721835A (en) * | 2020-06-28 | 2020-09-29 | 上海理工大学 | Intelligent monitoring method for grinding wheel state of hollow drill |
| CN111721835B (en) * | 2020-06-28 | 2023-03-31 | 上海理工大学 | Intelligent monitoring method for grinding wheel state of hollow drill |
| CN112207631A (en) * | 2020-09-21 | 2021-01-12 | 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 |
| CN112884027A (en) * | 2021-02-02 | 2021-06-01 | 北京航空航天大学 | Cutting process real-time state monitoring method and device based on pattern recognition |
| CN113624848A (en) * | 2021-08-10 | 2021-11-09 | 北京理工大学 | Cutting state identification method and system based on acoustic emission |
| CN115008254A (en) * | 2022-05-30 | 2022-09-06 | 河源富马硬质合金股份有限公司 | Cutter state monitoring method in high-speed milling process |
| CN115008254B (en) * | 2022-05-30 | 2024-06-04 | 河源富马硬质合金股份有限公司 | Method for monitoring state of cutter in high-speed milling process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110682159A (en) | Cutter wear state identification method and device | |
| Tran et al. | Effective multi-sensor data fusion for chatter detection in milling process | |
| WO2021004154A1 (en) | Method for predicting remaining life of numerical control machine tool | |
| CN111687689A (en) | Cutter wear state prediction method and device based on LSTM and CNN | |
| CN114619292B (en) | Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network | |
| CN111168471A (en) | Numerical control machine tool cutter wear monitoring method | |
| CN113927371A (en) | Cutter wear prediction method based on multi-sensor feature fusion | |
| CN110928237B (en) | Vibration signal-based numerical control machining center flutter online identification method | |
| CN112207631B (en) | Method for generating tool detection model, method, system, device and medium for detecting tool detection model | |
| CN114273977A (en) | MES-based cutter wear detection method and system | |
| CN111783544B (en) | Method for building diamond milling grinding head state monitoring system for processing ceramic mobile phone backboard | |
| CN117434153B (en) | Road nondestructive testing method and system based on ultrasonic technology | |
| CN113752089A (en) | Cutter state monitoring method based on singularity Leersian index | |
| CN115169409A (en) | Bridge structure natural vibration frequency identification and early warning method and device based on sliding window | |
| CN113269169A (en) | Bearing fault detection method and device | |
| CN112747921A (en) | Multi-sensor mechanical fault diagnosis method based on NA-MEMD | |
| CN112380992A (en) | Method and device for evaluating and optimizing accuracy of monitoring data in machining process | |
| CN115459868A (en) | Millimeter wave communication performance evaluation method and system in complex environment | |
| CN103203670A (en) | Grinding chatter prediction method based on maximum information entropy and directional divergence | |
| Wan et al. | Chatter detection for micro milling considering environment noises without the requirement of dominant frequency | |
| CN114813129B (en) | Rolling bearing acoustic signal fault diagnosis method based on WPE and EMD | |
| CN116482526A (en) | Analysis system for non-fault phase impedance relay | |
| CN114034375B (en) | Ultra-high voltage transmission line noise measurement system and method | |
| CN110111275A (en) | A kind of method of signal de-noising, system and computer storage medium | |
| CN114818813A (en) | Cutter wear value prediction algorithm based on self-attention convolution neural network |
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: 20200114 |
|
| RJ01 | Rejection of invention patent application after publication |