CN114918735A - PCC-LSTM-based milling cutter wear prediction method - Google Patents
PCC-LSTM-based milling cutter wear prediction method Download PDFInfo
- Publication number
- CN114918735A CN114918735A CN202210557444.XA CN202210557444A CN114918735A CN 114918735 A CN114918735 A CN 114918735A CN 202210557444 A CN202210557444 A CN 202210557444A CN 114918735 A CN114918735 A CN 114918735A
- Authority
- CN
- China
- Prior art keywords
- lstm
- wear
- value
- milling cutter
- signal
- 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
- 238000003801 milling Methods 0.000 title claims abstract description 46
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000004364 calculation method Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 10
- 230000009467 reduction Effects 0.000 claims abstract description 9
- 238000010606 normalization Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims abstract description 4
- 238000005299 abrasion Methods 0.000 claims description 17
- 238000003754 machining Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 4
- 230000003321 amplification Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 3
- 230000005484 gravity Effects 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims description 2
- 238000005520 cutting process Methods 0.000 abstract description 12
- 230000008569 process Effects 0.000 abstract description 8
- 238000004458 analytical method Methods 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 239000000463 material Substances 0.000 description 3
- 229910001018 Cast iron Inorganic materials 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 229910045601 alloy Inorganic materials 0.000 description 1
- 239000000956 alloy Substances 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23Q—DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING; MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
- B23Q17/00—Arrangements for observing, indicating or measuring on machine tools
- B23Q17/09—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool
- B23Q17/0952—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool during machining
- B23Q17/0957—Detection of tool breakage
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a PCC-LSTM-based milling cutter wear prediction method, which comprises the steps of collecting current, vibration and acoustic emission signals of a workbench and a spindle in a working process; denoising the acquired signal information by adopting a moving average method; normalizing the noise-reduced signal; extracting time domain and frequency domain signal characteristics of the data subjected to noise reduction and normalization processing, and storing all characteristic values in a characteristic matrix X; optimizing the features based on the Pearson correlation coefficient; and establishing an LSTM network for prediction. The invention provides a milling cutter wear rapid prediction method by fully utilizing the dimension reduction function of a PCC method and the analysis and prediction function of LSTM on time sequence, which can overcome the limitation that the calculation precision and the solving efficiency can not be coordinated and unified in the milling cutter wear prediction in the cutting process in the prior art, and realize the rapid and accurate prediction of the milling cutter wear.
Description
Technical Field
The invention relates to a PCC-LSTM-based milling cutter wear prediction method, and belongs to the technical field of machining process monitoring.
Background
In the production and manufacturing processes of mechanical part products, cutter abrasion can bring stress concentration and temperature rise of a cutting area, so that the quality of a machined workpiece is deteriorated, and even the precision of a machine tool is influenced in severe cases. The sensor is adopted to monitor the abrasion state of the milling cutter in the machining process, and the abrasion state of the milling cutter is predicted in advance, so that the method is one of important measures for improving the machining precision and is an important component of intelligent manufacturing.
In the prior art, algorithms such as hidden Markov, Bayesian and BP neural networks are adopted, in order to obtain an accurate prediction result, the algorithms need a large amount of data for training and prediction, and a large amount of input data means that a large amount of computing resources need to be consumed, so that the solving efficiency is reduced.
Disclosure of Invention
The invention provides a milling cutter wear prediction method based on PCC-LSTM, aiming at overcoming the technical problem that the calculation precision and the solving efficiency of the milling cutter wear prediction in the prior art can not be coordinated and unified.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a milling cutter wear prediction method based on PCC-LSTM comprises the following steps:
step 4, extracting time domain and frequency domain signal characteristics of the data subjected to noise reduction and normalization processing, wherein the time domain and frequency domain signal characteristics include an average value, a standard deviation, skewness, kurtosis, a peak-to-peak value, a root-mean-square, an amplitude (peak value) factor, a form factor, a pulse factor, a margin factor, a center-of-gravity frequency, a frequency variance and a mean-square frequency, 13 characteristic values are calculated in total, all sensors acquire the characteristic values, the number of which is num _ features 13 × num _ sensors, and all the characteristic values are stored in a characteristic matrix X;
and 8, inputting the optimal characteristics obtained in the step 5 into the network established in the step 7, and solving to obtain a predicted value, namely an output node of the LSTM network.
The invention solves the technical problem by adopting a mode of combining PCC and LSTM. Firstly, preferably obtaining a characteristic with larger correlation with cutter wear through a PCC algorithm, on one hand, preferably reducing the number of input characteristics of an LSTM network through the characteristic so as to improve the calculation efficiency, and on the other hand, ensuring that the preferred characteristic has strong correlation with the cutter wear through the PCC so as to ensure the calculation accuracy; secondly, the cutter wear prediction problem is converted into a time series prediction problem, the analysis and prediction advantages of the LSTM on the time series are fully exerted under the condition of fewer data samples, and the cutter wear prediction precision is improved.
The examples show that the actual wear values of the tool are: 0.23mm, 0.25mm, 0.28mm, 0.32mm, 0.36mm, 0.41mm, 0.48mm, 0.53 mm. The predicted values are respectively: 0.08mm, 0.22mm, 0.27mm, 0.32mm, 0.35mm, 0.40 mm, 0.48mm, 0.50mm, with an average error of about 11%. The predicted time for each tool wear value is less than 0.5 seconds. Therefore, the invention can improve the calculation efficiency and realize the coordination and unification of the calculation accuracy and the solving efficiency on the premise of ensuring the prediction accuracy.
Further, the current, vibration and acoustic emission signals in the step 1 are respectively obtained through a current sensor, a vibration sensor and an acoustic emission sensor, the acoustic emission sensor is installed on a workbench of a machining center, the vibration sensor is installed on a main shaft, and the signals are subjected to amplification and filtering to carry out data acquisition; the current sensor is connected to a power supply circuit of the motor.
The metal cutting process is essentially a chip deformation and contact friction process of a tool flank and a machined surface, during the milling process, the tool geometry changes due to abrasion, the changed tool geometry increases the non-uniformity of material deformation, and the friction characteristic of the tool flank and the machined surface is changed, so that the signals of cutting force, vibration, current, acoustic emission and the like change along with the change of the signals of cutting force, vibration, current, acoustic emission and the like, namely the change of the tool abrasion state can cause the change of the signals of cutting force, vibration, current, acoustic emission and the like generated during the machining process. Tool wear can therefore be judged by means of the current, vibration and acoustic emission signals.
Further, the normalized transfer function in step 3 is as follows:
wherein x is * Is a normalized value, x is the original value in the sample, x max And x min The maximum and minimum values of the sample data, respectively.
Further, the pearson correlation coefficient calculation formula in step 5 is:
wherein r is Pearson's correlation coefficient, X i For the corresponding sample signal characteristic values in different wear states,
is the average value of corresponding sample signal characteristic values in all wear states, Y i For the corresponding milling tool wear values in different wear states,
the average value of the corresponding milling cutter wear values in all wear states is obtained, i is the serial number of the wear states of the milling cutter, and n is the number of the wear states of the milling cutter; preferred feature numbers are:
and num _ features _ select, wherein num _ features _ selected is 0.25 num _ features, and the solution value of num _ features _ selected is rounded if it is a non-integer.
The invention has the beneficial effects that:
the milling cutter wear prediction method provided by the invention fully utilizes the dimensionality reduction function of the PCC method and the analysis and prediction function of the LSTM on the time sequence, can overcome the limitation that the calculation precision and the solving efficiency cannot be coordinated and unified in the milling cutter wear prediction in the cutting process, and realizes the rapid and accurate prediction of the milling cutter wear.
Drawings
FIG. 1 is a process flow of the present invention;
FIG. 2 is an example of a vibration signal collected in an embodiment of the present invention;
FIG. 3 shows the predicted results according to the embodiment of the present invention.
Detailed Description
The present invention is further described with reference to the following examples, which are provided for more clearly illustrating the technical solutions of the present invention and are not intended to limit the scope of the present invention.
In the embodiment, a Matsuura processing center is adopted in a data set experiment for milling, current, vibration and acoustic emission signals in the processing process are collected, and 6 sensors including 2 vibration sensors, 2 current sensors and 2 acoustic emission sensors are used. The acoustic emission sensor and the vibration sensor are respectively arranged on a workbench and a main shaft of the machining center, and data acquisition is carried out on signals through amplification and filtering. The ACOUSTIC emission sensor is WD 925(PHYSICAL AC Acoustic GROUP, frequency range up to 2 MHz); the vibration sensor is 7201-50, ENDEVCO. The signal from the spindle motor current sensor was recorded directly in the computer with a sampling frequency of 100 kHZ.
The milling cutter selects a face milling cutter with the diameter of 70mm, and comprises 6 cutting blades, the model of which is KC710 coating hard alloy blades. The experimental parameters refer to actual plant use parameters. The cutting speed was set at 200m/min (i.e. 826rev/min), the cutting depth was set at 0.75mm, the feed speed was set at 0.5mm/rev, the material being machined was cast iron, and the experimental sample size was 483 × 178 × 51 mm.
The test gathers the data set that 8 groups of cutters wearing state corresponds altogether, and 8 real values of flank face wearing capacity that wearing state corresponds are respectively: 0.23mm, 0.25mm, 0.28mm, 0.32mm, 0.36mm, 0.41mm, 0.48mm, 0.53 mm. The specific method comprises the following steps:
taking the 1 st wear state of the tool as an example, the raw data collected are shown in table 1 below.
TABLE 1
n is 0.25 × 100000 × 60/(360 × 826) ═ 5.044, rounded down to 5;
the data after noise reduction is shown in table 2, and the first 100 data are taken as the display here because the data size is large.
TABLE 2
wherein x is * Is a normalized value, x is the original value in the sample, x max And x min The maximum and minimum values of the sample data, respectively.
The normalized data are shown in table 3, and the top 100 data are displayed because of the large data size.
TABLE 3
Step 4, extracting time domain and frequency domain signal features of the data subjected to noise reduction and normalization processing, wherein the time domain and frequency domain signal features include an average value, a standard deviation, a skewness, a kurtosis, a peak-to-peak value, a root mean square, an amplitude (peak) factor, a form factor, a pulse factor, a margin factor, a center-of-gravity frequency, a frequency variance, a mean-square frequency and the like, 13 feature values are counted, all sensors obtain the feature values, the number of the feature values is num _ features ═ 13 × num _ sensors, and a calculation formula of the 13 feature values corresponding to each sensor is shown in the following table 4:
TABLE 4
In this embodiment, the number of the characteristic values obtained by the sensors is 78-13 × 6, and all the characteristic values are stored in the characteristic matrix X, where 1 tool wear state corresponds to 78 characteristic values (the tool wear state refers to a value at which the wear amount of 1 tool among 6 blades is the largest); wherein the characteristic numbered 1 is a mean value of the 1 st current signal, the characteristic numbered 2 is a standard deviation of the 1 st current signal, the characteristic numbered 3 is a skewness of the 1 st current signal, the characteristic numbered 4 is a kurtosis of the 1 st current signal, the characteristic numbered 5 is a peak-to-peak value of the 1 st current signal, the characteristic numbered 6 is a root mean square of the 1 st current signal, the characteristic numbered 7 is an amplitude (peak) factor of the 1 st current signal, the characteristic numbered 8 is a form factor of the 1 st current signal, the characteristic numbered 9 is a pulse factor of the 1 st current signal, the characteristic numbered 10 is a margin factor of the 1 st current signal, the characteristic numbered 11 is a center-of-gravity frequency of the 1 st current signal, the characteristic numbered 12 is a frequency variance of the 1 st current signal, the characteristic numbered 13 is a mean-square frequency of the 1 st current signal, the feature number 14 is an average value of the 2 nd current signal, the feature number 15 is a standard deviation of the 2 nd current signal, the feature number 16 is a skewness of the 2 nd current signal, the feature number 17 is a kurtosis of the 2 nd current signal, the feature number 18 is a peak-to-peak value of the 2 nd current signal, the feature number 19 is a root mean square of the 2 nd current signal, the feature number 20 is an amplitude (peak value) factor of the 2 nd current signal, the feature number 21 is a form factor of the 2 nd current signal, the feature number 22 is a pulse factor of the 2 nd current signal, the feature number 23 is a margin factor of the 2 nd current signal, the feature number 24 is a center-of-gravity frequency of the 2 nd current signal, the feature number 25 is a frequency variance of the 2 nd current signal, the feature number 26 is a mean-square frequency of the 2 nd current signal, the feature number 27 is an average value of the 1 st vibration signal, the feature number 28 is a standard deviation of the 1 st vibration signal, the feature number 29 is a skewness of the 1 st vibration signal, the feature number 30 is a kurtosis of the 1 st vibration signal, the feature number 31 is a peak-to-peak value of the 1 st vibration signal, the feature number 32 is a root-mean-square of the 1 st vibration signal, the feature number 33 is an amplitude (peak value) factor of the 1 st vibration signal, the feature number 34 is a form factor of the 1 st vibration signal, the feature number 35 is a pulse factor of the 1 st vibration signal, the feature number 36 is a margin factor of the 1 st vibration signal, the feature number 37 is a center-of-gravity frequency of the 1 st vibration signal, the feature number 38 is a frequency variance of the 1 st vibration signal, the feature number 39 is a mean-square frequency of the 1 st vibration signal, the feature number 40 is an average value of the 2 nd vibration signal, the feature number 41 is a standard deviation of the 2 nd vibration signal, the feature number 42 is a skewness of the 2 nd vibration signal, the feature number 43 is a kurtosis of the 2 nd vibration signal, the feature number 44 is a peak-to-peak value of the 2 nd vibration signal, the feature number 45 is a root-mean-square of the 2 nd vibration signal, the feature number 46 is an amplitude (peak value) factor of the 2 nd vibration signal, the feature number 47 is a form factor of the 2 nd vibration signal, the feature number 48 is a pulse factor of the 2 nd vibration signal, the feature number 49 is a margin factor of the 2 nd vibration signal, the feature number 50 is a center-of-gravity frequency of the 2 nd vibration signal, the feature number 51 is a frequency variance of the 2 nd vibration signal, the feature number 52 is a mean-square frequency of the 2 nd vibration signal, feature number 53 is an average of the 1 st acoustic emission signal, feature number 54 is a standard deviation of the 1 st acoustic emission signal, feature number 55 is a skewness of the 1 st acoustic emission signal, feature number 56 is a kurtosis of the 1 st acoustic emission signal, feature number 57 is a peak-to-peak of the 1 st acoustic emission signal, feature number 58 is a root-mean-square of the 1 st acoustic emission signal, feature number 59 is an amplitude (peak) factor of the 1 st acoustic emission signal, feature number 60 is a form factor of the 1 st acoustic emission signal, feature number 61 is a pulse factor of the 1 st acoustic emission signal, feature number 62 is a margin factor of the 1 st acoustic emission signal, feature number 63 is a center-of-gravity frequency of the 1 st acoustic emission signal, feature number 64 is a variance of the frequency of the 1 st acoustic emission signal, number 65 is characterized by the mean square frequency of the 1 st acoustic emission signal, number 66 is characterized by the mean of the 2 nd acoustic emission signal, number 67 is characterized by the standard deviation of the 2 nd acoustic emission signal, number 68 is characterized by the skewness of the 2 nd acoustic emission signal, number 69 is characterized by the kurtosis of the 2 nd acoustic emission signal, number 70 is characterized by the peak-to-peak value of the 2 nd acoustic emission signal, number 71 is characterized by the root mean square of the 2 nd acoustic emission signal, number 72 is characterized by the amplitude (peak) factor of the 2 nd acoustic emission signal, number 73 is characterized by the form factor of the 2 nd acoustic emission signal, number 74 is characterized by the pulse factor of the 2 nd acoustic emission signal, number 75 is characterized by the margin factor of the 2 nd acoustic emission signal, number 76 is characterized by the center of gravity frequency of the 2 nd acoustic emission signal, the feature numbered 77 is the variance in frequency of the 2 nd acoustic emission signal and the feature numbered 78 is the mean square frequency of the 2 nd acoustic emission signal. As described in table 5 below:
TABLE 5
wherein r is Pearson's correlation coefficient, X i For the corresponding sample signal characteristic values in different wear states,
is the average value of the corresponding sample signal characteristic values in all wear states, Y i For the corresponding milling tool wear values in different wear states,
and the average value of the corresponding milling cutter abrasion values in all abrasion states is shown, i is the serial number of the abrasion states of the milling cutter, and n is the number of the abrasion states of the milling cutter.
The above 20 preferred features are respectively: the average value of the 1 st current signal, the standard deviation of the 1 st current signal, the skewness of the 1 st current signal, the standard deviation of the 2 nd current signal, the kurtosis of the 2 nd current signal, the peak-to-peak value of the 2 nd current signal, the average value of the 1 st vibration signal, the standard deviation of the 1 st vibration signal, the skewness of the 1 st vibration signal, the kurtosis of the 1 st vibration signal, the pulse factor of the 1 st vibration signal, the margin factor of the 1 st vibration signal, the peak-to-peak value of the 2 nd vibration signal, the root mean square of the 2 nd vibration signal, the amplitude (peak) factor of the 2 nd vibration signal, the skewness of the 1 st acoustic emission signal, the root mean square of the 1 st acoustic emission signal, the average value of the 2 nd acoustic emission signal, the peak-to-peak value of the 2 nd acoustic emission signal, the root mean square of the 2 nd acoustic emission signal.
step 8, inputting the preferred characteristics obtained in step 5 into the networks established in step 6 and step 7, and solving to obtain predicted values, that is, output nodes of the LSTM network, in this embodiment, by inputting 20 characteristic values corresponding to 8 different wear states, the predicted values of the 8 different wear states are respectively: 0.08mm, 0.22mm, 0.27mm, 0.32mm, 0.35mm, 0.40 mm, 0.48mm, 0.50 mm. The predicted time for each tool wear value is less than 0.5 seconds.
According to the calculation method, the obtained milling cutter wear prediction result is shown in fig. 3, the abscissa in the graph represents different cutter wear states, namely different wear values of the rear tool face of the cutter, the ordinate represents a specific wear numerical value of the cutter, the unit is millimeter, the point in the graph represents a real cutter wear value, the line in the graph represents a predicted value obtained by the method, and the result shows that the coincidence degree between the real value and the predicted value is high, so that the milling cutter wear value can be rapidly predicted.
The rapid prediction method for the abrasion of the milling cutter provided by the invention is used for carrying out PCC characteristic dimension reduction by combining the characteristics of interrupted cutting, large signal data volume, high acquired signal dimension and the like in the milling process, and meanwhile, the abrasion state of the milling cutter in the processing process is evaluated by adopting the abrasion quantity of the rear cutter face, so that the limitation of the abrasion prediction of the milling cutter in the cutting processing process can be overcome, and the calculation precision and the solving efficiency can not be coordinated and unified.
If the scheme is popularized in related enterprises, the process reliability of production and manufacturing of high-added-value products made of difficult-to-process materials can be greatly improved, the product quality is further improved, and meanwhile, the milling cutter which does not reach the service life can be prevented from being abandoned, so that the cost of the enterprise is reduced. Therefore, the patent has higher use value and wide market prospect.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (5)
1. A milling cutter wear prediction method based on PCC-LSTM is characterized in that: the method comprises the following steps:
step 1, collecting current, vibration and acoustic emission signals of a workbench and a main shaft in a machining process, wherein the number of signal sources is num _ sensors;
step 2, denoising the acquired signal information by adopting a moving average method, wherein the average data item number is N, and the solving method of N is as follows: n is 0.25 f 60/(360N), and when the solved N is a non-integer, rounding is performed downwards, wherein f is the sampling frequency and the unit is Hz; n is the rotation speed of the milling cutter, and the unit is rotation/min;
step 3, carrying out normalization processing on the noise-reduced signals in the step 2;
step 4, extracting time domain and frequency domain signal characteristics of the data subjected to noise reduction and normalization processing, and storing all characteristic values in a characteristic matrix X;
step 5, optimizing the characteristics based on the Pearson correlation coefficient, and selecting a plurality of characteristics with larger Pearson correlation coefficient;
step 6, establishing an LSTM network which comprises 4 layers, wherein the 1 st layer is an LSTM layer and comprises 128 units; layer 2 is an LSTM layer, containing 64 cells; the 3 rd layer is an LSTM layer and comprises 32 units; the 4 th layer is a full connection layer and comprises 1 output node, and the loss function is MSE;
step 7, aiming at the network established in the step 6, training by adopting a cutter wear data set with a label, wherein the data set comprises current, vibration and sound emission sensor signals of a workbench and a main shaft acquired under each wear state of the cutter and a cutter wear value, the cutter wear value is evaluated by adopting the wear amount of a rear cutter face, and the training is finished when the value of a loss function MSE is less than 0.01 to obtain an LSTM network;
and 8, inputting the optimal characteristics obtained in the step 5 into the network established in the step 7, and solving to obtain a predicted value, namely an output node of the LSTM network.
2. A PCC-LSTM based milling tool wear prediction method according to claim 1, characterized in that: the current, vibration and acoustic emission signals are respectively obtained through a current sensor, a vibration sensor and an acoustic emission sensor, the acoustic emission sensor is installed on a workbench of a machining center, the vibration sensor is installed on a main shaft, and the signals are subjected to amplification and filtering to acquire data; the current sensor is connected to a power supply circuit of the motor.
3. A PCC-LSTM based milling tool wear prediction method according to claim 1, characterized in that: the normalized transfer function in step 3 is as follows:
wherein x is * Is a normalized value, x is the original value in the sample, x max And x min The maximum and minimum values of the sample data, respectively.
4. The PCC-LSTM based milling tool wear prediction method according to claim 1, characterized in that: in step 4, the characteristics include an average value, a standard deviation, a skewness, a kurtosis, a peak-to-peak value, a root mean square, an amplitude (peak) factor, a form factor, a pulse factor, a margin factor, a center of gravity frequency, a frequency variance and a mean square frequency, and 13 characteristic values are calculated in total, and the number of the characteristic values obtained by all the sensors is num _ features 13 num _ sensors.
5. The PCC-LSTM based milling tool wear prediction method according to claim 1, characterized in that: the pearson correlation coefficient calculation formula in step 5 is:
wherein r is Pearson's correlation coefficient, X i For corresponding sample signal characteristic values in different wear states,
is the average value of the corresponding sample signal characteristic values in all wear states, Y i For the corresponding milling tool wear values in different wear states,
the average value of the corresponding milling cutter abrasion values in all abrasion states is shown, i is the milling cutter abrasion state serial number, and n is the number of the milling cutter abrasion states; preferred feature numbers are:
and num _ features _ select, wherein num _ features _ selected is 0.25 × num _ features, and if the solution value of num _ features _ selected is a non-integer, it is rounded.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210557444.XA CN114918735A (en) | 2022-05-19 | 2022-05-19 | PCC-LSTM-based milling cutter wear prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210557444.XA CN114918735A (en) | 2022-05-19 | 2022-05-19 | PCC-LSTM-based milling cutter wear prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114918735A true CN114918735A (en) | 2022-08-19 |
Family
ID=82811316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210557444.XA Pending CN114918735A (en) | 2022-05-19 | 2022-05-19 | PCC-LSTM-based milling cutter wear prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114918735A (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919977A (en) * | 2015-12-25 | 2017-07-04 | 科大讯飞股份有限公司 | A kind of feedforward sequence Memory Neural Networks and its construction method and system |
| CN110472774A (en) * | 2019-07-19 | 2019-11-19 | 西北工业大学 | Variable working condition bottom tool method for predicting residual useful life based on LSTM network |
| CN110516735A (en) * | 2019-08-27 | 2019-11-29 | 天津科技大学 | A kind of natural gas line event category method based on LSTM network and Adam algorithm |
| CN110576336A (en) * | 2019-09-11 | 2019-12-17 | 大连理工大学 | Method for monitoring abrasion loss of deep hole machining tool based on SSAE-LSTM model |
| CN112435363A (en) * | 2019-08-26 | 2021-03-02 | 贵州大学 | Cutter wear state real-time monitoring method |
| CN113601261A (en) * | 2021-08-10 | 2021-11-05 | 中国科学院合肥物质科学研究院 | Monitoring method of online rapid optimization model for cutter |
| CN113927371A (en) * | 2021-11-05 | 2022-01-14 | 太原科技大学 | Cutter wear prediction method based on multi-sensor feature fusion |
| CN114102260A (en) * | 2021-11-22 | 2022-03-01 | 西安交通大学 | Mechanism-data fusion driven variable working condition cutter wear state monitoring method |
| CN114161227A (en) * | 2021-12-28 | 2022-03-11 | 福州大学 | Cutter wear loss monitoring method based on simulation feature and signal feature fusion |
-
2022
- 2022-05-19 CN CN202210557444.XA patent/CN114918735A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919977A (en) * | 2015-12-25 | 2017-07-04 | 科大讯飞股份有限公司 | A kind of feedforward sequence Memory Neural Networks and its construction method and system |
| CN110472774A (en) * | 2019-07-19 | 2019-11-19 | 西北工业大学 | Variable working condition bottom tool method for predicting residual useful life based on LSTM network |
| CN112435363A (en) * | 2019-08-26 | 2021-03-02 | 贵州大学 | Cutter wear state real-time monitoring method |
| CN110516735A (en) * | 2019-08-27 | 2019-11-29 | 天津科技大学 | A kind of natural gas line event category method based on LSTM network and Adam algorithm |
| CN110576336A (en) * | 2019-09-11 | 2019-12-17 | 大连理工大学 | Method for monitoring abrasion loss of deep hole machining tool based on SSAE-LSTM model |
| CN113601261A (en) * | 2021-08-10 | 2021-11-05 | 中国科学院合肥物质科学研究院 | Monitoring method of online rapid optimization model for cutter |
| CN113927371A (en) * | 2021-11-05 | 2022-01-14 | 太原科技大学 | Cutter wear prediction method based on multi-sensor feature fusion |
| CN114102260A (en) * | 2021-11-22 | 2022-03-01 | 西安交通大学 | Mechanism-data fusion driven variable working condition cutter wear state monitoring method |
| CN114161227A (en) * | 2021-12-28 | 2022-03-11 | 福州大学 | Cutter wear loss monitoring method based on simulation feature and signal feature fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | Research on the milling tool wear and life prediction by establishing an integrated predictive model | |
| CN106181579B (en) | A kind of Tool Wear Monitoring method based on multisensor current signal | |
| CN110059442B (en) | Turning tool changing method based on part surface roughness and power information | |
| CN111037365B (en) | Cutter state monitoring data set enhancing method based on generative countermeasure network | |
| CN111563301A (en) | Thin-wall part milling parameter optimization method | |
| Kilickap | Modeling and optimization of burr height in drilling of Al-7075 using Taguchi method and response surface methodology | |
| Quintana et al. | Surface roughness monitoring application based on artificial neural networks for ball-end milling operations | |
| CN104517033B (en) | A kind of numerical control processing technology parameter Multipurpose Optimal Method towards energy efficiency | |
| CN114619292B (en) | Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network | |
| CN111650890B (en) | Numerical control turning batch processing technological parameter energy-saving optimization method considering cutter abrasion | |
| CN113126564B (en) | Digital twin driven numerical control milling cutter abrasion on-line monitoring method | |
| CN111761411A (en) | Cutter health prediction device and method based on vibration signals | |
| CN114273974A (en) | Vibration signal-based tool runout parameter online estimation method in high-speed milling | |
| CN114326593B (en) | Cutter life prediction system and method | |
| Azimi et al. | An optimization on the DIN1. 2080 alloy in the electrical discharge machining process using ANN and GA | |
| CN114918735A (en) | PCC-LSTM-based milling cutter wear prediction method | |
| Zoghipour et al. | Modeling and optimization of drilling operation of lead-free brass alloys considering various cutting tool geometries and copper content | |
| CN112428025A (en) | Method for constructing two-dimensional wear graph of cutter to optimize safe cutting area | |
| CN110658782A (en) | Characteristic-based numerical control machine tool inherent energy efficiency prediction method | |
| Hossain et al. | Development of an artificial neural network algorithm for predicting the surface roughness in end milling of Inconel 718 alloy | |
| Padhy et al. | Use of multi-objective optimization technique (Taguchi-GRA Approach) in dry hard turning of inconel 625 | |
| CN212858795U (en) | Tool remaining service life prediction device based on vibration signals | |
| Nayak et al. | Performance optimization in electro-discharge machining using a suitable multiresponse optimization technique | |
| Li et al. | Tool Wear Monitoring Technology Based on Deep Learning of Machining State Features | |
| Lu et al. | Prediction model of the surface roughness of micro-milling single crystal copper |
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 |