CN116423292A - Milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation - Google Patents
Milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation Download PDFInfo
- Publication number
- CN116423292A CN116423292A CN202310543843.5A CN202310543843A CN116423292A CN 116423292 A CN116423292 A CN 116423292A CN 202310543843 A CN202310543843 A CN 202310543843A CN 116423292 A CN116423292 A CN 116423292A
- Authority
- CN
- China
- Prior art keywords
- frequency
- monitoring
- amplitude
- milling
- peak
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 103
- 238000003801 milling Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 49
- 230000001133 acceleration Effects 0.000 claims abstract description 33
- 230000000737 periodic effect Effects 0.000 claims abstract description 32
- 238000005070 sampling Methods 0.000 claims abstract description 21
- 238000001228 spectrum Methods 0.000 claims abstract description 21
- 239000013598 vector Substances 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims abstract description 8
- 238000001914 filtration Methods 0.000 claims description 13
- 230000003595 spectral effect Effects 0.000 claims description 11
- 230000004927 fusion Effects 0.000 claims description 7
- 230000007717 exclusion Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 abstract description 14
- 230000008859 change Effects 0.000 abstract description 7
- 238000012360 testing method Methods 0.000 description 12
- 238000005520 cutting process Methods 0.000 description 9
- 238000000605 extraction Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000000087 stabilizing effect Effects 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 3
- 238000003754 machining Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000007667 floating Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 230000006641 stabilisation Effects 0.000 description 2
- 238000011105 stabilization Methods 0.000 description 2
- 229910000838 Al alloy Inorganic materials 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 229910045601 alloy Inorganic materials 0.000 description 1
- 239000000956 alloy Substances 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000007704 transition Effects 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
- B23Q17/12—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring vibration
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation, which comprises the steps of firstly, acquiring a spindle three-way vibration signal in a milling process through a three-way vibration acceleration sensor; secondly, the peak value corresponding to the periodic peak and the peak value is found out in the signal frequency spectrum vector by using the peak value searching method, and then the peak value is substituted into a calculation formula to obtain the energy ratio; then calculating an amplitude standard deviation, and fusing the energy ratio extracted from each direction of the three-way vibration signal and the amplitude standard deviation into a characteristic layer to obtain a 6-dimensional monitoring index set; and finally, inputting the index set into a trained monitoring model to output a milling state monitoring result of the sampling time period. Compared with the traditional flutter monitoring method, the method can simultaneously meet three requirements of early monitoring, insensitivity of the provided monitoring index to the change of the processing technological parameters and real-time monitoring, and can accurately identify the amplitude and the frequency of the main flutter frequency.
Description
Technical Field
The invention belongs to the technical field of monitoring of machining states, and particularly relates to the field of on-line monitoring of milling chatter of high-speed milling, in particular to a milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation.
Background
Milling is the most common machining method, has the advantages of high efficiency, high precision, low cost and the like, and is widely applied to the high-end manufacturing fields of aviation, aerospace and the like. Milling chatter is a strong self-excited vibration generated in the milling process, and seriously affects the processing efficiency and the processing quality. Thus, milling chatter online monitoring technology has been a hot topic in high-end manufacturing for many years. The dither signal is a complex signal that is nonlinear, non-stationary, and multi-component, so it is difficult to accurately identify the dither in real time directly from the time domain waveform or spectrum of the sensor signal. For the flutter signals with complex components, the flutter components in the flutter signals are generally required to be separated or identified by utilizing a signal processing means, and then a proper monitoring index is constructed to reflect the change condition of the flutter components in the signals, so that the milling state is monitored on line.
A large number of students at home and abroad study the milling chatter monitoring problem based on the energy proportion of the chatter components in the signals. The collected flutter signals mainly comprise periodic components related to spindle frequency conversion and frequency multiplication thereof and flutter components related to low-order natural frequencies of milling process systems, and to calculate the energy ratio of the flutter components, the periodic components and the flutter components in the signals need to be separated first. At present, two ideas exist in the research, one of the ideas is to directly decompose signals by using various modal decomposition methods, and the component closest to the vibration is screened out from the signals to serve as the vibration component; another idea is to filter out periodic components in the signal indirectly through various filtering methods, and the remaining components in the signal can be regarded as flutter components.
Caliskan, canada, et al uses a kalman filter to separate the stable periodic signal from the raw vibration acceleration data of the spindle and workpiece, and uses a nonlinear Energy operator (Nonlinear Energy Operator, NEO) and an Energy separation algorithm (Energy Separation Algorithm, ESA) to extract an Energy Ratio (ER) in the time domain signal to distinguish between chatter and stability. The method comprises the steps of separating periodic components from a main shaft vibration signal and a sound pressure signal by utilizing a Kalman filter, and detecting 5 states of cutting, feeding, stabilizing, flutter and retracting by adopting a hybrid model combining a physical model based on energy ratio and a neural network model based on short-time Fourier time spectrum. However, the method for extracting the energy ratio is limited by the fact that the Kalman filter cannot adaptively filter, so that the periodic component filtering effect of the signal is affected by the change of the processing technological parameters, the filtering parameters need to be manually adjusted according to the processing technological parameters, and online monitoring cannot be performed under different working conditions.
Based on the existing search literature, the milling chatter monitoring method commonly used at present has the following problems: 1) Early monitoring cannot be achieved; 2) The provided monitoring index is sensitive to the change of the processing technological parameters; 3) It is difficult to meet the real-time requirements of online monitoring.
In view of the foregoing, there is a need for an online milling chatter monitoring method that can simultaneously satisfy the requirements of early monitoring, the proposed monitoring index is insensitive to process parameter changes, and real-time monitoring is performed, so as to be used for online monitoring in actual engineering.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an on-line milling chatter monitoring method integrating the energy ratio and the standard deviation of the amplitude; the original vibration acceleration signal is converted from a time domain to a frequency domain through fast Fourier transform (Fast Fourier Transform, FFT), all the vibration peaks and the periodic peaks are searched in the frequency domain by utilizing the vibration peak searching method, the energy of the vibration peaks is calculated, the energy of the periodic peaks in a frequency spectrum is calculated, the energy ratio value can be obtained according to the energy of the vibration peaks and the energy of the periodic peaks, and then the on-line monitoring of milling vibration can be realized by comprehensively assisting in monitoring the standard deviation value of the index amplitude.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: a milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation comprises the following steps:
and 5, finally, inputting the fused 6-dimensional monitoring index set into a trained three-class monitoring model, and outputting the milling state of the sampling time period.
The flutter peak searching method in the step 2 comprises 5 parts of contents of median filtering, mean filtering, standard deviation filtering, one-dimensional peak searching, threshold value removing and frequency removing which are sequentially carried out.
Before median filtering in step 2, the spectral vector Y is supplemented before and after (n Filter -1)/2 zeros to ensure that a median filtering operation is applicable to the front and back (n) of the spectral vector Y Filter -1)/2 frequencies.
In step 2, the frequency is considered to be greater than the tooth passing frequency ω when performing the chatter peak search t And the amplitude is larger than the flutter peak of the threshold value at the corresponding frequency, and if a plurality of flutter peaks are found between adjacent frequency multiplication of the spindle rotation frequency, only the flutter peak with the largest amplitude is reserved.
All peaks in the signal spectrum are obtained in the frequency domain through a one-dimensional peak finding algorithm, and the frequency is larger than the tooth passing frequency omega t And the peak value with the amplitude larger than the threshold value at the corresponding frequency is reserved, and the spindle rotation frequency omega of which each reserved peak value is nearest to the left and right of the reserved peak value is calculated s And the interval value of frequency multiplication thereof, if the interval value is smaller than the set valueThreshold value, interval value range is [0 ], omega s ) The peak is considered to be a periodic peak.
Through an open numerical control system, the main shaft rotating speed omega and the cutter tooth number N are acquired in real time t The spindle frequency conversion (omega) s =Ω/60) and over-tooth frequency (ω t =N t ω s ) Is calculated by the computer.
The energy ratio calculation formula in step 1 is defined as follows:
wherein N represents the number of the detected flutter peaks; omega CH,i Representing the ith dither frequency; k represents the number of the exclusion frequencies; omega REJ,k Represents the kth rejection frequency, the numerator being the energy of the dither component; the first term of the denominator is the energy of the amplitude at the excluded frequency instead of the periodic component energy, er=0 if no acceptable spectral peak is found from the FFT result, er=1 if no excluded spectral peak is found.
The calculation formula of the standard deviation of the amplitude in the step 3 is as follows:
wherein t is a time sequence number, and the values are 1,2, … and 2n; s and S m Respectively representing the time-domain amplitude vector of the original signal and its mean value.
And 5, the milling state in the step is blank cutting, stabilizing or vibrating.
Compared with the prior art, the invention has at least the following beneficial effects: the energy ratio of the monitoring index directly reflects the energy ratio of the flutter component in the original vibration acceleration signal, so that early monitoring of the flutter can be realized; according to the energy ratio extraction method, a Kalman filter is not needed, the calculation accuracy of the energy ratio is not influenced by the change of the processing technological parameters, and the energy ratio can be monitored under various working conditions; the real-time performance of the invention can meet the requirement of on-line monitoring, and can finish the signal processing of the previous sampling time period and give the monitoring result of the milling processing state before the signal acquisition of the next sampling time period is finished; the method can simultaneously meet three requirements of early monitoring, insensitivity of the provided monitoring index to the change of the processing technological parameters and real-time monitoring, and can accurately identify the amplitude and the frequency of the main vibration frequency.
Drawings
FIG. 1 is a schematic system diagram of the method of the present invention; wherein, 1 represents the headstock, 2 represents three-way acceleration sensor, 3 represents milling cutter, 4 represents data acquisition card, and 5 represents the computer.
FIG. 2 is a graph showing the results of a numerical experiment of the method of the present invention; wherein the abscissa represents the time-domain sample length in terms of points, the left ordinate represents the CPU time in terms of s, the right ordinate represents the ratio of the average CPU time to the sampling time (time ratio), and the dimensionless index has no units.
Fig. 3 is a flow chart of the method of the present invention.
Fig. 4 is a schematic diagram of variable-width processing according to an embodiment of the present invention.
FIG. 5 is a time domain plot of X-direction acceleration data and the variation curves and given monitoring result sequences of the flutter monitoring indicators (ER and sigma (a)) extracted therefrom according to an embodiment of the present invention; wherein fig. 5 (a) is a time domain plot of X-direction acceleration data, and fig. 5 (b) is a variation curve of flutter monitoring indexes (ER and σ (a)) extracted from the X-direction acceleration data and a given monitoring result sequence; the abscissa of fig. 5 (a) and (b) both represent time in s; the ordinate of FIG. 5 (a) shows the vibration signal amplitude in m/s 2 The method comprises the steps of carrying out a first treatment on the surface of the The left ordinate of FIG. 5 (b) represents the standard deviation of the amplitude in m/s extracted from the vibration signal 2 The method comprises the steps of carrying out a first treatment on the surface of the The right ordinate of fig. 5 (b) represents the energy ratio extracted from the vibration signal, and the dimensionless index has no units.
FIG. 6 is a graph showing X-direction, Y-direction and Z-direction acceleration data over 12.44-12.64 s for an embodiment of the present inventionA spectrogram; wherein, fig. 6 (a) is an X-direction acceleration data spectrum diagram; FIG. 6 (b) is a Y-direction acceleration data spectrum; FIG. 6 (c) is a graph of Z-direction acceleration data; the abscissa in FIGS. 6 (a), (b) and (c) each represents frequency in Hz; the ordinate in FIGS. 6 (a), (b) and (c) each represents the vibration signal amplitude in m/s 2 。
FIG. 7 is a confusion matrix diagram of test results obtained by the method of the present invention using two types of milling chatter monitoring model linear support vector machines (Linear Support Vector Machine, LSVM) and limit gradient lifting trees (eXtreme Gradient Boosting, XGBoost) as examples; wherein fig. 7 (a) is a confusion matrix diagram of the test results of the LSVM; fig. 7 (b) is a confusion matrix plot of the test results of XGBoost.
Detailed Description
A schematic system diagram of the method of the present invention is shown in fig. 1.
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the present invention, but are not intended to limit the invention in any way. It should be noted that variations and modifications could be made by those skilled in the art without departing from the inventive concept. These are all within the scope of the present invention.
A milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation comprises the following steps:
step 1: acquiring real-time vibration signals of a spindle in the milling process through a three-way vibration acceleration sensor arranged at the end part of the spindle box;
step 2: obtaining a frequency spectrum of a real-time vibration signal of a main shaft through fast Fourier transform, converting the signal from a time domain to a frequency domain, searching all the dither peaks and the periodic peaks in the frequency domain by using a dither peak searching algorithm, and calculating the energy ratio of the dither peaks and the periodic peaks according to the energy of the dither peaks and the periodic peaks to obtain the energy ratio value of the dither peaks and the periodic peaks in a sampling time period; the method specifically comprises the following steps:
all peaks in the signal spectrum are obtained in the frequency domain through a one-dimensional peak finding algorithm, and the frequency is larger than the tooth passing frequency (omega t ) And amplitude valuePeak retention greater than the threshold at the corresponding frequency, threshold a at each frequency th (i) The calculation formula of (2) is as follows:
median(Y)(i)=median({Y(k)})
a th (i)=σ(Y)+median(Y)(i)+Y m
wherein i is a frequency sequence number, and the values are 1,2, … and n; k is the frequency number in the median filter window and takes the value of i- (n) Filter -1)/2,i-(n Filter -1)/2+1,…,i+(n Filter -1)/2; y represents that the FFT result of the original signal is [ n×1 ]]Magnitude vectors at the frequencies; y is Y m Sigma (Y) and median (Y) (i) represent the mean, standard deviation and i-th median filter value, respectively, of the FFT result of the original signal, where the median filter window size n Filter =25。
Then, the spindle rotation frequency (ω) at which each of the held peaks is nearest to the left and right thereof is calculated s ) And the interval value of frequency multiplication, if the interval value is smaller than the set threshold value, the peak value is considered to belong to the periodic peak, otherwise, the peak value is considered to belong to the flutter peak. As an example, the set threshold is 0.1 omega s 。
The energy ratio of the flutter peak and the periodic peak is defined as follows:
wherein N represents the number of the detected flutter peaks; omega CH,i Representing the ith dither frequency; k represents the number of the exclusion frequencies; omega REJ,k Indicating the kth exclusion frequency. The molecules are the energy of the flutter component; the first term of the denominator is the energy of the amplitude at the rejection frequency instead of the periodic component energy. In addition, e.g.If no acceptable spectral peak is found from the FFT result, er=0, no exclusionary spectral peak is found, er=1.
wherein t is a time sequence number, and the values are 1,2, … and 2n; s and S m Respectively representing the time-domain amplitude vector of the original signal and its mean value.
and 5, finally, inputting the fused 6-dimensional monitoring index set into a trained three-class monitoring model, and outputting milling states in a sampling time period, wherein the milling states are blank cutting, stable or flutter.
The time complexity is a key parameter for evaluating the real-time performance of the monitoring method, and the time consumption of the algorithm is mainly concentrated on the monitoring index extraction part (extracting ER and sigma (a)), so that the invention analyzes the theoretical time complexity of the monitoring index extraction part.
(1) Temporal complexity of extraction of ER
The procedure in this section includes a fast fourier transform (O (nlogn)), a flutter peak search algorithm (median filter (O (n) 2 logn)), the mean value (O (N)) and standard deviation (O (2N)) of the spectral magnitudes, the search peak value (O (logn)), the threshold exclusion (O (N)), the frequency exclusion (O (n× (n+k))), and the calculation of ER (O (2N)), the total time complexity of extracting ER is:
T ER =O(nlogn)+O(n 2 logn)+O(n)+O(2n)+O(logn)+O(n)+O(n×(N+K))+O(2N)
(2) Extracting time complexity of sigma (a)
The procedure in this section only finds the standard deviation of the time domain signal amplitude (O (4 n)), so the total time complexity of extraction is:
T σ(a) =O(4n)
because of O (n) 2 logn) is much larger than the other parts, and in summary, the time complexity of the available monitoring index extraction part is:
T≈O(n 2 logn)
from this, the time complexity of the whole chatter monitoring algorithm should be nonlinear, according to O (n 2 logn) is mixed and increased along with the frequency domain sample length n, and because the time domain sample length L is twice (linear relation) the frequency domain sample length n, the time complexity of the whole flutter monitoring algorithm still keeps the mixed and increased relation of the square and the logarithm along with the time domain sample length L, and the increase is characterized by being slow and fast. Consider sliding time window parameters: time window duration (t) w =L/f s =2n/f s ) And time window interval duration (t s ) Using f p Representing the floating point operand per second for the processor, the processing time of the signal segment in the sliding window may be expressed as:
the actual duration of the update signal is t s For real-time systems, the processing time is at most equal to the update sampling time, i.e.:
t cpu ≤t s ≤t w
when t w =t s At the same time, a critical floating point per second operand (the critical floating-point operations per second, FLOPS) f can be obtained c :
If the time domain sample length l=2048 is taken, the sampling frequency f s =10240 Hz, spindle rotation frequency ω s Frequency of teeth passing ω = 138.33Hz t = 276.66Hz, f c =[1024×log 2 (1024)×10240]/2FLOPS=0.0524GFLOPS。
The test platform is configured as follows: windows10×64, intel (R) Core (TM) i5-6200U CPU@2.30GHz (2C, 4T,2.7GHz,2.8GHz, IMC,2× 256kB L2,3MB L3,100MHz FSB), 16g DDR3, and SiSoftware Sandra Professional Home 2018 software, the single Core computing capability f of the platform under test is obtained by performance testing of an aggregation algorithm p 10.84GFLOPS, and the theoretical calculated critical value f c 0.0524GFLOPS, meets f p >f c Real-time requirements of (2).
For each time domain sample length, many different types of samples were used to test the CPU time running on a single kernel (the CPU time was for the entire chatter monitoring algorithm, including spindle vibration data reading, sample construction, monitoring index extraction, and monitoring model prediction), and an average curve was plotted, with the numerical experimental test results shown in fig. 2.
As is clear from fig. 2, the average CPU time line is below the sampling time line, and there is a feasible region (such as the cross-sectional line region in fig. 2) which is formed by the region sandwiched by the sampling time line (such as the double-dashed line in fig. 2) and the CPU time upper limit (such as the dashed line in fig. 2), which illustrates that the present invention can realize real-time monitoring.
When the time-domain sample length is too short, there is an unstable region, so the time-domain sample length is not preferable to be too short. Meanwhile, as the length of the time domain sample increases, the lower bound of the time window interval length increases, and the feasible domain shows a trend of increasing before decreasing, which is consistent with the theoretical analysis, in addition, the CPU time margin of the algorithm is also noted to be large, and the configuration of the test platform can be completely reduced.
Examples:
the main flow of the milling chatter on-line monitoring method integrating the energy ratio and the amplitude standard deviation of the main shaft vibration acceleration signal is shown in figure 3.
On-line monitoring of chatter in the high-speed milling process is carried out on a 7050 aviation aluminum alloy thin-wall plate, the sampling frequency is 10240Hz, a 3-edge hard alloy end mill is adopted as a cutter, the diameter of the cutter is 20mm, the helix angle is 45 degrees, the length of the cutter body is 104mm, and the suspension length of the cutter is 75mm during clamping. Bao Biban is clamped on a workbench through a vice, the thickness is 40mm, and cutting parameters are as follows: the spindle rotation speed is 6000r/min, the cutter feed rate is kept constant at 240mm/min in the milling process, the cutter is reversely milled along the feed direction in fig. 4, the radial cutting width is continuously increased from 0mm to 15mm, and the milling process is dry cutting.
(1) Acquisition of signals
Vibration information during milling is collected through three-way vibration acceleration sensors (the sensitivity of X direction, Y direction and Z direction is respectively 50.2mV/g, 48.6mV/g and 48.5 mV/g) arranged at the end part of the spindle box, and the collected X-direction vibration acceleration signals are shown in fig. 5 (a). As can be seen from FIG. 5 (a), the cutter is in idle state at the stage of 0-0.4 s, and the signal amplitude is very small. The tool enters a milling state from idle after 0.4s, and the amplitude of the signal increases steeply at the moment of 10.95 s. After 14.7s the tool was completely withdrawn from the workpiece, the amplitude of the signal was rapidly reduced.
Fig. 6 is a spectrum of X-, Y-and Z-direction acceleration data within 12.44-12.64 s of the vibration acceleration signal of this case, wherein the spectrum of X-direction acceleration data (e.g. fig. 6 (a)) shows that the chatter peak is dominant in the spectrum, the spectrum of Y-direction acceleration data (e.g. fig. 6 (b)) has chatter peaks that occur but do not dominant, and the spectrum of Z-direction acceleration data (e.g. fig. 6 (c)) has small amplitude, and the reason for this may be that the vibration caused by chatter is mainly distributed in the machining plane (xoy plane) and the Z-direction is along the arbor direction, so that the component of the vibration acceleration caused by chatter in the Z-direction is very small. According to the frequency spectrum of the X-direction acceleration data and the comprehensive judgment of the vibration patterns of the machined workpiece surface, the milling process can be considered to generate vibration after 10.95 s.
(2) Extraction of monitoring indicators
Taking 2048 sampling points as a sample, carrying out sliding time window sampling interception on X-direction, Y-direction and Z-direction acceleration data, extracting an energy ratio and an auxiliary monitoring index amplitude standard deviation from the X-direction, Y-direction and Z-direction acceleration data, wherein the change curve of the flutter monitoring index (ER and sigma (a)) of the extracted X-direction acceleration data and a given monitoring result sequence are shown as a figure 5 (b), and the coding meanings of the monitoring results in the figure are respectively: 0 denotes a cut, 1 denotes a steady state, 2 denotes a chatter (the monitoring result and ER share one vertical axis).
From the monitoring result sequence in fig. 5 (b), the method of the present invention can monitor the generation of the chatter in the early stage (transition stage) of the chatter and give accurate early warning, i.e., the method of the present invention can realize early monitoring.
Because two monitoring indexes can be extracted from each direction of acceleration data, the invention considers that after feature layer fusion is carried out on the two monitoring indexes extracted from each direction of acceleration data, 6 monitoring index sets obtained by fusion are used as the input of a monitoring model, so that the operation is more convenient in application, rather than respectively taking the two monitoring indexes extracted from each direction of acceleration data as the input of the monitoring model, and decision layer fusion is carried out after the monitoring result of each direction of acceleration data is obtained.
(3) Monitoring of chatter conditions
And finally, inputting the monitoring index set subjected to feature layer fusion into a trained monitoring model, outputting a milling state monitoring result in a sampling time period by the monitoring model, and if the result is blank or stable, carrying out early warning, and if the result is flutter, carrying out early warning through a software platform. According to the invention, two machine learning models, namely a linear support vector machine and a limit gradient lifting tree, are selected as milling flutter monitoring models, the effectiveness of the monitoring index provided by the invention is verified through the test results of the two machine learning models, and the two machine learning models are not used for limiting the invention.
Training, verifying and testing the two types of monitoring models by utilizing a large number of index data sets containing three types of samples of air cutting, stabilizing and flutter, wherein test results show that the following conclusions can be drawn about LSVM and XGBoost constructed by the energy ratio and amplitude standard deviation extracted by the method: the average monitoring accuracy of LSVM on the air cutting, stabilizing and vibrating is 92.59%, wherein the monitoring accuracy of each of the air cutting, stabilizing and vibrating is 100.00%, 77.78% and 100.00% (as in fig. 7 (a)); the average accuracy of XGBoost monitoring for air-cut, stabilization and chatter was up to 87.58%, where the accuracy of monitoring for air-cut, stabilization and chatter was 100.00%, 68.94% and 93.79%, respectively (as in fig. 7 (b)).
In summary, the invention provides an on-line milling chatter monitoring method integrating the energy ratio and the amplitude standard deviation; firstly, converting an original vibration acceleration signal from a time domain to a frequency domain through fast Fourier transform (Fast Fourier Transform, FFT), secondly, searching all the vibration peaks (spectrum peaks related to low-order natural frequencies of a milling process system) and periodic peaks (spectrum peaks related to spindle frequency conversion and frequency multiplication thereof) in the frequency domain by utilizing the vibration peak searching algorithm provided by the invention, thirdly, calculating the energy of the vibration peaks, calculating the energy of the periodic peaks in the frequency spectrum, finally, substituting the energy of the vibration peaks and the energy of the periodic peaks into an energy ratio calculation formula constructed by the invention to obtain energy ratio values, and comprehensively assisting in monitoring the standard deviation values of the index amplitude values to realize on-line monitoring of the vibration.
Claims (9)
1. The milling chatter on-line monitoring method integrating the energy ratio and the amplitude standard deviation is characterized by comprising the following steps of:
step 1, acquiring real-time vibration signals of a spindle in the milling process through a three-way vibration acceleration sensor arranged at the end part of a spindle box;
step 2, obtaining a frequency spectrum of a real-time vibration signal of a main shaft through fast Fourier transform, converting the signal from a time domain to a frequency domain, searching all the dither peaks and the periodic peaks in the frequency domain by using a dither peak searching method, and calculating the energy ratio of the dither peaks and the periodic peaks according to the energy of the dither peaks and the periodic peaks to obtain the energy ratio value of the dither peaks and the periodic peaks in a sampling time period;
step 3, calculating the amplitude standard deviation value of the sampling time period by using the time domain amplitude of the vibration signal;
step 4, carrying out feature layer fusion on the energy ratio value and the amplitude standard deviation value of each direction of the three-way vibration signal to obtain a 6-dimensional monitoring index set;
and 5, finally, inputting the fused 6-dimensional monitoring index set into a trained three-class monitoring model, and outputting the milling state of the sampling time period.
2. The method for on-line monitoring milling vibration according to claim 1, wherein the method for searching vibration peaks in step 2 comprises 5 parts of content of median filtering, mean filtering, standard deviation filtering, one-dimensional peak searching, threshold value removing and frequency removing which are sequentially carried out.
3. The method for on-line monitoring milling chatter as defined in claim 2, wherein before the median filtering in step 2, the spectral vectors Y are supplemented before and after each other (n Filter -1)/2 zeros to ensure that a median filtering operation is applicable to the front and back (n) of the spectral vector Y Filter -1)/2 frequencies.
4. The method for on-line monitoring milling vibration according to claim 2, wherein in step 2, the vibration peak search is performed by considering the frequency greater than the tooth passing frequency ω t And the amplitude is larger than the flutter peak of the threshold value at the corresponding frequency, and if a plurality of flutter peaks are found between adjacent frequency multiplication of the spindle rotation frequency, only the flutter peak with the largest amplitude is reserved.
5. The milling chatter on-line monitoring method of integrated energy ratio and amplitude standard deviation as set forth in claim 2, wherein all peaks in the signal spectrum are obtained in the frequency domain by a one-dimensional peak finding algorithm, and the frequency is greater than the tooth passing frequency ω t And the peak value with the amplitude larger than the threshold value at the corresponding frequency is reserved, and the spindle rotation frequency omega of which each reserved peak value is nearest to the left and right of the reserved peak value is calculated s And the interval value of frequency multiplication thereof, if the interval value is smaller than the set threshold value, intervalThe value range is [0 ], omega s ) The peak is considered to be a periodic peak.
6. The on-line monitoring method for milling chatter of integrated energy ratio and amplitude standard deviation as defined in claim 5, wherein the main shaft rotation speed Ω and the number of teeth of the cutter N are collected in real time by opening a numerical control system t The spindle frequency conversion (omega) s =Ω/60) and over-tooth frequency (ω t =N t ω s ) Is calculated by the computer.
7. The method for on-line monitoring milling chatter according to claim 1, wherein the energy ratio calculation formula in step 1 is defined as follows:
wherein N represents the number of the detected flutter peaks; omega CH,i Representing the ith dither frequency; k represents the number of the exclusion frequencies; omega REJ,k Represents the kth rejection frequency, the numerator being the energy of the dither component; the first term of the denominator is the energy of the amplitude at the excluded frequency instead of the periodic component energy, er=0 if no acceptable spectral peak is found from the FFT result, er=1 if no excluded spectral peak is found.
8. The on-line monitoring method for milling chatter according to claim 1, wherein the amplitude standard deviation in the step 3 is calculated as follows:
wherein t is a time sequence number, and the values are 1,2, … and 2n; s and S m Respectively representing the time-domain amplitude vector of the original signal and its mean value.
9. The method for on-line monitoring milling chatter as defined in claim 1, wherein the milling condition in step 5 is a blank cut, a stable or a chatter.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310543843.5A CN116423292A (en) | 2023-05-15 | 2023-05-15 | Milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310543843.5A CN116423292A (en) | 2023-05-15 | 2023-05-15 | Milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116423292A true CN116423292A (en) | 2023-07-14 |
Family
ID=87084052
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310543843.5A Pending CN116423292A (en) | 2023-05-15 | 2023-05-15 | Milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116423292A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116755414A (en) * | 2023-08-22 | 2023-09-15 | 山东新巨龙能源有限责任公司 | Ore mining equipment supervision system based on Internet of things |
CN117708721A (en) * | 2023-12-13 | 2024-03-15 | 西南交通大学 | Winding loosening deformation fault identification method for FFT (fast Fourier transform) decomposition of transformer vibration signal |
-
2023
- 2023-05-15 CN CN202310543843.5A patent/CN116423292A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116755414A (en) * | 2023-08-22 | 2023-09-15 | 山东新巨龙能源有限责任公司 | Ore mining equipment supervision system based on Internet of things |
CN116755414B (en) * | 2023-08-22 | 2023-11-07 | 山东新巨龙能源有限责任公司 | Ore mining equipment supervision system based on Internet of things |
CN117708721A (en) * | 2023-12-13 | 2024-03-15 | 西南交通大学 | Winding loosening deformation fault identification method for FFT (fast Fourier transform) decomposition of transformer vibration signal |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116423292A (en) | Milling chatter on-line monitoring method integrating energy ratio and amplitude standard deviation | |
Yesilli et al. | On transfer learning for chatter detection in turning using wavelet packet transform and ensemble empirical mode decomposition | |
Shah et al. | Tool wear prediction in face milling of stainless steel using singular generative adversarial network and LSTM deep learning models | |
CN109158953B (en) | A kind of cutting-tool wear state on-line monitoring method and system | |
Chen et al. | Development of a novel online chatter monitoring system for flexible milling process | |
CN111687689A (en) | Cutter wear state prediction method and device based on LSTM and CNN | |
CN108982107B (en) | It is a kind of based on morphology and it is multiple dimensioned arrangement entropy mean value bearing fault quantify trend diagnosis method | |
CN110355608A (en) | Based on the tool abrasion prediction technique from attention mechanism and deep learning | |
Chen et al. | Chatter detection for milling using novel p-leader multifractal features | |
CN108629864B (en) | Vibration-based electric spindle radial precision characterization method and system | |
CN110928237B (en) | Vibration signal-based numerical control machining center flutter online identification method | |
Wenping et al. | Vibration signal-based chatter identification for milling of thin-walled structure | |
CN109605128B (en) | Milling chatter online detection method based on power spectrum entropy difference | |
CN112098088A (en) | Rolling bearing fault diagnosis method based on KICA-fractal theory | |
Han et al. | ESPRIT-and HMM-based real-time monitoring and suppression of machining chatter in smart CNC milling system | |
CN116861313A (en) | Kalman filtering working condition identification method and system based on vibration energy trend | |
Liu et al. | A chatter detection method in milling based on gray wolf optimization VMD and multi-entropy features | |
Elbestawi et al. | Process monitoring in milling by pattern recognition | |
CN112846938B (en) | Main shaft rotation precision degradation traceability system under cutting working condition | |
CN114638265B (en) | Milling chatter judging method based on signal convolution neural network | |
Zhao et al. | Tool wear monitoring based on an improved convolutional neural network | |
CN115422978A (en) | Workpiece surface roughness prediction method | |
Benkedjouh et al. | Tool condition monitoring based on mel-frequency cepstral coefficients and support vector regression | |
CN109389039A (en) | A kind of optimization core extreme learning machine on-line monitoring cutting-tool wear state method and system | |
Sichen et al. | Early chatter identification based on optimized VMD with multi-band information fusion and compression method in robotic milling process |
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 |