A method for detecting blade breakage of a dicing saw blade

Abstract

The present invention provides a method for detecting broken blades of dicing machine blades, including the following steps: Step 1: Downsample each sampling point, taking the average with a step size of 2; Step 2: Calculate the median and the maximum value for the obtained average, and calculate the altitude value through the median and the maximum value; Step 3: When it is determined that the altitude value is within a predetermined range, perform filtering processing on the sampling data to obtain a low-pass filter; Step 4: Calculate the kurtosis of the low-pass filter; Step 5: When it is determined that the kurtosis is lower than a preset value, calculate the positions of the peaks and valleys of the low-pass filter; Step 6: Calculate the slope, the rising amplitude, and the falling amplitude of the filter according to the positions of the peaks and valleys; Step 7: Determine the slope, the rising amplitude, and the falling amplitude. When the slope is not greater than the set threshold, the rising amplitude is not less than the rising amplitude limit value, and the falling amplitude is not less than the falling amplitude threshold, it is determined that the blade is damaged.

Description

Technical Field

[0001] The present invention relates to the field of dicing machines, and particularly to a method for detecting broken blades of a dicing machine. Background Art

[0002] A dicing machine is a precision machining device mainly based on the principle of high-speed rotating grinding wheel. When it works, its machining accuracy is affected by the accuracy of the device itself and process parameters. An important factor is the wear and breakage degree of the tool.

[0003] Due to the special working nature of the blade, blade breakage often occurs during the dicing production process. At this time, if the blade continues to cut, it will affect the dicing quality and even damage the cutting piece. In order to ensure the dicing quality of the cutting piece and prevent the cutting piece from being damaged, it is required that the blade stops cutting in time when breakage occurs during the cutting process. Therefore, it is necessary to perform real-time detection on the blade to timely detect the blade breakage situation, stop continuous cutting and replace the tool body. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: aiming at the technical defect of the need to timely detect the blade breakage situation of the dicing machine, a method for detecting broken blades of a dicing machine is provided.

[0005] To solve the above technical problem, the present invention provides a method for detecting broken blades of a dicing machine. The method for detecting broken blades of the dicing machine includes the following steps:

[0006] Step 1: Downsample each sampling point, and take the average with a step size of 2;

[0007] Step 2: Calculate the median and the maximum value of the obtained average value, and calculate the altitude value through the median and the maximum value;

[0008] Step 3: When it is determined that the altitude value is within a predetermined range value, perform filtering processing on the sampling data to obtain a low-pass filter;

[0009] Step 4: Calculate the kurtosis of the low-pass filter;

[0010] Step 5: When it is determined that the kurtosis is lower than a preset value, calculate the positions of the wave peaks and wave valleys of the low-pass filter;

[0011] Step 6: Calculate the slope, the rising amplitude, and the falling amplitude of the filter according to the positions of the wave peaks and wave valleys;

[0012] Step 7: Determine the slope, the rising amplitude, and the falling amplitude. When the slope is not greater than the set threshold value, the rising amplitude is not less than the rising amplitude limit value, and the falling amplitude is not less than the falling amplitude threshold value, it is determined that the blade is broken.

[0013] The method for detecting broken blades of a dicing machine provided by the present invention may also have the following feature: in step three, the low-pass filtering includes performing first-order low-pass filtering, second-order low-pass filtering and first-order low-pass filtering on the sampled data through a digital low-pass filter.

[0014] The method for detecting a broken blade of a dicing machine provided by the present invention may further have the following characteristics: a first-order low-pass filtering formula of a first-order digital low-pass filter is:

[0015] SigA n =a0*Xavg n +(1-a0)*SigA n-1

[0016] Where: SigA n is the primary filter output value at sampling point n, SigA n-1 is the output value of the first filter at the sampling point n-1, a0 is 0.314, Xavg n is the average of n sampling points, and * is multiplication calculation.

[0017] The method for detecting a broken blade of a dicing machine provided by the present invention may also have the following characteristics: the second-order low-pass filter formula is:

[0018]

[0019] Where: SigB n is the secondary filter output value at sampling point n, SigB n-1 is the secondary filtering output value at the n-1 sampling point, SigB n-2 is the secondary filter output value at the n-2 sampling point, b0=Wc 2 *Tsw 2 , a1=4+4*ly*Wc*Tsw+b0, Wc=2*1000*π, Tsw=0.0005, a2=-8+2*b0,a3=4-4*ly*Wc*Tsw+a0,*denotes multiplication calculation.

[0020] The method for detecting a broken blade of a dicing machine provided by the present invention may also have the following characteristics: the first-order low-pass filter formula is:

[0021] SigC n =a4*SigB n +(1-a4)*SigC n-1

[0022] Where: SigC n is the low-pass filter output value at sampling point n, SigC n-1is the low-pass filter output value at the n-1 sampling point, a4 is 0.314, and a is 0.314.

[0023] The method for detecting a broken blade of a dicing machine provided by the present invention may further have the following characteristics: the criteria for judging the peaks and troughs are:

[0024] When SigC n -SigC n-1 ≥0, the stage waveform is rising,

[0025] When SigC n -SigC n-1 <0, the phase waveform is descending;

[0026] ΔSig n =SigC n -SigC n-1

[0027] When ΔSig n-1 >0, and ΔSig n ≤0, judged as a peak,

[0028] When ΔSig n-1 <0, and ΔSig n ≤0, it is judged as a trough.

[0029] The method for detecting a broken blade of a dicing machine provided by the present invention may also have the following characteristics: the formula for kurtosis is:

[0030]

[0031] Where Kurl is the kurtosis, x i is the data value of sampling point i, x med is the median of all sampling point data values, To calculate the sum.

[0032] The method for detecting a broken blade of a dicing machine provided by the present invention may also have the following feature: the slope k is calculated based on the peak P and the trough V1, and the peak P and the trough V2 in the waveform to obtain k1 and k2.

[0033]

[0034] The method for detecting a broken blade of a dicing machine provided by the present invention may also have the following feature: the rising amplitude of the filter is ΔY1, which is the peak value in the rising interval of the waveform minus the starting point of the interval.

[0035] The method for detecting a broken blade of a dicing machine provided by the present invention may also have the following feature: the filtering decrease amplitude ΔY2 is the peak value in the waveform decrease interval minus the interval start point.

[0036] The beneficial effects of the present invention are as follows:

[0037] In the method for detecting broken blades of the dicing saw blades of the present invention, the following steps are included:

[0038] Step 1: Sampling is performed by using the downsampling method, where the average value is taken with a step size of 2 to obtain the average value of each sampling data. The sampling formula is as follows:

[0039] Xavg n-1 =(X 2n-2 +X 2n-1 ) / 2

[0040] In the formula: Xavg n-1 is the sampling number at n - 1, X 2n-2 is the sampling number at 2n - 2, and X 2n-1 is the sampling number at 2n - 1;

[0041] Step 2: Calculate the median and the maximum value for the obtained average value, and calculate the altitude value through the median and the maximum value;

[0042] Step 3: When it is determined that the altitude value is within the predetermined range value, perform filtering processing on the sampling data to obtain low - pass filtering;

[0043] Step 4: Calculate the kurtosis for the low - pass filtering;

[0044] Step 5: When it is determined that the kurtosis is lower than the preset value, calculate the positions of the wave peaks and wave valleys of the low - pass filtering;

[0045] Step 6: Calculate the slope, the rising amplitude, and the falling amplitude of the filtering according to the positions of the wave peaks and wave valleys;

[0046] Step 7: Determine the slope, the rising amplitude, and the falling amplitude. When the slope is not greater than the set threshold value, the rising amplitude is not less than the rising amplitude limit value, and the falling amplitude is not less than the falling amplitude threshold value, it is determined that the blade is damaged;

[0047] Through the above - mentioned method for detecting broken blades, by performing filtering calculations on the collected data, then calculating the wave peaks and wave valleys of the obtained waveform, and using the settings on the wave peaks and wave valleys to calculate and determine whether the blade is damaged. This detection method can send more indistinguishable fuzzy data into the loop detection logic for comprehensive judgment, rather than directly triggering an alarm as damaged data. Therefore, the false alarm rate of this detection method is lower and the detection efficiency is higher. In addition, this detection method can run on platforms with lower configurations such as ARM, significantly reducing the cost. Moreover, by using the method of filtering the data, the interference of environmental impacts on the recognition accuracy of the algorithm is minimized as much as possible, improving the calculation speed. Description of the Drawings

[0048] Figure 1 is a flowchart of the broken tool detection method for the dicing blade in this embodiment;

[0049] Figure 2 is a waveform diagram of the first low-pass filter in this embodiment;

[0050] Figure 3 is a waveform diagram of the second low-pass filter in this embodiment;

[0051] Figure 4 is a waveform diagram of the third low-pass filter in this embodiment;

[0052] Figure 5 is Figure 4 an enlarged schematic view of part A in Detailed Description of the Invention

[0053] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] As Figure 1 shown, the broken tool detection method for the dicing blade includes the following steps:

[0055] Step 1: First, sampling is performed by using the downsampling method. In the sampling of this embodiment, the average value is taken with a step size of 2 to obtain the sampling data of each sampling point. The sampling formula is as follows:

[0056] Xavg n-1 =(X 2n-2 +X 2n-1 ) / 2

[0057] In the formula: Xavg n is the sampling number at the sampling point n - 1, X 2n-2 is the sampling number at the sampling point 2n - 2, X 2n-1 is the sampling number at the sampling point 2n - 1, and n is a natural number;

[0058] Step 2: Calculate the median x med and the maximum value Xavg.max for the sampling average values of all sampling points, and calculate the altitude value through the median and the maximum value:

[0059] Height = Xavg.max - Xavg.med

[0060] Where: Height is the altitude, Xavg.max is the maximum value of all sample averages, and Xavg.med is the median of the sample averages;

[0061] Step 3: Judge the altitude value calculated in step 2. The judgment rule is: set the altitude upper limit value H uplim , lower limit of altitude H low1im When the altitude value Height is not lower than the upper altitude limit, the blade is determined to be broken; when the altitude value Height is not higher than the lower altitude limit, the blade is determined to be normal; when the altitude value Height is between the upper altitude limit and the lower altitude limit, proceed to step 4;

[0062] Step 4: Filter each sample average and perform three low-pass filtering. First, input each sample average into the digital low-pass filter to obtain a filtered output value. The filtering formula is:

[0063] SigA n =a0*Xavg n +(1-a0)*SigA n-1

[0064] Where: SigA n is the primary filter output value at sampling point n, SigA n-1 is the output value of the first filter at the sampling point n-1, a0 is 0.314, Xavg n is the average of n sampling points, * is multiplication calculation;

[0065] Then the obtained primary filter output value is input into the digital low-pass filter again for secondary filtering to obtain the secondary filter output value. The secondary filtering formula is:

[0066]

[0067] Where: SigB n is the secondary filter output value at sampling point n, SigB n-1 is the secondary filtering output value at the n-1 sampling point, SigB n-2 is the secondary filter output value at the n-2 sampling point, b0=Wc 2 *Tsw 2 , a1=4+4*ly*Wc*Tsw+b0, Wc=2*1000*π, Tsw=0.0005, a2=-8+2*b0,a3=4-4*ly*Wc*Tsw+a0, * is multiplication calculation;

[0068] The obtained secondary filtering output value is then input into a digital low-pass filter for tertiary filtering to obtain a tertiary filtering output value. The tertiary filtering formula is as follows:

[0069] SigC n = a4 * SigB n +(1 - a4) * SigC n-1

[0070] In the formula: SigC n is the low-pass filtering output value at the nth sampling point, SigC n-1 is the low-pass filtering output value at the (n - 1)th sampling point, and a4 is 0.314.

[0071] Step Five: Calculate the kurtosis for the waveform obtained after filtering processing. The calculation formula is as follows:

[0072]

[0073] In the formula, Kurl is the kurtosis, x i is the data value at the ith sampling point, x med is the median of all sampling point data values, is the summation calculation.

[0074] Step Six: Make a judgment based on the obtained kurtosis. The judgment rule is: Set the kurtosis limit value K lim . When the kurtosis is greater than or equal to the kurtosis limit value, it is judged that the blade is broken. When the kurtosis is less than the preset value, proceed to Step Seven;

[0075] Step Seven: Based on the waveform obtained by filtering, judge the rising and falling positions of the waveform, and calculate the position of each wave valley and wave peak through the rising waveform and the falling waveform. The judgment calculation formula is as follows:

[0076] When SigC n - SigC n-1 ≥0, the stage waveform is rising,

[0077] When SigC n - SigC n-1 <0, the stage waveform is falling;

[0078] ΔSig n = SigC n - SigC n-1

[0079] When ΔSig n-1 >0, and at the same time ΔSig n ≤0, it is judged as a wave peak,

[0080] When ΔSig n-1 <0, and at the same time ΔSign ≤0, it is judged as a trough.

[0081] Step 8: If Figure 5 As shown, according to the values of the peak P, trough V1 and trough V2 in the waveform, the slope k is calculated to obtain k1 and k2

[0082]

[0083] Where y p is the longitudinal coordinate value of the peak P, is the vertical coordinate value of the trough V1, x p is the transverse coordinate value of the peak P, is the horizontal coordinate value of the trough V1, is the longitudinal coordinate value of the trough V2, is the horizontal coordinate value of the trough V2.

[0084] Step 9: Set the threshold E according to the slopes k1 and k2, |k1+k2|≤E, obtain the filter rise amplitude ΔY1 and rise time ΔX according to the trough V1 and the peak P, and obtain the filter fall amplitude ΔY2 according to the trough V2 and the peak P

[0085] ΔY1=Y P -Y v1

[0086] ΔY2=Y P -Y v2

[0087] ΔX=X v1 -X p

[0088] Set the rising limit value Y up , the limit of the decline is Y down , rise time limit X t , when the conditions are met,

[0089] |k1+k2|≤E, ΔY1≥Y up , ΔY2≥Y down

[0090] Determine if the blade is broken.

[0091] In this embodiment, the threshold value in the above steps is set to E=16, Y up =100,Y down =80, K lim =30, H uplim =30, H lowlim =120. In addition, in step 7, set two valleys V n-1 ,V n , when Vn -V n-1 <4, the trough V n It is a small disturbance and can be ignored.

[0092] In addition, if Figures 2 to 4 As shown, waveform B in the figure is the waveform after filtering.

[0093] like Figure 2 As shown in FIG, the effect after the first filtering of each sample average is shown. According to the attached figure, it can be seen that after a low-pass filtering, the noise and sharp symbols are smoothed, but the broken knife feature is retained.

[0094] like Figure 3 As shown in the figure, it is the effect of the second filtering of the waveform after the first filtering. According to the attached figure, it can be seen that after the waveform is filtered by the second-order low-pass filter, the waveform amplitude is reduced and the waveform is slightly delayed, but it is smoother than the original waveform and the noise amplitude is smaller.

[0095] like Figure 4 As shown, this is the effect of the third filtering of the waveform of the secondary filtering. According to the attached figure, it can be seen that after passing through the digital low-pass filter again, the noise is effectively reduced, the noise near the broken knife feature is less, and the waveform is smoother.

[0096] According to the broken blade detection method in the above embodiment, the collected data is filtered and calculated, and then the peaks and troughs of the obtained waveform are calculated. The settings on the peaks and troughs are used to determine whether the blade is damaged. This detection method can send more difficult-to-distinguish fuzzy data into the loop detection logic for comprehensive judgment, rather than directly triggering an alarm as damaged data. Therefore, the false alarm rate of this detection method will be lower and the detection efficiency will be higher. In addition, this detection method can run on platforms with lower configurations such as ARM, significantly reducing costs. Moreover, by filtering the data, the interference of environmental influences on the algorithm recognition accuracy can be minimized, thereby improving the estimation speed.

[0097] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A method for detecting broken blades of a dicing machine blade, characterized in that, Including: Step 1: Downsample each sampling point, taking the average with a step size of 2; Step 2: Calculate the median and the maximum value for the obtained average values, and calculate the altitude value through the median and the maximum value; Step 3: When it is determined that the altitude value is within the predetermined range, perform filtering processing on the sampling data to obtain a low-pass filter; Step 4: Calculate the kurtosis for the low-pass filter; Step 5: When it is determined that the kurtosis is lower than the preset value, calculate the peak and valley positions of the low-pass filter; Step 6: Calculate the slope, the rising amplitude, and the falling amplitude of the filter according to the peak and valley positions; Step 7: Determine the slope, the rising amplitude, and the falling amplitude. When the slope is not greater than the set threshold, the rising amplitude is not less than the rising amplitude limit value, and the falling amplitude is not less than the falling amplitude threshold, it is determined that the blade is damaged.

2. The method for detecting blade breakage of a dicing saw blade according to claim 1, wherein: In step 3, the low-pass filter includes performing first-order low-pass filtering, second-order low-pass filtering, and first-order low-pass filtering on the sampling data through a digital low-pass filter.

3. The method for detecting blade breakage of a dicing saw blade according to claim 2, wherein: The first-order low-pass filtering formula of the digital low-pass filter is: SigA n = a0 * Xavg n + (1 - a0) * SigA n-1 Where: SigA n is the first filtered output value at the nth sampling point, SigA n-1 is the first filtered output value at the (n - 1)th sampling point, a0 is 0.314, Xavg n is the average value of the nth sampling point, and * represents multiplication calculation.

4. The method for detecting blade breakage of a dicing saw blade according to claim 3, wherein: The second-order low-pass formula is: Where: SigB n is the secondary filtering output value at the nth sampling point, SigB n-1 is the secondary filtering output value at the (n - 1)th sampling point, SigB n-2 is the secondary filtering output value at the (n - 2)th sampling point, b0 = Wc 2 *Tsw 2 , a1 = 4 + 4 * ly * Wc * Tsw + b0, Wc = 2 * 1000 * π, Tsw = 0.0005, a2 = -8 + 2 * b0, a3 = 4 - 4 * ly * Wc * Tsw + a0, * represents multiplication calculation.

5. The method for detecting blade breakage of a dicing saw blade according to claim 4, wherein: The first-order low-pass filtering formula is: SigC n = a4 * SigB n + (1 - a4) * SigC n-1 Where: SigC n is the low-pass filtered output value at the nth sampling point, SigC n-1 is the low-pass filtered output value at the (n-1)th sampling point, and a4 is 0.

314.

6. The method for detecting blade breakage of a dicing saw blade according to claim 5, wherein: The judgment criteria for the peak and valley are: When SigC n -SigC n-1 ≥0, the stage waveform is rising, When SigC n -SigC n-1 < 0, the phase waveform is decreasing; ΔSig n = SigC n - SigC n-1 When ΔSig n-1 > 0, and at the same time ΔSig n ≤ 0, it is determined as a wave crest. When ΔSig n-1 < 0, and at the same time ΔSig n ≤ 0, it is determined as a trough.

7. The method for detecting blade breakage of a dicing saw blade according to claim 1, wherein: The formula for kurtosis is: where Kurl is kurtosis, and x i is the data value at the i-th sampling point, and x med is the median of all sampling point data values, is the summation calculation.

8. The method for detecting blade breakage of a dicing saw blade according to claim 1, wherein: Calculate the slope k according to the peak P and valley V1, and peak P and valley V2 in the waveform, and obtain k1 and k2.

9. The method for detecting blade breakage of a dicing saw blade according to claim 1, wherein: The rising amplitude of the filter ΔY1 is the peak value within the rising interval of the waveform minus the starting point of the interval.

10. The method for detecting blade breakage of a dicing saw blade according to claim 1, wherein: The falling amplitude of the filter ΔY2 is the peak value within the falling interval of the waveform minus the starting point of the interval.

Images