Multi-feature error laser displacement sensor quality detection method and device

Abstract

The invention relates to the technical field of laser displacement sensor manufacturing, in particular to a multi-characteristic-error laser displacement sensor quality detection method and device capable of efficiently and rapidly completing quality detection of a correlation type laser displacement sensor, and the method comprises the following steps: directly irradiating a light beam emitted by a sensor onto a measuring plate to obtain a diffraction light beam; the diffracted beam is converted into a waveform signal, the waveform signal comprises n peak groups, each peak group comprises two edges, and n is larger than or equal to 1; calculating peak group parameters and errors, comparing the obtained errors with a preset error threshold value, when the errors are smaller than the preset error threshold value, judging that the quality of the device is qualified and recording parameters in the process, and when the errors are larger than the preset error threshold value, shortening the length of the preset weight coefficient table and recalculating the errors. And when the error cannot be smaller than the preset error threshold value all the time, judging that the quality of the device is unqualified, and carrying out optical path and hardware debugging on the laser displacement sensor device again.

Description

Technical Field:

[0001] The present invention relates to the technical field of manufacturing laser displacement sensors, and specifically to a method and device for detecting the quality of a laser displacement sensor with multi-feature errors that can efficiently and quickly complete the quality detection of a transmissive laser displacement sensor. Background Art:

[0002] The transmissive laser displacement measuring instrument uses an incident laser beam, an image sensor and the surface of the object to be measured to form a transmissive projection optical path. When the laser beam is projected onto the surface of the object to be measured, the reflected laser spot will be captured by the image sensor. By analyzing the characteristic information such as the position and shape of the laser spot on the image sensor, and combining the known projection angle of the laser beam and the parameters of the image sensor, the occlusion geometric relationship of the object to be measured on the optical path can be calculated, so as to obtain the required displacement measurement information. It has the advantages of high precision, non-contact measurement, real-time performance, flexibility and strong anti-interference ability, and plays an important role in the fields of industrial automation, intelligent manufacturing, quality inspection, scientific research experiments, etc. In the future, with the continuous progress of technology and the continuous expansion of application fields, the transmissive laser displacement measuring instrument will develop in the directions of miniaturization, intelligence, integration and high precision.

[0003] During the mass production process of the transmissive laser displacement measuring instrument, in order to ensure the consistency of the measurement accuracy of the laser displacement sensor and improve the production efficiency of the sensor, it is usually necessary to detect the quality of the sensor and quickly identify sensors with poor accuracy and unqualified quality. At present, in order to improve the production efficiency of the laser displacement sensor, a linearity measurement method is proposed in the existing solutions. By using the waveform signal generated by the sensor irradiating an equal-width occluding object, the quality of the sensor is detected by judging the linearity between the measured values of multiple occlusion positions and the actual values. Since the measurement speed of the linearity of the sensor is fast and the measurement cost is low, the production efficiency of the sensor is improved. However, this solution can only measure the linearity of each position within the measurement range of the sensor, and cannot determine the measurement error and measurement accuracy of equal-width occlusion at each position. In subsequent production, there will be a situation where the linearity measurement of the sensor passes, but the measurement accuracy at different occlusion positions varies greatly. When the customer uses it, it is impossible to ensure that the measurement accuracy of the object at each position meets the production conditions. Summary of the Invention:

[0004] The present invention aims at the disadvantages and deficiencies existing in the prior art, and proposes a method and device for detecting the measurement accuracy of a sensor and quickly identifying products with poor accuracy and unqualified quality, which can detect multi-feature errors of a laser displacement sensor.

[0005] The present invention is achieved by the following measures:

[0006] A method for quality detection of a laser displacement sensor with multi-feature error, characterized by including the following steps:

[0007] Step 1: Direct the beam emitted by the sensor onto the measurement plate to obtain a diffracted beam, and convert the diffracted beam into a waveform signal. The waveform signal contains n peak groups, and each peak group contains two edges, where n≥1;

[0008] Step 2: Calculate the parameters and errors of the peak groups, and the steps are as follows:

[0009] Step 2-1: Calculate the left and right edge distances and edge positions of each peak group. The distance data is recorded as variable D = [d1, d2, ……, d n , the left edge position is recorded as variable L = [l1, l2, ……, l n , the right edge position is recorded as variable R = [r1, r2, ……, r n . After obtaining the left and right edge positions, take the midpoint position of the left and right edge distances as the center position of each peak group. The center position data is recorded as variable Q = [q1, q2, ……, q n . Record the actual positions of multiple peak groups in the preset direction, which is recorded as variable X = [x1, x2, ……, x n ;

[0010] Step 2-2: Obtain the width standard data of the measurement plate occlusion, and calculate the deviation of D based on this. The error data is recorded as variable F = [f1, f2, ……, f n ;

[0011] Step 2-3: Superimpose the error data F with the position feature X in the preset direction to generate a new feature K;

[0012] Step 2-4: Perform fitting or regression processing on variables X, Q, L, R, and K, and calculate the corresponding errors;

[0013] Step 3: Compare the error obtained in Step 2 with the preset error threshold. When the error is less than the preset error threshold, it is determined that the device quality is qualified, and the parameters in the process are recorded. When the error is greater than the preset error threshold, shorten the length of the preset weight coefficient table, repeat Step 2, recalculate the error and execute Step 3. When repeating N times, N≥3, and the error still cannot be less than the preset error threshold, it is determined that the device quality is unqualified, and the optical path and hardware of the laser displacement sensor device are re-debugged.

[0014] In Step 3 of the present invention, when the error is greater than the preset error threshold, the method for shortening the preset weight coefficient table is as follows:

[0015] Step 3-1: Let the error be E, and the preset error threshold be T. According to the magnitude of the error value E, calculate the adjustment ratio P of the weight coefficient table, where P = (E - T) / T * α, and α is the adjustment coefficient, which can be set according to the actual situation, with 0.3 ≤ α ≤ 0.9;

[0016] Step 3-2: According to the adjustment ratio P, shorten the length of the preset weight coefficient table. The minimum length of the preset weight coefficient table is 3, and the length L of the new weight coefficient table new = L old ×(1 - P), where L old is the length of the original weight coefficient table.

[0017] In step 1 of the present invention, the calculation methods for the two edges of the peak group in the peak group are as follows:

[0018] Step 1-1: Let the one-dimensional waveform signal be Y, that is, Y = [y1, y2, …, y n , perform signal preprocessing on Y, and based on the preset weight coefficient table W = [w1, w2, ……, w k , with the length set to k, perform smoothing processing on the sequence Y based on the sliding window and the weight coefficient table;

[0019] Step 1-2: For each data point y i in Y, take a subsequence with an odd length k centered on this point. The subsequence is represented as Y i:i+k-1 = [y i-k / / 2 , yi -k / / 2+1 , …, y i , …, y i+k / / 2 , where k / / 2 represents the integer part of k divided by 2;

[0020] Step 1-3: Sort the subsequence Y i:i+k-1 to obtain Y sorted,i:i+k-1 = [y (1),i , y (2),i , …, y (k),i , where y (j),i represents the j-th smallest element after sorting;

[0021] Step 1-4: Multiply the sorted subsequence by the weight coefficient table W and sum to obtain the processed data point y i′ , and the formula is:

[0022]

[0023] Secondly, obtain the change rate around each point of the sequence Y′, perform convolution processing on the signal sequence Y′, with the convolution kernel E = [-1, 1], to obtain a new sequence C, where each element c i is the dot product of the convolution kernel and the subsequence at the corresponding position in the signal:

[0024] c i = (E·Y i:i+1 ) = y i+1 -y i For i = 2, 3, …, n - 1;

[0025] Step 1 - 5: Sum the consecutive sequence points greater than 0 in sequence C, set a threshold 1, and take the position of the last point of the sequence exceeding the threshold 1 as the initial right edge. Similarly, sum the consecutive sequence points less than 0, set a threshold 2, and take the position of the last point of the sequence exceeding the threshold 2 as the initial left edge. Replace the noise data between the previous initial right edge and the next initial left edge with edge value data to eliminate interference. Then, set an edge threshold to obtain the final positions of the left and right edges.

[0026] In step 2 - 3 of the present invention, the method for generating the new feature K is as follows: Normalize the position data X and the calculated error data F of each occlusion along the preset direction respectively, so that the two variables become the same order of magnitude, and then add the two variables to obtain the new feature K.

[0027] In the present invention, when fitting processing is adopted in step 2 - 4, it specifically includes the following steps:

[0028] Step 2 - 4 - 1: Perform linear fitting on variable X with variable Q, variable L, variable R, and variable K respectively;

[0029] Step 2 - 4 - 2: Calculate four fitting errors, and superimpose the errors to obtain the corresponding error value.

[0030] In step 2 - 2 of the present invention, the calculation method of the error is as follows: Obtain the actual value of the width, subtract this actual value from each calculated distance result in variable D to obtain the deviation value of variable D, which is re - denoted as variable F, and then take the absolute value of F.

[0031] The present invention also proposes a device for a quality detection method of a laser displacement sensor based on the multi - feature error as described above. It is provided with a measurement platform. It is characterized in that two laser displacement sensors and a controllable linear slide rail are horizontally placed on the measurement platform. There is a slider on the linear slide rail that is matched with the slide rail. There is a groove for fixing the measurement plate in the middle of the slider. The measurement plate is placed in the groove and is automatically moved to the middle of the two laser displacement sensors and fixed by the linear slide rail. There are two or more equally - wide occlusions arranged along the preset direction and parallel to each other on the measurement plate. Each occlusion has two edges arranged oppositely along the preset direction. The two laser displacement sensors are arranged in a counter - shooting manner onto the measurement plate to obtain diffracted light beams.

[0032] The present invention processes waveform signals of multiple peak groups through a sliding window and an adjustable weight coefficient table, calculates the left and right edges and the width of occlusion (i.e., the distance between the left and right edges), which serves as the basic feature of the peak group; combines the position information with the occlusion width error data, while detecting the linearity of the peak group, also considering the error magnitude or consistency of the width data measured by the peak group at different positions; fits or regresses the peak group position information with multiple peak group feature variables to obtain the error, and judges the error of equal-width occlusion irradiated by the sensor at each position based on the magnitude of the error, so as to determine whether the measurement accuracy of the sensor meets the conditions; compared with the prior art, the method proposed by the present invention can not only measure the linearity of the sensor, but also determine the consistency of the measurement accuracy of the sensor for objects at different positions and the magnitude of the measurement error. At a relatively low measurement cost, it can comprehensively evaluate the quality of the sensor, greatly improving the production efficiency of the sensor. BRIEF DESCRIPTION OF THE DRAWINGS:

[0033] Attached Figure 1 is a flowchart of the present invention.

[0034] Attached Figure 2 is a schematic diagram of data processing in an embodiment of the present invention.

[0035] Attached Figure 3 is a schematic diagram of an equal-width occlusion measurement plate used in an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION:

[0036] The present invention will be further described below in conjunction with the drawings and embodiments.

[0037] Embodiment:

[0038] This example proposes a device and method based on multi-feature error that can efficiently and quickly detect whether the quality of a transmissive laser displacement sensor product is qualified.

[0039] In this example, the measuring device consists of two laser displacement sensors and a measuring plate. Two laser displacement sensors are horizontally placed on the measuring platform, and the measuring plate is placed in the middle of the two sensors. The two sensors shoot at each other to measure the size of the measuring plate placed in the middle. The measuring plate has a plurality of equal-width occlusions arranged along a preset direction and parallel to each other. Each occlusion has two edges arranged oppositely along the preset direction. The light beam emitted by the sensor is directly irradiated onto the measuring plate to obtain a diffracted light beam, and the diffracted light beam is converted into a one-dimensional waveform signal. This waveform signal contains n peak groups, and each peak group contains left and right edges.

[0040] Next, the characteristic parameters and errors of the peak group are calculated to determine whether the quality of the device is qualified. The steps are as follows:

[0041] Step (1): Calculate the left and right edge distances and the edge distance positions of each peak group. Denote the distance data as variable D = [d1, d2, ……, dn], the left edge position as variable L = [l1, l2, ……, ln], and the right edge position as variable R = [r1, r2, ……, rn].

[0042] Step (2): After obtaining the left and right edge positions, take the midpoint position of the left and right edge distances as the center position of each peak group. Denote the center position data as variable Q = [q1, q2, ……, qn].

[0043] Step (3): Record the actual positions of multiple peak groups in the preset direction, denoted as variable X = [x1, x2, ……, xn].

[0044] Step (4): Obtain the standard data of the width blocked by the measurement plate to calculate the deviation of D. Denote the error data as variable F = [f1, f2, ……, fn]. Specifically: obtain the actual value of the width, subtract this actual value from each calculated distance result in variable D to get the deviation value of variable D, and re-denote it as variable F. Then, take the absolute value of F.

[0045] Step (5): Superimpose the error data F along the preset direction with the position feature X to generate a new feature K. Specifically: standardize the position data X of each occlusion and the calculated error data F along the preset direction respectively to make the two variables of the same order of magnitude, and then add the two variables to get the new feature K.

[0046] Step (6): Perform regression on variables X, Q, L, R, and K, and calculate their regression errors.

[0047] Step (7): Compare the regression error with the preset error threshold. When the error is greater than the preset error threshold, recalculate the edge distance and the edge distance position of each peak group, repeat Steps 2 - 6, and recalculate the error. When the error is less than the preset error threshold, it is determined that the device quality is qualified, and record the parameters in the process. When after repeating N times (N ≥ 3), the error still cannot be less than the preset error threshold, it is determined that the device quality is unqualified, and re-debug the functions such as the optical path and hardware of the laser displacement sensor device.

[0048] The calculation methods for the two edges of the peak group in Step (1) of this example are as follows:

[0049] Step a: Let the one-dimensional waveform signal be Y, that is, Y = [y1, y2, …, yn]. Perform signal preprocessing on Y. Based on the preset weight coefficient table W = [w1, w2, ……, wk] with a length of k, perform smoothing processing on the sequence Y based on the sliding window and the weight coefficient table.

[0050] Step b: For each data point yi in Y, take a subsequence of odd length k centered at this point. The subsequence is denoted as Yi:i+k-1 = [yi-k / / 2, yi-k / / 2+1, …, yi, …, yi+k / / 2], where k / / 2 represents the integer part of k divided by 2;

[0051] Step c: Sort the subsequence Yi:i+k-1 to obtain Ysorted,i:i+k-1 = [y(1),i, y(2),i, …, y(k),i], where y(j),i represents the j-th smallest element after sorting;

[0052] Step d: Multiply the sorted subsequence by the weight coefficient table W and sum to obtain the processed data point yi′. The formula is:

[0053]

[0054] Secondly, obtain the change rate around each point of the sequence Y′, and perform convolution processing on the signal sequence Y′. The convolution kernel E = [-1, 1], and a new sequence C can be obtained. Each element ci is the dot product of the convolution kernel and the subsequence at the corresponding position in the signal: ci = (E·Yi:i+1) = yi+1 - yi for i = 2, 3, …, n-1

[0055] Finally, sum the consecutive sequence points in sequence C that are greater than 0, set a threshold 1, and take the position of the last point of the sequence that exceeds threshold 1 as the initial right edge. Similarly, sum the consecutive sequence points that are less than 0, set a threshold 2, and take the position of the last point of the sequence that exceeds threshold 2 as the initial left edge. Replace the noise data between the previous initial right edge and the next initial left edge with the edge value data to eliminate interference. Then, set an edge threshold, obtain the final positions of the left and right edges, and then calculate the distance value between the left and right edges of the peak group.

[0056] The present invention processes waveform signals of multiple peak groups through a sliding window and an adjustable weight coefficient table, calculates the left and right edges and the width of occlusion (i.e., the distance between the left and right edges), which serves as the basic features of the peak group; merges the position information with the occlusion width error data, taking into account the error magnitude or consistency of the width data measured by the peak group at different positions while detecting the linearity of the peak group; regresses the peak group position information with multiple peak group feature variables to obtain the regression error, and judges the error of the equal-width occlusion irradiated at each position of the sensor based on the magnitude of the regression error, so as to determine whether the measurement accuracy of the sensor meets the conditions; compared with the prior art, the method proposed by the present invention can not only measure the linearity of the sensor, but also determine the consistency of the measurement accuracy of the sensor at different positions and the magnitude of the measurement error. With relatively low measurement costs, it can comprehensively evaluate the quality of the sensor, greatly improving the production efficiency of the sensor.

Claims

1. A quality detection method for a laser displacement sensor with multi - feature errors, characterized in that, Including the following steps: Step 1: Direct the light beam emitted by the sensor onto the measurement plate to obtain a diffracted light beam, and convert the diffracted light beam into a waveform signal. The waveform signal includes n peak groups, and each peak group includes two edges, where n≥1; Step 2: Calculate the parameters and errors of the peak group, and the steps are as follows: Step 2-1: Calculate the left and right edge distances and edge positions of each peak group. The distance data is recorded as variable D = [d1, d2, ……, d n , the left edge position is recorded as variable L = [l1, l2, ……, l n , the right edge position is recorded as variable R = [r1, r2, ……, r n . After obtaining the left and right edge positions, take the midpoint position of the left and right edge distances as the center position of each peak group. The center position data is recorded as variable Q = [q1, q2, ……, q n , and record the actual positions of multiple peak groups along the preset direction, which is recorded as variable X = [x1, x2, ……, x n ; Step 2-2: Obtain the standard width data blocked by the measuring plate, and calculate the deviation of D based on this. The error data is recorded as variable F = [f1, f2, ……, f n ; Step 2-3: Superimpose the error data F on the position feature X along the preset direction to generate a new feature K; Step 2-4: Perform fitting or regression processing on the variables X, Q, L, R, and K, and calculate the corresponding errors; Step 3: Compare the error obtained in Step 2 with the preset error threshold. When the error is less than the preset error threshold, it is determined that the device quality is qualified, and the parameters in the process are recorded. When the error is greater than the preset error threshold, shorten the length of the preset weight coefficient table, repeat Step 2, recalculate the error and execute Step 3. When the error still cannot be less than the preset error threshold after repeating N times, where N≥3, it is determined that the device quality is unqualified, and the optical path and hardware of the laser displacement sensor device are re-debugged.

2. The quality detection method of a laser displacement sensor with multiple feature errors according to claim 1, characterized in that, In Step 3, when the error is greater than the preset error threshold, the method for shortening the preset weight coefficient table is as follows: Step 3-1: Let the error be E and the preset error threshold be T. According to the magnitude of the error value E, calculate the adjustment ratio P of the weight coefficient table, where P=(E - T) / T*α, and α is the adjustment coefficient, with 0.3≤α≤0.9; Step 3-2: According to the adjustment ratio P, shorten the length of the preset weight coefficient table. The minimum length of the preset weight coefficient table is 3, and the length L of the new weight coefficient table new = L old × (1 - P), where L old is the length of the original weight coefficient table.

3. A method for quality inspection of a laser displacement sensor with multi-feature errors according to claim 1, characterized in that, In Step 1, the calculation method for the two edges of the peak group in the peak group is as follows: Step 1-1: Let the one-dimensional waveform signal be Y, i.e., Y = [y1, y2, …, y n , perform signal preprocessing on Y, and based on the preset weight coefficient table W = [w1, w2, ……, w k , with the length set to k, perform smoothing processing on the sequence Y based on the sliding window and the weight coefficient table; Step 1-2: For each data point y in Y i , taking a subsequence of odd length k centered at this point, and the subsequence is denoted as Y i:i+k-1 = [y i-k / / 2 , yi -k / / 2+1 , …, y i , …, y i+k / / 2 , where k / / 2 represents the integer part of k divided by 2; Step 1-3: Sort the subsequence Y i:i+k-1 to obtain Y sorted,i:i+k-1 = [y (1),i , y (2),i , …, y (k),i , where y (j),i represents the j-th smallest element after sorting; Step 1-4: Multiply the sorted subsequence by the weight coefficient table W and sum to obtain the processed data point y i′ , and the formula is: Secondly, obtain the change rates around each point of the sequence Y′, and perform convolution processing on the signal sequence Y′. The convolution kernel E = [-1, 1], and a new sequence C is obtained, where each element c i is the dot product of the convolution kernel and the subsequence at the corresponding position in the signal: c i = (E·Y i:i+1 ) = y i+1 -y i For i = 2, 3, …, n - 1; Step 1-5: Sum the sequence points in sequence C that are continuously greater than 0, set a threshold 1, and take the position of the last point in the sequence exceeding the threshold 1 as the initial right edge. Similarly, sum the sequence points that are continuously less than 0, set a threshold 2, and take the position of the last point in the sequence exceeding the threshold 2 as the initial left edge. Replace the noise data between the previous initial right edge and the next initial left edge with edge value data to eliminate interference. Then, set an edge threshold to obtain the final positions of the left and right edges.

4. A quality detection method for a laser displacement sensor with multi-feature errors according to claim 1, characterized in that In Step 2-3, the method for generating the new feature K is as follows: Standardize the position data X of each occlusion and the calculated error data F along the preset direction respectively, so that the two variables become the same order of magnitude, and then add the two variables to obtain the new feature K.

5. A quality detection method for a laser displacement sensor with multi-feature errors according to claim 1, characterized in that, When fitting processing is adopted in Step 2-4, it specifically includes the following steps: Step 2-4-1: Perform linear fitting on the variable X with the variables Q, L, R, and K respectively; Step 2-4-2: Calculate four fitting errors, and superimpose the errors to obtain the corresponding error value.

6. A method for quality inspection of a laser displacement sensor with multi-feature errors according to claim 1, characterized in that, In Step 2-2, the calculation method for the error is as follows: Obtain the actual value of the width, subtract this actual value from each calculated distance result in the variable D to obtain the deviation value of the variable D, which is re-recorded as the variable F, and then take the absolute value of F.

7. An apparatus for a method of quality inspection of a laser displacement sensor based on the multi-feature error as described in any one of claims 1-6, provided with a measurement platform, characterized in that, Two laser displacement sensors and a controllable linear slide rail are horizontally placed on a measurement platform. A slider that matches the slide rail is provided on the linear slide rail. There is a groove for fixing a measurement plate in the middle of the slider. The measurement plate is placed in the groove and is automatically moved by the linear slide rail to be fixed between the two laser displacement sensors. Two or more equal-width light-blocking elements are provided on the measurement plate and are arranged parallel to each other along a preset direction. Each light-blocking element has two edges that are oppositely arranged along the preset direction. The two laser displacement sensors are arranged to emit light beams towards each other to the measurement plate, and diffracted light beams are obtained.

Images