CN111855803B - A laser-ultrasonic high signal-to-noise ratio imaging method for metal additive manufacturing micro-defects - Google Patents

Abstract

The invention discloses a laser ultrasonic high signal-to-noise ratio imaging method for metal additive manufacturing micro-defects. The steps include: scanning a laser ultrasonic system to obtain three-dimensional matrix data; performing noise reduction processing on the data; Perform similarity calculation and maximum amplitude time shift calculation for all signals; perform normalization and translation processing on signals that satisfy both similarity conditions and amplitude shift conditions; extract the corresponding maximum amplitude time from the three-dimensional data. Two-dimensional matrix; outlier elimination processing and image rendering are performed on the matrix to obtain high signal-to-noise ratio images of defects. The beneficial effects of the invention are as follows: the invention can eliminate the influence of the surface roughness of the additive product on the laser ultrasonic signal and the image, separate the micro-defect signal and the strong background noise signal, and improve the detection of defects without grinding the rough surface. probability, laying the foundation for laser ultrasonic online inspection of additive manufacturing.

Description

A laser ultrasonic high signal-to-noise ratio imaging method for metal additive manufacturing micro-defects

technical field

The invention relates to the technical field of laser ultrasonic nondestructive testing, in particular to a laser ultrasonic high signal-to-noise ratio imaging method for metal additive manufacturing micro-defects.

Background technique

Metal additive manufacturing is a transformative advanced manufacturing technology. Due to its point-by-point accumulation characteristics, metal additive manufacturing will inevitably produce defects in the manufacturing process, and most of the defects are in the micrometer scale. The non-contact inspection technology represented by laser ultrasound is considered to be the most feasible method for online inspection of additive manufacturing.

Laser ultrasound can generate high-frequency ultrasonic waves by exciting the additive surface with a laser, thereby realizing the detection of micro-defects. The commonly used laser ultrasonic C-scan method can realize the imaging of surface micro-defects. However, since the receiver of laser ultrasound is sensitive to the surface roughness of the material, the received ultrasound signal is usually accompanied by a strong noise signal. In actual detection, these noise signals are often stronger than the ultrasonic signals generated by micro-defects, so that the defect signals are submerged in the noise signals, causing misjudgment and missed detection of defects.

In order to realize the accurate inspection of defects, most of the existing literatures grind the rough surface of the additive parts to smooth, so as to ensure the high signal-to-noise ratio of the laser receiver. However, for in-line inspection of additive manufacturing, adding grinding units can interfere with the manufacturing process. Therefore, how to achieve high signal-to-noise ratio imaging of micro-defects in additive manufacturing by means of post-signal processing without damaging the rough surface and interfering with the printing process is a key point restricting the practical application of laser ultrasound.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a laser ultrasonic high signal-to-noise ratio imaging method for metal additive manufacturing micro-defects in view of the deficiencies of the prior art, so as to realize the detection of micro-defects without removing the rough surface of the additive parts. Imaging with high signal-to-noise ratio avoids missing defect detection caused by submerged defect signals by noise signals, and provides guarantee for the realization of laser ultrasonic online inspection of additive manufacturing.

The technical scheme adopted in the present invention is: a laser ultrasonic high signal-to-noise ratio imaging method for metal additive manufacturing micro-defects, the method comprises the following steps:

S1. Use the laser ultrasonic system to scan two-dimensional grids on the surface of the additive parts, complete the acquisition of ultrasonic data, and obtain three-dimensional matrix data A(m,n,t), m=1...M, n=1...N, t=1...t _q , where m is the number of scanning rows, n is the number of scanning columns, and t is the signal acquisition length;

S2. Perform noise reduction processing on the collected ultrasonic signals to obtain an ultrasonic data matrix A1(m,n,t);

S3. From the noise-reduced ultrasound data matrix, select a group of signals A1 (m ₀ , n ₀ , t) in which only surface waves exist as reference signals;

S4. Compare all the ultrasonic signals A1(m,n,t) after noise reduction processing with the reference signal for similarity, and obtain the waveform similarity coefficient Nc(m,n);

S5. Extract the time t0 corresponding to the maximum surface wave amplitude of the reference signal A1 (m ₀ , n ₀ , t), extract the time t (m, n) corresponding to the maximum surface wave amplitude of all ultrasonic signals after noise reduction, and then divide the two Subtract them to obtain the time offset Δt(m,n)=t(m,n)-t0 of all ultrasonic signals and the reference signal;

S6. Judge all the noise-reduced ultrasonic signals A1(m,n,t) in turn. When the waveform similarity coefficient is greater than the threshold Nc0 and the time offset Δt is less than the threshold Δt0, the signal is normalized and waveform offset processing to obtain a normalized amplitude value of 1;

S7. From the processed signals, extract the amplitude values corresponding to all signals at time t0 to form a two-dimensional matrix A2(m,n);

S8. Eliminate outliers on the two-dimensional matrix A2(m,n), when the values of the four adjacent elements before, after, left and right of an element in the two-dimensional matrix are equal and the normalized amplitude is 1 , then set the element value to 1;

S9. Perform image drawing on the two-dimensional matrix processed in S8, so as to obtain a defect image with a high signal-to-noise ratio.

According to the above solution, in S2, the noise reduction processing method may be one of wavelet noise reduction processing, Hilbert Huang noise reduction processing, and deep learning self-encoding noise reduction processing.

According to the above scheme, in S4, the waveform similarity coefficient Nc(m,n) is obtained by using the waveform similarity function of the following formula:

In the above formula, s ₁ (t) represents the reference signal, and s ₂ (t) represents the ultrasonic signal A1 (t) corresponding to the specific positions m and n after noise reduction processing.

According to the above scheme, in S6, the normalization processing method is: dividing the signal by its maximum amplitude value to obtain the normalized amplitude value 1; the normalized signal is:

According to the above scheme, in S6, the method of waveform offset processing is to offset the signal by Δt; the signal after waveform offset processing is:

In the formula, t _q is the element number corresponding to the offset position of the signal array, and t _N is the length of the signal array. .

According to the above scheme, in S6, the method for determining the threshold Nc0 is: calculate the similarity coefficient of the reference signal and the ultrasonic signal in the defect-free area one by one, and take the smallest similarity coefficient as the threshold Nc0. The method for determining the threshold Δt0 is as follows: calculate the time offset between the reference signal and the ultrasonic signal in the defect-free area one by one, and take the maximum time offset as the threshold Δt0.

According to the above scheme, in S1, the laser ultrasonic system includes a pulsed laser for ultrasonic signal excitation and a laser Doppler vibrometer for ultrasonic signal reception. During the scanning process, the pulsed laser and the laser Doppler vibrometer are used for The spot spacing of the instrument remains fixed.

According to the above scheme, in S1, the size of the scanning grid is not greater than the target detection accuracy.

The beneficial effects of the present invention are as follows: the present invention obtains high signal-to-noise ratio imaging for additive micro-defect laser ultrasonic detection through the synthesis of signal noise reduction, similarity comparison, normalization processing and waveform translation processing, and solves the problem of The problem of strong noise in the laser ultrasonic system caused by the surface roughness of the additive material realizes the separation of micro-defects and strong background noise, thereby improving the detection probability and quantitative accuracy of defects. The invention realizes high signal-to-noise ratio imaging of defects without removing the surface roughness of the additive parts, which is of great significance for the subsequent intelligent automatic identification of defects, and lays a foundation for the application of laser ultrasonic technology and the online detection of additive manufacturing. important foundation.

Description of drawings

FIG. 1 is a schematic diagram of scanning a laser ultrasound system according to a specific embodiment of the present invention.

FIG. 2 is a schematic diagram of the original defect signal in this embodiment.

FIG. 3 is a schematic diagram of the noise reduction effect in this embodiment.

FIG. 4 is a schematic diagram of signal normalization processing and translation processing in this embodiment.

FIG. 5 is a schematic diagram of defect high signal-to-noise ratio imaging in this embodiment.

Detailed ways

For a better understanding of the present invention, the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

The 316L stainless steel additive manufacturing sample was prepared by powder printing method. The average surface roughness of the actual test sample was Ra7.5μm. Laser drilling was used on the surface of the sample. The processing depth was 50μm and the diameters were 100μm and 50μm. kind of holes. Based on the above samples, the present invention provides a method for accurately measuring the defect size of rough parts based on laser ultrasonic imaging, which specifically includes the following steps:

S1. Use a laser ultrasonic system to scan a two-dimensional grid on the surface of an additive product (that is, the sample in this embodiment), complete the acquisition of ultrasonic data, and obtain three-dimensional matrix data A(m,n,t),m =1...M, n=1...N, t=1...t _q , where m and n represent the position where the signal is acquired, m is the number of scanning rows, n is the number of scanning columns, and t is the signal acquisition length.

In the present invention, the laser ultrasonic system includes a pulsed laser for ultrasonic signal excitation, a laser Doppler vibrometer for ultrasonic signal reception, and a pulsed laser (that is, an excitation device) and a laser Doppler vibrometer during the scanning process. The spot spacing of the vibrometer (ie the receiver) remains fixed.

The scan grid size is not larger than the target detection accuracy. In this embodiment, as shown in FIG. 1 , the grid size is set to 50 μm; the number of scanning rows in the three-dimensional matrix is M=400, the number of scanning columns is N=130; the signal acquisition length t _q =500, and the acquired data draws the defect image as shown in the figure 2 shown.

S2. Perform noise reduction processing on the collected ultrasonic signals to obtain an ultrasonic data matrix A1(m,n,t).

In the present invention, the noise reduction processing method may be one of wavelet noise reduction processing, Hilbert Huang noise reduction processing and deep learning self-encoding noise reduction processing. In this embodiment, deep learning self-encoding is used to perform noise reduction processing on the collected ultrasound signals, and the ultrasound data matrix after noise reduction processing is shown in FIG. 3 .

S3. From the denoised ultrasound data matrix, select a group of signals A1 (m ₀ , n ₀ , t) in which only surface waves exist as reference signals.

S4. Compare all ultrasonic signals A1(m,n,t) after noise reduction processing with the reference signal A1( _m0 , _n0 ,t) for similarity to obtain waveform similarity coefficient Nc(m,n).

In the present invention, the waveform similarity coefficient Nc(m,n) is obtained by calculating the waveform similarity function shown in the following formula,

In the above formula, s ₁ (t) represents the reference signal, and s ₂ (t) represents the ultrasonic signal A1(t) corresponding to the specific positions m and n after noise reduction processing.

S5. Extract the time t0 corresponding to the maximum surface wave amplitude of the reference signal A1 (m ₀ , n ₀ , t), extract the time t (m, n) corresponding to the maximum surface wave amplitude of all ultrasonic signals after noise reduction, and then divide the two Subtract the above to obtain the time offset Δt(m,n)=t(m,n)−t0 of all ultrasonic signals and the reference signal.

In this embodiment, the time t0=208 corresponding to the maximum value of the surface wave amplitude of the reference signal A1 (m ₀ , n ₀ , t).

S6. Judging all the denoised ultrasonic signals A1(m,n,t) in turn, when the waveform similarity coefficient Nc(m,n) is greater than the threshold Nc0 and the time offset Δt is less than the threshold Δt0, the The signal is subjected to normalization processing and waveform offset processing to obtain a normalized amplitude value of 1.

In the present invention, the normalization processing method is: dividing the signal satisfying the requirement by its maximum amplitude value to obtain the normalized amplitude value 1; the normalized signal is:

In the present invention, the waveform offset processing method is: offset the normalized signal by Δt; the offset signal is:

In the formula, t _q is the element number corresponding to the offset position of the signal array, and t _N is the length of the signal array.

In the present invention, the method for determining the threshold Nc0 is as follows: calculate the similarity coefficient of the reference signal and the ultrasonic signal in the defect-free area one by one, and take the smallest similarity coefficient as the threshold Nc0. The method for determining the threshold Δt0 is as follows: calculate the time offset between the reference signal and the ultrasonic signal in the defect-free area one by one, and take the maximum time offset as the threshold Δt0.

In this embodiment, the threshold value Nc0=0.6, and the threshold value Δt0=5.

S7. From the processed signals, extract the amplitude values corresponding to all the signals at time t0 to form a two-dimensional matrix A2(m,n).

S8. Eliminate outliers on the two-dimensional matrix A2(m,n), when the values of the four adjacent elements before, after, left and right of an element in the two-dimensional matrix are equal and the normalized amplitude is 1 , then set the element value to 1.

S9. Perform image drawing on the two-dimensional matrix processed in S8, so as to obtain a defect image with a high signal-to-noise ratio, as shown in FIG. 5 .

Compared with the schematic diagram of the original defect signal shown in Figure 2, the image processed by the technical solution of the present invention is shown in Figure 5, the background noise of the image is completely eliminated, and a pure inspection image is obtained: for a 100 μm defect, the surrounding After the noise effect is eliminated, the defect boundary is revealed, which facilitates the subsequent defect size measurement; for a 50μm defect, the defect boundary in the original defect signal schematic diagram was originally submerged by the noise signal and could not be displayed, but it is fully displayed in Figure 5.

The invention first eliminates the noise of a single signal caused by the surface roughness of the additive through signal noise reduction processing, and secondly, through waveform normalization and translation processing, it eliminates the uneven separation of surface roughness and the possible noise caused by the noise reduction algorithm. For the problem of defect amplitude and fluctuation at the corresponding moment, the abnormal signal caused by the detection system during the scanning process is eliminated again through the outlier elimination process, and finally a noise-free defect image is obtained. The invention can be applied to high signal-to-noise ratio imaging of additive micro-defect laser ultrasonic inspection, solves the problem of strong noise in the laser ultrasonic system caused by the surface roughness of the additive, realizes the separation of micro-defects and strong background noise, and improves the defect rate. detection probability and quantitative accuracy. The present invention can realize the high signal-to-noise ratio imaging of defects without removing the surface roughness of the additive parts, which is of great significance for the subsequent intelligent automatic identification of defects, and is a useful tool for the application of laser ultrasonic technology and the online detection of additive manufacturing. Lay important foundations.

The above-mentioned embodiments are only intended to illustrate the technical concept and characteristics of the present invention, and the purpose thereof is to enable those who are familiar with the art to understand the content of the present invention and implement them accordingly, and cannot limit the protection scope of the present invention. All equivalent changes or modifications made according to the spirit of the present invention should be included within the protection scope of the present invention.

Claims (8)

1. A laser ultrasonic high signal-to-noise ratio imaging method for metal additive manufacturing micro-defects, characterized in that the method comprises the following steps: S1. Use the laser ultrasonic system to scan two-dimensional grids on the surface of the additive parts, complete the acquisition of ultrasonic data, and obtain three-dimensional matrix data A(m,n,t), m=1...M, n=1...N, t=1...t _q , where m is the number of scanning rows, n is the number of scanning columns, and t is the signal acquisition length; S2. Perform noise reduction processing on the collected ultrasonic signals to obtain an ultrasonic data matrix A1(m,n,t); S3. From the noise-reduced ultrasound data matrix, select a group of signals A1 (m ₀ , n ₀ , t) in which only surface waves exist as reference signals; S4. Compare all the ultrasonic signals A1(m,n,t) after noise reduction processing with the reference signal for similarity, and obtain the waveform similarity coefficient Nc(m,n); S5. Extract the time t0 corresponding to the maximum surface wave amplitude of the reference signal A1 (m ₀ , n ₀ , t), extract the time t (m, n) corresponding to the maximum surface wave amplitude of all ultrasonic signals after noise reduction, and then divide the two The time offset Δt(m,n)=t(m,n)-t0 of all ultrasonic signals and the reference signal is obtained; S6. Judge all the noise-reduced ultrasonic signals A1(m,n,t) in turn. When the waveform similarity coefficient is greater than the threshold Nc0 and the time offset Δt is less than the threshold Δt0, the signal is normalized and waveform offset processing to obtain a normalized amplitude value of 1; S7. From the processed signals, extract the amplitude values corresponding to all signals at time t0 to form a two-dimensional matrix A2(m,n); S8. Eliminate outliers on the two-dimensional matrix A2(m,n), when the values of the four adjacent elements before, after, left and right of an element in the two-dimensional matrix are equal and the normalized amplitude is 1 , then set the element value to 1; S9. Perform image drawing on the two-dimensional matrix processed in S8, so as to obtain a defect image with a high signal-to-noise ratio. 2. The laser ultrasound high signal-to-noise ratio imaging method according to claim 1, wherein in S2, the noise reduction processing method can be wavelet noise reduction processing, Hilbert Huang noise reduction processing and deep learning One of the auto-encoding noise reduction processes. 3. The laser ultrasound high signal-to-noise ratio imaging method according to claim 1, wherein in S4, the waveform similarity coefficient Nc(m,n) is obtained by using the waveform similarity function of the following formula:

In the above formula, s ₁ (t) represents the reference signal, and s ₂ (t) represents the ultrasonic signal A1 (t) corresponding to the specific positions m and n after noise reduction processing. 4. The laser ultrasound high signal-to-noise ratio imaging method according to claim 3, wherein in S6, the normalization processing method is: dividing the signal by its maximum amplitude value to obtain a normalized amplitude value 1 ; The normalized signal is:

In the formula, s ₀ (t) is a one-dimensional array composed of amplitude values, and max(s ₀ (t)) refers to the maximum amplitude value taken from the one-dimensional array s0 (t). 5. The laser ultrasound high signal-to-noise ratio imaging method according to claim 3, characterized in that, in S6, the method of waveform shift processing is to shift the signal by Δt; the signal after waveform shift processing is:

In the formula, t _q is the element number corresponding to the offset position of the signal array, and t _N is the length of the signal array. 6. The laser ultrasound high signal-to-noise ratio imaging method according to claim 3, wherein in S6, the method for determining the threshold Nc0 is: calculating the similarity coefficient of the reference signal and the ultrasound signal of the defect-free area one by one, taking The minimum similarity coefficient is used as the threshold Nc0; the method for determining the threshold Δt0 is: calculate the time offset between the reference signal and the ultrasonic signal in the defect-free area one by one, and take the maximum time offset as the threshold Δt0. 7. The laser ultrasound high signal-to-noise ratio imaging method according to claim 1, wherein in S1, the laser ultrasound system comprises a pulsed laser for ultrasound signal excitation, a laser Doppler for ultrasound signal reception For the laser vibrometer, the spot spacing of the pulsed laser and the laser Doppler vibrometer remains fixed during the scanning process. 8 . The laser ultrasound high signal-to-noise ratio imaging method according to claim 1 , wherein, in S1 , the size of the scanning grid is not greater than the target detection accuracy. 9 .

Images