Laser deposition acoustic signal online noise reduction method and system based on dynamic adaptive filtering

Abstract

The invention provides a laser deposition acoustic signal online noise reduction method and system based on dynamic adaptive filtering. The method comprises the following steps: acquiring an original acoustic signal in a laser directional energy deposition process; determining a path risk coefficient according to the laser deposition path; obtaining the powder flow at the current moment, and calculating a dynamic powder impact noise mask; when the risk coefficient is greater than a threshold value, setting the basic weight as a preset value, otherwise, taking the proportion of the signal-to-noise ratio of the mechanical vibration noise in the total signal-to-noise ratio as the basic weight; compensating the basic weight according to the path risk coefficient to generate a final weight; performing denoising and band-pass filtering on the original acoustic signal to obtain a first signal; for the time-frequency spectrum of the original acoustic signal, generating a time-frequency mask through a dynamic powder impact noise mask, and suppressing the noise to obtain a second signal; and carrying out weighted fusion on the first signal and the second signal by adopting the final weight to obtain a denoised and reconstructed molten pool acoustic signal. The noise reduction effect is better.

Description

Technical Field

[0001] This invention relates to laser deposition acoustic signal processing technology, and more particularly to an online noise reduction method and system for laser deposition acoustic signals based on dynamic adaptive filtering. Background Technology

[0002] Laser-directed energy deposition (LDED) technology, a key branch of metal additive manufacturing, has been widely used in aerospace, high-end equipment, and other fields due to its ability to efficiently form complex components. This process relies on real-time monitoring of the acoustic emission signal from the molten pool to assess the forming quality (typical defect characteristic frequencies are concentrated in the 10-15kHz range). However, in actual operating conditions, the acoustic signal is severely contaminated: mechanical vibrations of the joints during multi-axis robot operation generate low-frequency interference of 200-800Hz, especially causing sudden changes in noise amplitude during path turns and speed changes; simultaneously, the continuous impact of the metal powder flow on the molten pool or substrate generates 2-5kHz broadband white noise, whose spectral energy overlaps with the key acoustic characteristic frequency bands of the molten pool, causing interference. Furthermore, fluctuations in the powder feed airflow and environmental vibrations further degrade the signal-to-noise ratio, leading to a decrease in the accuracy of molten pool condition identification.

[0003] Existing noise reduction technologies generally employ a single filter architecture (such as a fixed band-stop filter or a general adaptive filter), which has inherent technical limitations. Because they fail to decouple the time-frequency characteristics of mechanical vibration noise and powder impact noise, traditional methods mistakenly suppress the 10-15kHz characteristic frequency band of the molten pool during noise filtering, resulting in a measured signal distortion rate exceeding 30%. Furthermore, most methods use fixed-parameter filtering, leading to an inability to respond in real-time to sudden increases in mechanical vibration energy that could cause false triggering of the monitoring system during dynamic path execution. Existing technologies also lack effective suppression capabilities for powder impact noise, which exhibits non-stationary characteristics. These bottlenecks restrict the application of LDED technology in high-reliability fields. Summary of the Invention

[0004] To address the problems existing in the prior art, the purpose of this invention is to provide an online noise reduction method and system for laser-deposited acoustic signals based on dynamic adaptive filtering, which offers better noise reduction performance.

[0005] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0006] An online noise reduction method for laser-deposited acoustic signals based on dynamic adaptive filtering includes the following steps:

[0007] (1) Acquire the raw acoustic signals during the laser-directed energy deposition process;

[0008] (2) Extract path features based on the laser deposition path, and determine the main frequency of mechanical vibration noise and the path risk coefficient based on the path features;

[0009] (3) Obtain the current powder flow rate of laser deposition, and calculate the powder impact noise frequency band and dynamic powder impact noise mask based on the powder flow rate;

[0010] (4) When the path risk coefficient is greater than or equal to the preset threshold, set the basic weight to the preset weight; otherwise, proceed to step (5).

[0011] (5) Based on the signals in the original acoustic signal located in the main frequency band of mechanical vibration noise, the frequency band of powder impact noise and the frequency band of molten pool sound, calculate the signal-to-noise ratio of mechanical vibration noise and the signal-to-noise ratio of powder impact noise, and take the proportion of mechanical vibration noise signal-to-noise ratio in the total signal-to-noise ratio as the basic weight.

[0012] (6) Compensate the basic weights according to the path risk coefficient to generate vibration channel weights;

[0013] (7) Remove the mechanical vibration noise component from the original acoustic signal and perform bandpass filtering on the remaining signal in the molten pool acoustic frequency band to obtain the first signal; the mechanical vibration noise component is determined by the main frequency of mechanical vibration noise;

[0014] (8) Obtain the time spectrum of the original acoustic signal, and generate a time-frequency mask based on the dynamic powder impact noise mask. After suppressing the powder impact noise in the time spectrum using the time-frequency mask, transform it to the time domain and use it as the second signal.

[0015] (9) The first signal and the second signal are weighted and fused by the vibration channel weight to obtain the denoised and reconstructed molten pool acoustic signal.

[0016] Furthermore, step (2) specifically includes:

[0017] (2.1) Obtain the laser deposition path, which includes the coordinates and motion velocity at each time point;

[0018] (2.2) Extract path features based on the coordinates and movement speed at each time point, wherein the path features are path types;

[0019] (2.3) Based on the path type, the main frequency of mechanical vibration noise is obtained according to the preset path type-mechanical vibration noise main frequency mapping rule;

[0020] (2.4) Based on the path type, the path risk coefficient is obtained according to the preset path type-path risk coefficient mapping rule.

[0021] Furthermore, step (3) specifically includes:

[0022] (3.1) Obtain the powder flow rate of laser deposition at the current moment, and calculate the center frequency of powder impact noise based on the powder flow rate;

[0023] (3.2) Determine the frequency band of powder impact noise based on the center frequency of powder impact noise;

[0024] (3.3) Determine the dynamic powder impact noise mask based on the center frequency of the powder impact noise.

[0025] Furthermore, step (5) specifically includes:

[0026] (5.1) Extract the signal located in the main frequency band of mechanical vibration noise from the original acoustic signal and calculate the mechanical vibration noise power;

[0027] (5.2) Extract the signal located in the frequency band of powder impact noise from the original acoustic signal and calculate the powder impact noise power;

[0028] (5.3) Extract the signal located in the molten pool acoustic frequency band from the original acoustic signal and calculate the molten pool acoustic signal power;

[0029] (5.4) Determine the signal-to-noise ratio of mechanical vibration noise based on the ratio of the power of the acoustic signal of the molten pool to the power of the mechanical vibration noise; determine the signal-to-noise ratio of powder impact noise based on the ratio of the power of the acoustic signal of the molten pool to the power of the powder impact noise.

[0030] (5.5) Calculate the weight of the mechanical vibration noise signal-to-noise ratio in the sum of the mechanical vibration noise signal-to-noise ratio and the powder impact noise signal-to-noise ratio, and use it as the basic weight.

[0031] Furthermore, step (6) specifically includes:

[0032] The basic weights are compensated based on the path risk coefficient, and the mechanical vibration noise weights are generated according to the following formula:

[0033] α final = a + bR + c(α - d)

[0034] In the formula, α final The values ​​represent the mechanical vibration noise weights, where a, b, c, and d are compensation coefficients, R is the path risk coefficient, and α is the base weight.

[0035] Furthermore, step (7) specifically includes:

[0036] (7.1) Perform EMD intrinsic mode decomposition on the original acoustic signal and extract the intrinsic mode components where the mechanical vibration noise is located;

[0037] (7.2) Remove the intrinsic mode components containing mechanical vibration noise from the original acoustic signal;

[0038] (7.3) The remaining signal is bandpass filtered in the molten pool acoustic band to obtain the first signal.

[0039] Furthermore, step (8) specifically includes:

[0040] (8.1) Perform STFT transformation on the original acoustic signal to obtain the time spectrum of the original acoustic signal;

[0041] (8.2) Based on the dynamic powder impact noise mask, the time-frequency mask is calculated according to the following formula:

[0042]

[0043] In the formula, M(t,f) is the time-frequency mask, η(Q) is the powder impact noise mask, and S(t,f) is the time spectrum;

[0044] (8.3) Multiply the time-frequency mask by the time-frequency spectrum to obtain the time-frequency spectrum that suppresses powder impact noise;

[0045] (8.4) Perform inverse STFT transformation on the time spectrum of the powder impact noise suppression to obtain the second signal in the time domain.

[0046] Furthermore, step (9) specifically includes: using mechanical vibration noise weights to weight and fuse the first and second signals according to the following formula to obtain the denoised and reconstructed molten pool acoustic signal:

[0047] s out (t)=α final s vib (t)+(1-α final )·s powder (t)

[0048] In the formula, s out (t) represents the denoised and reconstructed molten pool acoustic signal, α final Indicates the vibration channel weight, s vib (t), s powder (t) represent the first signal and the second signal, respectively.

[0049] An online noise reduction system for laser-deposited acoustic signals based on dynamic adaptive filtering includes:

[0050] The acoustic signal acquisition subsystem is used to acquire the raw acoustic signals during the laser directional energy deposition process;

[0051] The process path dynamic prediction module is used to extract path features based on the laser deposition path, and determine the main frequency of mechanical vibration noise and the path risk coefficient based on the path features;

[0052] The noise feature analysis and channel scheduling module is used to obtain the powder flow rate of laser deposition at the current moment, calculate the powder impact noise frequency band and dynamic powder impact noise mask based on the powder flow rate; when the path risk coefficient is greater than or equal to a preset threshold, the basic weight is set to the preset weight; otherwise, based on the signals in the original acoustic signal located in the main frequency band of mechanical vibration noise, the frequency band of powder impact noise and the frequency band of molten pool sound, the mechanical vibration noise signal-to-noise ratio and the powder impact noise signal-to-noise ratio are calculated, and the proportion of the mechanical vibration noise signal-to-noise ratio in the total signal-to-noise ratio is used as the basic weight; the basic weight is compensated according to the path risk coefficient to generate vibration channel weight;

[0053] The mechanical vibration noise channel module is used to remove the mechanical vibration noise component from the original acoustic signal and perform bandpass filtering on the remaining signal in the molten pool acoustic frequency band to obtain the first signal;

[0054] The powder impact noise channel module is used to acquire the time spectrum of the original acoustic signal, and generate a time-frequency mask based on the dynamic powder impact noise mask. After suppressing the powder impact noise in the time spectrum using the time-frequency mask, the signal is transformed to the time domain and used as a second signal.

[0055] The signal denoising and reconstruction module is used to weight and fuse the first and second signals using vibration channel weights to obtain the denoised and reconstructed molten pool acoustic signal.

[0056] Furthermore, the acoustic signal acquisition subsystem includes a laser generator, a control system, a three-axis robotic arm, an inert gas supply device, a forming base, a directional microphone, a data acquisition card, and a powder conveying mechanism. The laser generator is positioned above the forming base. The control system is connected to both the laser generator and the three-axis robotic arm. The laser generator is connected to the powder conveying mechanism to achieve photo-powder coupling. The inert gas supply device is connected to the powder conveying mechanism. The directional microphone is fixed to the side of the forming base and is connected to the data acquisition card to acquire the raw acoustic signal.

[0057] Compared with the prior art, the beneficial effects of this invention are:

[0058] 1. This invention simplifies traditional full-band analysis by real-time analysis of the process path, reducing data size by more than 80% through dual-channel separate processing of 200-800Hz mechanical vibration and 2-5kHz powder noise. Simultaneously, it utilizes an adaptive filtering architecture to replace manual parameter tuning, meeting the real-time monitoring requirements of the laser deposition process.

[0059] 2. By employing an EMD-STFT collaborative filtering mechanism (mechanical channel sieve vibration noise component, powder channel dynamic masking) and a weighted fusion algorithm, accurate extraction of the key frequency band (10-15kHz) of the molten pool is achieved. Compared to traditional fixed filters, this method reduces the signal distortion rate from >30% to ≤8% under 90° path change conditions, increases the molten pool feature retention rate to 92% (experimental verification data), and can automatically identify defect features such as bubble rupture and splashing. Attached Figure Description

[0060] Figure 1 A schematic flowchart of an online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering provided in an embodiment of the present invention;

[0061] Figure 2 This is a schematic diagram of the acoustic signal acquisition subsystem in an embodiment of the present invention;

[0062] Figure 3 This is a comparison diagram of signal processing in the mechanical vibration channel in an embodiment of the present invention;

[0063] Figure 4 This is a comparison diagram of powder noise channel signal processing in an embodiment of the present invention;

[0064] Figure 5 This is a schematic diagram of the dual-channel weighted fusion signal in an embodiment of the present invention. Detailed Implementation

[0065] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0066] Example 1

[0067] This invention provides an online noise reduction method for laser-deposited acoustic signals based on dynamic adaptive filtering, such as... Figure 1 As shown, it includes the following steps:

[0068] (1) Acquire the original acoustic signals during the laser directional energy deposition process.

[0069] Among them, the acquisition of the original acoustic signal can be achieved through... Figure 2The acoustic signal acquisition subsystem shown includes a laser generator 1, a control system 2, a three-axis robotic arm 3, an inert gas supply device 4, a forming base 5, a directional microphone 6, a data acquisition card 7, and a powder conveying mechanism 8. The laser generator 1 is positioned above the forming base 5. The control system 2 is connected to both the laser generator 1 and the three-axis robotic arm 3. The laser generator 1 is connected to the powder conveying mechanism 8 to achieve photo-powder coupling. The inert gas supply device 4 is connected to the powder conveying mechanism 8. The directional microphone 6 is fixed to the side of the forming base 5 for acquiring acoustic signals and is connected to the data acquisition card 7.

[0070] During data acquisition, the control system 2 drives the three-axis robotic arm 3 to move according to the preset processing path (G-code), simultaneously controlling the laser generator 1 to emit a laser beam, and synchronously controlling the powder conveying mechanism 8 to transport metal powder to the molten pool area through the airflow provided by the inert gas supply device 4. A directional microphone 6 collects acoustic signals generated near the molten pool in real time, including acoustic emission from the molten pool, mechanical vibration noise, and powder impact noise. The collected analog acoustic signals are converted into digital signals via analog-to-digital conversion (ADC) using a data acquisition card 7. In this embodiment, the laser power is kept constant at 900W and the scanning speed at 9mm / s. By changing the process path, signal acquisition experiments are conducted under stable cladding and corner cladding process conditions. The experimental material is 316L stainless steel powder, the laser spot diameter is 2mm, and the powder feed rate is 7-8g / min. The collected acoustic signals are sound pressure data, mainly acquired using LabVIEW's DAQ module with a sampling frequency of X. The sound pressure data is saved as a TDMS file. By fully leveraging existing experience in laser additive manufacturing and the advantages of TDMS files—high speed, easy access, and convenience—we can manufacture acoustic signals generated under different process paths using the same process parameters and acquire them in real time.

[0071] (2) Dynamic prediction of process road: Based on the laser deposition path, the path features are extracted, and the main frequency of mechanical vibration noise and the path risk coefficient are determined based on the path features.

[0072] Step (2) specifically includes:

[0073] (2.1) Obtain the laser deposition path, which includes the coordinates and motion speed at each time.

[0074] The laser deposition path is typically a sequence of G-codes or similar motion instructions, containing Cartesian coordinates (X, Y, Z) at each timestamp and motion velocity parameters.

[0075] (2.2) Extract path features based on the coordinates and movement speed at each time point. The path features are path types.

[0076] By using coordinates and velocity, the position point at the next moment can be calculated, and then the path type of that position point can be obtained, such as the path segment type being a straight line, an arc, or an acute angle.

[0077] (2.3) Based on the path type, the mechanical vibration noise main frequency is obtained according to the preset path type-mechanical vibration noise main frequency mapping rule.

[0078] The path type-mechanical vibration noise dominant frequency mapping rule can be set as follows:

[0079]

[0080] This indicates the dominant frequency of mechanical vibration noise.

[0081] (2.4) Based on the path type, the path risk coefficient is obtained according to the preset path type-path risk coefficient mapping rule.

[0082] The specific mapping rule between path type and path risk coefficient is as follows:

[0083]

[0084] R represents the path risk coefficient.

[0085] The following steps involve noise characteristic analysis and channel scheduling, specifically steps (3) to (6).

[0086] (3) Obtain the powder flow rate of laser deposition at the current moment, and calculate the powder impact noise frequency band and dynamic powder impact noise mask based on the powder flow rate.

[0087] Step (3) specifically includes:

[0088] (3.1) Obtain the current powder flow rate of laser deposition, and calculate the center frequency of powder impact noise based on the powder flow rate; wherein, the powder flow rate is specifically the powder flow rate at point 8 of the powder conveying mechanism, which can be measured by a flow meter; the center frequency of powder impact noise can be calculated according to the following formula:

[0089] B n (Q) = 150 × Q + 2500

[0090] Where Q is the powder flow rate (g / min), 150 is the proportionality coefficient (Hz·min / g), and 2500 is the basic bandwidth (Hz).

[0091] (3.2) The frequency band of powder impact noise is determined according to the center frequency of powder impact noise using the following formula;

[0092] Powder impact noise frequency band = [B n(Q)-500,B n (Q)+500]HZ;

[0093] For example, when Q = 8 g / min, B n If (Q) = 3700Hz, then the frequency band of powder impact noise is 3200-4200Hz.

[0094] (3.3) The dynamic powder impact noise mask is determined according to the following formula based on the center frequency of the powder impact noise:

[0095]

[0096] In the formula, η(Q) is the dynamic powder impact noise mask.

[0097] (4) When the path risk coefficient is greater than or equal to the preset threshold, set the basic weight to the preset weight; otherwise, execute step (5). Specifically, the preset threshold can be set to 1.0, and when the path risk coefficient R≥1.0, the basic weight α=0.8 is set.

[0098] (5) Based on the signals in the original acoustic signal located in the main frequency band of mechanical vibration noise, the frequency band of powder impact noise and the frequency band of molten pool sound, calculate the signal-to-noise ratio of mechanical vibration noise and the signal-to-noise ratio of powder impact noise, and use the proportion of mechanical vibration noise signal-to-noise ratio in the total signal-to-noise ratio as the basic weight.

[0099] Step (5) specifically includes:

[0100] (5.1) Extract the signal located in the main frequency band of mechanical vibration noise from the original acoustic signal and calculate the mechanical vibration noise power;

[0101] The dominant frequency band of mechanical vibration noise is [200, 800 Hz], therefore the specific formula for calculating the power of mechanical vibration noise is:

[0102]

[0103] In the formula, S(f) represents the mechanical vibration noise power, S(f) represents the FFT spectrum of the original acoustic signal, f represents the frequency, and N represents the number of sampling points in that frequency band.

[0104] (5.2) Extract the signal located in the frequency band of powder impact noise from the original acoustic signal and calculate the powder impact noise power;

[0105] The frequency band of powder impact noise is determined by the center frequency of powder impact noise, and the formula for calculating powder impact noise power is:

[0106]

[0107] In the formula, denoted as powder impact noise power, and M as the number of sampling points in this frequency band.

[0108] (5.3) Extract the signal located in the molten pool acoustic frequency band from the original acoustic signal and calculate the molten pool acoustic signal power;

[0109] The acoustic frequency band of the molten pool is 10-15kHz; therefore, the formula for calculating the acoustic signal power of the molten pool is:

[0110]

[0111] In the formula, P sig Where is the power of the acoustic signal in the molten pool, and K is the number of sampling points in this frequency band.

[0112] (5.4) Determine the signal-to-noise ratio of mechanical vibration noise based on the ratio of the power of the acoustic signal of the molten pool to the power of the mechanical vibration noise; determine the signal-to-noise ratio of powder impact noise based on the ratio of the power of the acoustic signal of the molten pool to the power of the powder impact noise.

[0113] Mechanical vibration noise signal-to-noise ratio:

[0114] Powder impact noise signal-to-noise ratio:

[0115] (5.5) Calculate the weight of the mechanical vibration noise signal-to-noise ratio in the sum of the mechanical vibration noise signal-to-noise ratio and the powder impact noise signal-to-noise ratio, and use it as the basic weight.

[0116]

[0117] α is the base weight.

[0118] (6) Compensate the basic weights according to the path risk coefficient to generate vibration channel weights.

[0119] Specifically, the mechanical vibration noise weight can be generated according to the following formula:

[0120] α final = a + bR + c(α - d)

[0121] In the formula, α final The values ​​represent the mechanical vibration noise weights, where a, b, c, and d are compensation coefficients, R is the path risk coefficient, and α is the base weight.

[0122] In practical applications, α can be set. final =0.6+0.2R+0.2(α-0.5).

[0123] (7) Remove the mechanical vibration noise component from the original acoustic signal and perform bandpass filtering on the remaining signal in the molten pool acoustic band to obtain the first signal.

[0124] Step (7) specifically includes:

[0125] (7.1) Perform EMD intrinsic mode decomposition on the original acoustic signal and extract the intrinsic mode components where the mechanical vibration noise is located;

[0126] Among them, EMD intrinsic mode decomposition is performed on the original acoustic signal:

[0127]

[0128] y(t) is the original acoustic signal, IMF k (t) represents the k-th IMF intrinsic mode component, r s (t) is the residual component. The mechanical vibration noise component is determined by the dominant frequency of the mechanical vibration noise, for example... 300, 400, and 600 Hz correspond to IMF3, IMF4, and IMF5 respectively. The corresponding component is taken as the intrinsic mode component (IMF) of mechanical vibration noise. vib .

[0129] (7.2) Remove the intrinsic mode components containing mechanical vibration noise from the original acoustic signal:

[0130] x(t)=y(t)-∑IMF vib

[0131] (7.3) The remaining signal is bandpass filtered in the molten pool acoustic frequency band to obtain the first signal:

[0132] s vib (t)=BPF(x(t); f c ±Δf=10kHz±2.5kHz)

[0133] s vib (t) represents the first signal, specifically as follows: Figure 3 As shown, BPF represents bandpass filtering.

[0134] (8) Obtain the time spectrum of the original acoustic signal, generate a time-frequency mask based on the dynamic powder impact noise mask, suppress the powder impact noise in the time spectrum using the time-frequency mask, transform it to the time domain, and use it as the second signal.

[0135] Step (8) specifically includes:

[0136] (8.1) Perform STFT transformation on the original acoustic signal to obtain the time spectrum S(t,f) of the original acoustic signal;

[0137] (8.2) Based on the dynamic powder impact noise mask, the time-frequency mask is calculated according to the following formula:

[0138]

[0139] In the formula, M(t,f) is the time-frequency mask, η(Q) is the powder impact noise mask, and S(t,f) is the time spectrum;

[0140] (8.3) Multiply the time-frequency mask by the time-frequency spectrum to obtain the time-frequency spectrum for suppressing powder impact noise: S′(t,f)=M(t,f)·S(t,f);

[0141] (8.4) Perform an inverse STFT transform on the time spectrum S′(t,f) for suppressing powder impact noise to obtain the second signal s in the time domain. powder (t) = ISTFT[S′(t,f)], such as Figure 4 As shown.

[0142] (9) The first signal and the second signal are weighted and fused by the vibration channel weight to obtain the denoised and reconstructed molten pool acoustic signal.

[0143] s out (t)=α final s vib (t)+(1-α final )·s powder (t)

[0144] In the formula, s out (t) represents the denoised and reconstructed molten pool acoustic signal, such as Figure 5 As shown, α final This indicates the weight of the vibration channel.

[0145] To further illustrate the superiority of this method, such as Figure 3 As shown, the original signal is an acoustic signal collected under stable cladding conditions. The signal characteristics fluctuate steadily, with an amplitude between 0.3 and 0.5. After processing, the signal curve is smoothed, and the amplitude is between 0.2 and 0.3. This patent reduces the noise of the original signal by 40% while retaining 92% of its characteristics. Figure 4 As shown, the original signal is an acoustic signal collected under corner cladding, characterized by severe jitter and spikes > 0.8. After processing, the signal fluctuation is stabilized, with an amplitude of 0.25 ± 0.05. This patent reduces the abnormality of the original signal by 70%. It can be seen that the original signal achieves a significant noise reduction effect through the method of this invention.

[0146] After obtaining the denoised and reconstructed acoustic signal of the molten pool, it can be used for defect feature identification, such as inputting it into the molten pool condition monitoring system to realize online defect feature identification, including molten pool porosity identification (10-15kHz feature analysis) or online alarm for spatter defects.

[0147] Example 2

[0148] This invention provides an online noise reduction system for laser-deposited acoustic signals based on dynamic adaptive filtering, comprising:

[0149] The acoustic signal acquisition subsystem is used to acquire the raw acoustic signals during the laser directional energy deposition process;

[0150] The process path dynamic prediction module is used to extract path features based on the laser deposition path, and determine the main frequency of mechanical vibration noise and the path risk coefficient based on the path features;

[0151] The noise feature analysis and channel scheduling module is used to obtain the powder flow rate of laser deposition at the current moment, calculate the powder impact noise frequency band and dynamic powder impact noise mask based on the powder flow rate; when the path risk coefficient is greater than or equal to a preset threshold, the basic weight is set to the preset weight; otherwise, based on the signals in the original acoustic signal located in the main frequency band of mechanical vibration noise, the frequency band of powder impact noise and the frequency band of molten pool sound, the mechanical vibration noise signal-to-noise ratio and the powder impact noise signal-to-noise ratio are calculated, and the proportion of the mechanical vibration noise signal-to-noise ratio in the total signal-to-noise ratio is used as the basic weight; the basic weight is compensated according to the path risk coefficient to generate vibration channel weight;

[0152] The mechanical vibration noise channel module is used to remove the mechanical vibration noise component from the original acoustic signal and perform bandpass filtering on the remaining signal in the molten pool acoustic frequency band to obtain the first signal;

[0153] The powder impact noise channel module is used to acquire the time spectrum of the original acoustic signal, and generate a time-frequency mask based on the dynamic powder impact noise mask. After suppressing the powder impact noise in the time spectrum using the time-frequency mask, the signal is transformed to the time domain and used as a second signal.

[0154] The signal denoising and reconstruction module is used to weight and fuse the first and second signals using vibration channel weights to obtain the denoised and reconstructed molten pool acoustic signal.

[0155] The acoustic signal acquisition subsystem includes a laser generator, a control system, a three-axis robotic arm, an inert gas supply device, a forming base, a directional microphone, a data acquisition card, and a powder conveying mechanism. The laser generator is positioned above the forming base. The control system is connected to both the laser generator and the three-axis robotic arm. The laser generator is connected to the powder conveying mechanism to achieve photo-powder coupling. The inert gas supply device is connected to the powder conveying mechanism. The directional microphone is fixed to the side of the forming base and is connected to the data acquisition card to acquire the raw acoustic signal.

[0156] The system provided in this embodiment of the invention can be used to execute the method provided in Embodiment 1 of the invention, and has the corresponding functions and beneficial effects of executing the method.

[0157] It is worth noting that in the embodiments of the above system, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of the present invention.

[0158] The embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art will clearly understand that each implementation can be achieved using software plus necessary general-purpose hardware platforms, or it can be implemented solely through hardware, as long as the function or purpose can be achieved.

[0159] It should be understood that the embodiments and descriptions above are only the principles, main features and advantages of the present invention. Various changes and modifications can be made to the present invention without departing from the spirit and scope of the invention, and all such changes and modifications fall within the protection scope of the present invention.

Claims

1. An online noise reduction method for laser-deposited acoustic signals based on dynamic adaptive filtering, characterized in that... Includes the following steps: (1) Acquire the raw acoustic signals during the laser-directed energy deposition process; (2) Extract path features based on the laser deposition path, and determine the main frequency of mechanical vibration noise and the path risk coefficient based on the path features; (3) Obtain the current powder flow rate of laser deposition, and calculate the powder impact noise frequency band and dynamic powder impact noise mask based on the powder flow rate; (4) When the path risk coefficient is greater than or equal to the preset threshold, set the basic weight to the preset weight; otherwise, proceed to step (5). (5) Based on the signals in the original acoustic signal located in the main frequency band of mechanical vibration noise, the frequency band of powder impact noise and the frequency band of molten pool sound, calculate the signal-to-noise ratio of mechanical vibration noise and the signal-to-noise ratio of powder impact noise, and take the proportion of mechanical vibration noise signal-to-noise ratio in the total signal-to-noise ratio as the basic weight. (6) Compensate the basic weights according to the path risk coefficient to generate vibration channel weights; (7) Remove the mechanical vibration noise component from the original acoustic signal and perform bandpass filtering on the remaining signal in the molten pool acoustic frequency band to obtain the first signal; the mechanical vibration noise component is determined by the main frequency of mechanical vibration noise; (8) Obtain the time spectrum of the original acoustic signal, and generate a time-frequency mask based on the dynamic powder impact noise mask. After suppressing the powder impact noise in the time spectrum using the time-frequency mask, transform it to the time domain and use it as the second signal. (9) The first signal and the second signal are weighted and fused by the vibration channel weight to obtain the denoised and reconstructed molten pool acoustic signal.

2. The online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 1, characterized in that, Step (2) specifically includes: (2.1) Obtain the laser deposition path, which includes the coordinates and motion velocity at each time point; (2.2) Extract path features based on the coordinates and movement speed at each time point, wherein the path features are path types; (2.3) Based on the path type, the main frequency of mechanical vibration noise is obtained according to the preset path type-mechanical vibration noise main frequency mapping rule; (2.4) Based on the path type, the path risk coefficient is obtained according to the preset path type-path risk coefficient mapping rule.

3. The online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 1, characterized in that, Step (3) specifically includes: (3.1) Obtain the powder flow rate of laser deposition at the current moment, and calculate the center frequency of powder impact noise based on the powder flow rate; (3.2) Determine the frequency band of powder impact noise based on the center frequency of powder impact noise; (3.3) Determine the dynamic powder impact noise mask based on the center frequency of the powder impact noise.

4. The online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 1, characterized in that, Step (5) specifically includes: (5.1) Extract the signal located in the main frequency band of mechanical vibration noise from the original acoustic signal and calculate the mechanical vibration noise power; (5.2) Extract the signal located in the powder impact noise frequency band from the original acoustic signal and calculate the powder impact noise power; (5.3) Extract the signal located in the molten pool acoustic frequency band from the original acoustic signal and calculate the molten pool acoustic signal power; (5.4) Determine the signal-to-noise ratio of mechanical vibration noise based on the ratio of the power of the acoustic signal of the molten pool to the power of the mechanical vibration noise; determine the signal-to-noise ratio of powder impact noise based on the ratio of the power of the acoustic signal of the molten pool to the power of the powder impact noise. (5.5) Calculate the weight of the mechanical vibration noise signal-to-noise ratio in the sum of the mechanical vibration noise signal-to-noise ratio and the powder impact noise signal-to-noise ratio, and use it as the basic weight.

5. The online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 1, characterized in that, Step (6) specifically includes: The basic weights are compensated based on the path risk coefficient, and the mechanical vibration noise weights are generated according to the following formula: α final =a+bR+c(α-d) In the formula, α final The values ​​represent the mechanical vibration noise weights, where a, b, c, and d are compensation coefficients, R is the path risk coefficient, and α is the base weight.

6. The online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 1, characterized in that, Step (7) specifically includes: (7.1) Perform EMD intrinsic mode decomposition on the original acoustic signal and extract the intrinsic mode components of the mechanical vibration noise; (7.2) Remove the intrinsic mode components containing mechanical vibration noise from the original acoustic signal; (7.3) The remaining signal is bandpass filtered in the molten pool acoustic band to obtain the first signal.

7. The online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 1, characterized in that, Step (8) specifically includes: (8.1) Perform STFT transformation on the original acoustic signal to obtain the time spectrum of the original acoustic signal; (8.2) Based on the dynamic powder impact noise mask, the time-frequency mask is calculated according to the following formula: In the formula, M(t,f) is the time-frequency mask, η(Q) is the powder impact noise mask, and S(t,f) is the time spectrum; (8.3) Multiply the time-frequency mask by the time-frequency spectrum to obtain the time-frequency spectrum that suppresses powder impact noise; (8.4) Perform inverse STFT transformation on the time spectrum of the powder impact noise suppression to obtain the second signal in the time domain.

8. The online noise reduction method for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 1, characterized in that, Step (9) specifically includes: The first and second signals are weighted using mechanical vibration noise weights and then fused according to the following formula to obtain the denoised and reconstructed molten pool acoustic signal: s out (t)=α final s vib (t)+(1-α final )·s powder (t) In the formula, s out (t) represents the denoised and reconstructed molten pool acoustic signal, α final Indicates the vibration channel weight, s vib (t), s powder (t) represent the first signal and the second signal, respectively.

9. An online noise reduction system for laser-deposited acoustic signals based on dynamic adaptive filtering, characterized in that, include: The acoustic signal acquisition subsystem is used to acquire the raw acoustic signals during the laser directional energy deposition process; The process path dynamic prediction module is used to extract path features based on the laser deposition path, and determine the main frequency of mechanical vibration noise and the path risk coefficient based on the path features; The noise feature analysis and channel scheduling module is used to obtain the powder flow rate of laser deposition at the current moment, calculate the powder impact noise frequency band and dynamic powder impact noise mask based on the powder flow rate; when the path risk coefficient is greater than or equal to the preset threshold, the basic weight is set to the preset weight; otherwise, based on the signals in the original acoustic signal located in the main frequency band of mechanical vibration noise, the frequency band of powder impact noise and the frequency band of molten pool sound, the mechanical vibration noise signal-to-noise ratio and the powder impact noise signal-to-noise ratio are calculated, and the proportion of mechanical vibration noise signal-to-noise ratio in the total signal-to-noise ratio is used as the basic weight. The basic weights are compensated based on the path risk coefficient to generate vibration channel weights; The mechanical vibration noise channel module is used to remove the mechanical vibration noise component from the original acoustic signal and perform bandpass filtering on the remaining signal in the molten pool acoustic frequency band to obtain the first signal; The powder impact noise channel module is used to acquire the time spectrum of the original acoustic signal, and generate a time-frequency mask based on the dynamic powder impact noise mask. After suppressing the powder impact noise in the time spectrum using the time-frequency mask, the signal is transformed to the time domain and used as a second signal. The signal denoising and reconstruction module is used to weight and fuse the first and second signals using vibration channel weights to obtain the denoised and reconstructed molten pool acoustic signal.

10. The online noise reduction system for laser deposition acoustic signals based on dynamic adaptive filtering according to claim 9, wherein the acoustic signal acquisition subsystem includes a laser generator, a control system, a three-axis robotic arm, an inert gas supply device, a forming base, a directional microphone, a data acquisition card, and a powder conveying mechanism; wherein, The laser generator is positioned above the forming base. The control system is connected to both the laser generator and the three-axis robotic arm. The laser generator is connected to the powder conveying mechanism to achieve photo-powder coupling. The inert gas supply device is connected to the powder conveying mechanism. The directional microphone is fixed to the side of the forming base and is connected to the data acquisition card to acquire the original sound signal.

Images