Plywood defect detection system and detection method

Abstract

The invention relates to the technical field of nondestructive testing, in particular to a plywood defect detection system and method, and the plywood defect detection system comprises a multi-mode sensor array, a central processing unit, a dynamic feedback control unit and a mechanical arm. The multi-mode sensor array cooperatively works with the multi-frequency ultrasonic transmitting-receiving pair through the acoustic vibration coupling sensor, and the frequency band combination and the frequency sweeping range are adjusted in real time in combination with the dynamic feedback control unit, so that accurate detection is realized. The method solves or at least relieves the problem that the sensitivity is insufficient when a traditional plywood defect detection method is used for distinguishing a tiny degumming area and normal bonding textures, and meanwhile the problem that frequency response overlapping of non-uniform adhesive layer curing and wood natural textures cannot be effectively distinguished through traditional acoustic detection is solved; the invention provides a plywood defect detection system and a detection method.

Description

Technical Field

[0001] The present invention belongs to the technical field of non-destructive testing, and in particular relates to a plywood defect detection system and a detection method. Background Art

[0002] Currently, in ultrasonic detection of internal debonding defects in plywood, due to insufficient sensitivity, the commonly used single-frequency inspection has difficulty distinguishing some very small debonding areas from normal bonding textures, resulting in loss of details and affecting inspection accuracy. To improve inspection accuracy, it is necessary to use a multi-band combination or other supplementary means to increase signal details.

[0003] Therefore, traditional acoustic inspection methods can be used as a supplementary means to improve inspection results by checking the curing condition of the glue layer. However, traditional acoustic inspection is prone to misjudgment due to frequency overlap when distinguishing between uneven curing of the glue layer and the natural texture of the wood. Due to the similarity of frequency responses, curing defects and the wood texture are difficult to separate, resulting in inaccurate inspection results. However, traditional inspection methods are not suitable for complex board structures. Due to the material diversity and internal unevenness of plywood, inspection techniques must also be diverse. For example, uneven glue layer thickness and the natural texture of the wood make it more difficult to distinguish signal details, resulting in large fluctuations in results. Therefore, traditional inspection methods are prone to deviations for complex boards, resulting in equipment instability and low efficiency.

[0004] For example, the invention patent with publication number CN117571829A discloses a plywood surface defect detection device. To address the problems existing in the prior art, the following solution is proposed: it includes a conveying shell, the side inner wall of the conveying shell is rotatably connected to two conveying rollers, the side outer wall of the conveying roller is sleeved with a conveyor belt, and one side outer wall of the conveying shell is fixedly connected to a conveying motor for driving the conveying rollers to rotate. The top outer wall of the conveying shell is fixedly connected to an outer shell, and a downward pressure shell that can slide up and down is provided inside the outer shell. The top outer wall of the downward pressure shell is fixedly connected to an adjustment cylinder, and the top outer wall of the outer shell is rotatably connected to a downward pressure knob. When there are scratches or bumps on the plywood surface, gas sprayed on the plywood surface will produce a buzzing sound, which will cause the sound sensor to detect changes in the sound waves. The staff can then judge the surface defects of the plywood by the changes in the sound waves, which is more accurate, convenient and quick. This invention uses a sound sensor to detect surface scratches / bumps, relying on gas injection to produce a buzzing sound, and only solves the problem of single-side defect detection.

[0005] For example, the utility model patent with publication number CN201488965U discloses a plywood defect detection hammer, which includes a hammer body and a handle. Its characteristics are: a columnar piezoelectric ceramic is vertically embedded in the hammer body, with the lower end exposed as a striking point. A neon tube is installed on the hammer body, and the two poles of the neon tube are electrically connected to the upper end of the piezoelectric ceramic and the output line respectively. The hammer body and the handle are connected by a spring. When the detection hammer strikes the plywood, if there is a hollow or bark inside the board, the impact force is buffered and weakened, the output voltage of the piezoelectric ceramic column is low, and the neon tube does not emit light or the light becomes weak. By observing and distinguishing, the internal quality of the artificial wood board can be understood. This utility model has the advantages of simple structure and intuitive display. The invention is a mechanical detection hammer that determines internal hollowness / bark by the change of piezoelectric ceramic voltage. It has no active scanning capability and low accuracy. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings of the existing technology, solve or at least alleviate the problem of insufficient sensitivity of traditional plywood defect detection methods in distinguishing between tiny debonding areas and normal bonding textures, and at the same time improve the problem of frequency response overlap between traditional acoustic detection and the inability to effectively distinguish between uneven curing of the glue layer and the natural texture of the wood, and provide a plywood defect detection system and detection method.

[0007] To achieve the above objectives, the present invention provides the following technical solutions: a plywood defect detection system, comprising:

[0008] Multimodal sensor array, supported by a ring base,

[0009] a central processing unit connected to the sensor array via a pre-processing circuit;

[0010] Dynamic feedback control unit, integrated into the central processing unit, used to adjust sensor parameters in real time;

[0011] A robotic arm carrying a multimodal sensor array to achieve three-dimensional scanning;

[0012] Among them, the multimodal sensor array includes:

[0013] The acoustic-vibration coupling sensors, which are evenly spaced on the outer ring, are used to collect vibration signals by attaching to the surface of the plywood;

[0014] The multi-frequency ultrasonic transmitter-receiver pairs arranged symmetrically on the inner ring are used to transmit penetrating ultrasonic pulses.

[0015] Preferably, the spatial position of the acoustic vibration coupling sensor and the ultrasonic transmitter-receiver pair satisfies:

[0016] The acoustic vibration sensor is 60mm away from the center of the ring base, and the ultrasonic transmitter and receiver are 30mm away from the center.

[0017] The dynamic feedback control unit includes:

[0018] Frequency band reorganization module, which adjusts the ultrasonic frequency band combination according to the defect probability distribution;

[0019] The sweep frequency focusing module dynamically narrows the acoustic vibration sweep frequency range according to the defect type.

[0020] A plywood defect detection method based on the above system comprises:

[0021] S1: Parameter-adaptive multimodal signal acquisition, with real-time adjustment of the acoustic vibration sweep range and ultrasonic frequency band combination through a dynamic feedback control unit;

[0022] S2: Time-frequency domain feature decoupling analysis, phase attenuation modeling and energy distribution construction of the acquired signal, and generation of fusion feature tensor;

[0023] S3: Defect classification and dynamic feedback optimization, perform texture benchmark comparison and defect judgment, and feed back the judgment results to steps S1 and S2 to form a closed loop.

[0024] In order to further realize the present invention, the following technical solutions may be preferably used:

[0025] Preferably, the step S1 includes:

[0026] S101: Receive the defect type distribution fed back in step S3;

[0027] S102: If the probability of debonding defects exceeds a first threshold, activating a high-frequency dominant ultrasonic mode;

[0028] S103: If the probability of uneven curing defects exceeds a second threshold, the focused acoustic vibration is swept within a narrow band range of 40 Hz to 800 Hz.

[0029] Preferably, step S2 includes:

[0030] S201: Dynamically select a window function based on the signal-to-noise ratio of the acoustic vibration signal: when the signal-to-noise ratio is lower than a set value, enable the Hanning window;

[0031] S202: Perform weighted energy calculation on the multi-frequency ultrasonic echoes, and the weight ratio is dynamically adjusted according to the frequency band instruction of step S1.

[0032] Preferably, step S3 includes:

[0033] S301: Calculate the Euclidean distance between the fused feature tensor and the texture feature library, and output the deviation;

[0034] S302: When the deviation is greater than the third threshold and the phase attenuation characteristic exceeds the fourth threshold, it is determined that the adhesive layer is unevenly cured;

[0035] S303: When the deviation is greater than the third threshold and the high-frequency echo attenuation rate exceeds the fifth threshold, it is determined to be a minor debonding.

[0036] Preferably, the step S3 further includes:

[0037] S304: Reversely input the fused feature tensor of the misjudged sample into the convolutional neural network of step S2, and update the convolution kernel weight online;

[0038] S305: Automatically expand the texture features of the normal area into the texture feature library.

[0039] Preferably, the ultrasonic frequency band optimization instruction generated in step S3 is fed back to the ultrasonic frequency band selection module in step S1 in real time;

[0040] The window function selection strategy generated in step S3 is fed back to the acoustic vibration signal modeling module in step S2 in real time.

[0041] Preferably, closed-loop feedback is triggered immediately after each single test is completed; the acoustic vibration sweep range is dynamically narrowed according to the distribution density of curing defects in continuous testing.

[0042] Preferably, the robotic arm adjusts the sensor pressing force in real time according to the signal-to-noise ratio data fed back in step S2;

[0043] When the signal-to-noise ratio is lower than the warning value, the contact pressure of the acoustic vibration sensor on the plywood surface is increased.

[0044] The beneficial effects of the present invention are:

[0045] The present invention adopts a multimodal sensor array and a dynamic feedback control method. By adjusting the ultrasonic frequency band combination and the acoustic vibration sweep range in real time, it can more easily capture the signals of small-area degumming areas under complex textures, greatly improve the detection sensitivity, and overcome the defect of detail loss in single-frequency ultrasonic detection; and, by decoupling and analyzing the time-frequency domain characteristics of the acoustic vibration signal and the ultrasonic echo signal, combined with the texture benchmark comparison technology, it can overcome the misjudgment problem caused by the overlap of frequency responses in traditional acoustic detection methods, and improve the ability to distinguish between uneven curing and the natural texture of wood; in addition, through closed-loop triggering and result-based parameter adaptive rules, dynamic adjustment of detection parameters is achieved, which can keep the detection effect stable, solve the problem of large sensitivity changes of the detection method under different materials and process conditions, and improve detection stability, especially in the edge area of ​​​​plywood. By increasing the sweep density and implementing vibration compensation, the detection accuracy and detection stability can be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 4 is a block diagram of the detection system of the present invention.

[0047] Figure 2Flow chart of step S1 of the detection method of the present invention.

[0048] Figure 3 Flow chart of step S2 of the detection method of the present invention.

[0049] Figure 4 Flow chart of step S3 of the detection method of the present invention.

[0050] Figure 5 Flowchart of the closed-loop feedback mechanism of the detection method of the present invention DETAILED DESCRIPTION

[0051] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0053] Example 1

[0054] Traditional detection methods are not suitable for complex board structures. Plywood presents challenges due to its diverse varieties and internal stratification. For example, uneven glue layer thickness and the natural variability of wood complicate signal interpretation, leading to significant fluctuations in test results.

[0055] like Figure 1As shown, this embodiment discloses a plywood defect detection system, wherein the plywood defect detection system includes: a multimodal sensor array, the multimodal sensor array is arranged at the end of the robotic arm and fixed to the robotic arm through a threaded connection; a central processing unit, the central processing unit is connected to the multimodal sensor array through a high-speed signal conditioning module, the central processing unit is used to receive signals transmitted by the high-speed signal conditioning module, and filter and amplify the signals; a dynamic feedback control unit, the dynamic feedback control unit is built into the central processing unit, the dynamic feedback control unit is connected to the multimodal sensor array, so that the central processing unit adjusts the parameters of the sensors in the multimodal sensor array in real time through the dynamic feedback control unit; the robotic arm includes six joints, each of the joints is connected to a servo motor, and each of the servo motors is connected to the central processing unit via a CAN bus.

[0056] The multimodal sensor array includes several equally spaced acoustic-vibration coupling sensors arranged in an outer circle and symmetrically distributed multi-frequency ultrasonic transmitter-receiver pairs arranged in an inner circle. The acoustic-vibration coupling sensors include a spring sheet and an adhesive sheet adhered to the plywood surface for collecting vibration signals. The multi-frequency ultrasonic transmitter-receiver pair is connected to an annular base via conductive silicone and is used to transmit penetrating ultrasonic pulses. The spatial positioning of the acoustic-vibration coupling sensors and the multi-frequency ultrasonic transmitter-receiver pair satisfies the following conditions: the acoustic-vibration coupling sensors have a radius of 60 mm from the center of the annular base, and the multi-frequency ultrasonic transmitter-receiver pair has a radius of 30 mm from the center of the annular base. This arrangement of the two sensors allows them to detect different locations on the plywood and complement each other.

[0057] The dynamic feedback control unit includes a frequency band reorganization module and a sweep frequency focusing module. Among them, the frequency band reorganization module selectively enables or disables the ultrasonic frequency band combination according to the defect probability distribution, and the sweep frequency focusing module selectively narrows the acoustic vibration sweep frequency range according to the defect type. In this embodiment, when a high-risk area for debonding is detected, the frequency band reorganization module enables the high-frequency dominant mode, in which the 1MHz pulse accounts for 70%-80%, the 200kHz pulse accounts for 20%-25%, and the 50kHz pulse accounts for 0-5%. In the concentrated area of ​​uneven curing, the frequency band reorganization module enables the medium and low frequency mode, in which the 200kHz pulse accounts for 50%, the 50kHz pulse accounts for 40%, and the 1MHz pulse accounts for 10%. When the curing defects are detected three times in a row, the sweep frequency focusing module narrows the sweep frequency range from 20Hz-5kHz to 40Hz-800Hz, and automatically increases the sweep frequency density in the edge area of ​​the plate to 1.5 times that of the center area.

[0058] The present invention uses a multimodal sensor array combined with dynamic feedback control to adjust the ultrasonic frequency band combination and acoustic vibration sweep range in real time during the detection process, effectively capturing signals of tiny debonding areas under complex textures and significantly improving detection sensitivity. Compared with traditional single-frequency ultrasonic detection methods, the present invention avoids the problem of detail loss through multi-band combination and weight distribution strategy.

[0059] Example 2

[0060] like Figure 2-Figure 5 As shown, in order to better enable the system of embodiment 1 to work better, this embodiment provides a plywood defect detection method, which includes:

[0061] S1: Parameter-adaptive multimodal signal acquisition, with real-time adjustment of the acoustic vibration sweep range and ultrasonic frequency band combination through a dynamic feedback control unit;

[0062] S2: Time-frequency domain feature decoupling analysis, phase attenuation modeling and energy distribution construction of the acquired signal, and generation of fusion feature tensor;

[0063] S3: Defect classification and dynamic feedback optimization, perform texture benchmark comparison and defect judgment, and feed back the judgment results to steps S1 and S2 to form a closed loop.

[0064] The step S1 comprises:

[0065] S101: Receive the defect probability distribution map fed back in step S3, where the debonding defect area is marked with a heat map;

[0066] S102: If the debonding defect probability exceeds the first threshold, in the high-risk area where the debonding probability is greater than 60%, the high-frequency dominant ultrasonic mode is activated, with 1 MHz pulses accounting for 70%-80%, 200 kHz pulses accounting for 20%-25%, and 50 kHz pulses accounting for 0-5%;

[0067] S103: If the probability of uneven curing defects exceeds the second threshold, in the concentrated area where the probability of uneven curing is >55%, the medium and low frequency mode is enabled, with 200kHz pulses accounting for 50%, 50kHz pulses accounting for 40%, and 1MHz pulses accounting for 10%. The focused acoustic vibration sweep is in the narrow band range of 40Hz-800Hz. When curing defects are detected three times in a row, the sweep range is narrowed from 20Hz-5kHz to 40Hz-800Hz. In the edge area of ​​the plate, the sweep density is automatically increased to 1.5 times that of the center area.

[0068] The step S2 comprises:

[0069] S201: Dynamically select a window function based on the signal-to-noise ratio of the acoustic vibration signal: When the signal-to-noise ratio of the acoustic vibration signal is less than 30dB, the Hanning window is forcibly enabled for short-time Fourier transform; when the signal-to-noise ratio is ≥30dB, a rectangular window is used to retain high-frequency components;

[0070] S202: Calculate weighted energy of the multi-frequency ultrasonic echoes, with the weight ratio dynamically adjusted according to the frequency band instruction of step S1;

[0071] In high-frequency dominant mode, the energy matrix weight distribution is: 1MHz echo weight 0.7-0.8, 200kHz echo weight 0.15-0.25, 50kHz echo weight 0.05;

[0072] In the medium and low frequency mode, the energy matrix weight distribution is: 200kHz echo weight 0.5, 50kHz echo weight 0.4, 1MHz echo weight 0.1.

[0073] The step S3 comprises:

[0074] S301: Calculate the Euclidean distance between the fused feature tensor and the texture feature library, and dynamically calculate the deviation;

[0075] The pine wood texture benchmark library uses a preset Euclidean distance threshold of T1 = 0.35, the birch wood texture benchmark library uses a preset Euclidean distance threshold of T1 = 0.28, and the mixed wood panel uses a weighted threshold of T1 = 0.32 ± 0.03;

[0076] S302: Determination of curing defects: when the deviation is greater than T1 and the phase delay is greater than 120% of the reference value, it is determined that the curing is uneven, and a second scanning verification is performed on the determination area, with the acoustic vibration sweep frequency focused to 200Hz±50Hz;

[0077] S303: Debonding defect determination: When the deviation is greater than T1 and the 1MHz echo attenuation rate is greater than 45dB / m, it is determined to be a minor debonding defect. For debonding less than 0.5mm, 1.5MHz enhanced scanning is enabled.

[0078] The step S3 further includes:

[0079] S304: Processing of misclassified samples: Collecting feature tensors of confused samples with deviations within the range of T1 ± 0.05. Each update injects 50-100 groups of samples into the CNN training set. CNN is optimized online, and the convolution kernel weights are updated using momentum optimization. Every 200 detections trigger the full network gradient descent.

[0080] S305: The texture library is dynamically expanded. Texture features of normal areas are stored in the library after wavelet denoising. The upper limit of the texture library capacity is 5,000 groups, and the LRU elimination mechanism is adopted.

[0081] The ultrasonic frequency band optimization instruction generated in step S3 is fed back to the ultrasonic frequency band selection module in step S1 in real time. The ultrasonic frequency band instruction contains a priority mark: debonding risk > curing risk > conventional scanning. The ultrasonic frequency sweep instruction is accompanied by regional coordinates, and the positioning error of the robotic arm is <0.1mm.

[0082] The window function selection strategy generated in step S3 is fed back to the acoustic vibration signal modeling module in step S2 in real time. The window function strategy is directly transmitted to the FPGA through the DMA channel, and the weight distribution table is updated in real time in a 128-bit vector format.

[0083] Closed-loop feedback is triggered immediately after each single test is completed. The feedback process is started within 50ms after the completion of a single test. The closed loop is triggered for each test in the key area. The acoustic vibration sweep range is dynamically narrowed according to the distribution density of curing defects in continuous testing. When the density of curing defects is greater than 5 / ㎡, the sweep range narrowing rate is increased to 2 times. After 10 consecutive correct judgments, the threshold T1 is automatically relaxed to 110% of the original value.

[0084] The robotic arm adjusts the sensor pressing force in real time according to the signal-to-noise ratio data fed back in step S2, and increases the contact pressure of the acoustic vibration sensor on the plywood surface when the signal-to-noise ratio is lower than the warning value;

[0085] The sensor's pressing force is controlled in stages. When the signal-to-noise ratio is 30-35dB, the contact pressure is maintained at 5N±0.5N. When the signal-to-noise ratio is 25-30dB, the pressure is increased to 7N±0.8N. When the signal-to-noise ratio is less than 25dB, the pressure is increased stepwise to 10N, with a maximum of three increments.

[0086] When the detection speed is greater than 0.5m / s, the anti-phase vibration cancellation algorithm is enabled, and piezoelectric ceramic dampers are configured at the joints of the robotic arm with an attenuation amplitude of ≥60%.

[0087] Example 3

[0088] In order to enable relevant personnel in this technical field to better understand and implement the present invention, the implementation principle of the present invention is further described in detail below in conjunction with a specific application scenario of the present invention.

[0089] First, the robotic arm is initially positioned at the upper left corner of the plywood surface. Driven by a servo motor, the arm moves along its path. A vibroacoustic coupling sensor and a multi-frequency ultrasonic transmitter-receiver pair, attached to the plywood surface, collect signals simultaneously with the arm's movement. The vibroacoustic coupling sensor contacts the plywood surface via a spring plate, collecting vibration signals. The multi-frequency ultrasonic transmitter-receiver pair is fixed to a ring-shaped base via conductive silicone, alternatingly transmitting and receiving ultrasonic pulses. When the robotic arm reaches the edge of the plywood, the dynamic feedback control unit automatically increases the vibroacoustic sweep frequency range based on a built-in algorithm. The sweep frequency focusing module increases the sweep frequency density to 1.5 times that of the center area to compensate for insufficient vibroacoustic signal acquisition at the edge. The vibration signals collected by the vibroacoustic coupling sensor are then transmitted to the central processing unit via a high-speed signal conditioning module, which filters and amplifies the received signals. The signals then enter the time-frequency domain feature decoupling analysis module to generate a fused feature tensor.

[0090] When defects cluster at the intersection of the plywood grain, such as in pine and birch plywood, the dynamic feedback control unit resets the rapid reassembly module's instructions based on the defect probability distribution map, generating different ultrasonic frequency band combinations. If the ultrasonic test results indicate a debonding probability greater than 60%, the high-frequency dominant combination mode is activated, with 1 MHz pulses accounting for 70%-80%, 200 kHz pulses accounting for 20%-25%, and 50 kHz pulses accounting for 0-5%. This frequency combination can detect small debonding areas less than 1 mm wide while avoiding loss of detail caused by high-frequency signal attenuation. If uneven curing occurs in the plywood, the dynamic feedback control unit activates a combination of medium and low frequencies. This reduces misjudgment caused by overlapping frequency responses of different frequency signals, especially when uneven curing of the glue layer resembles the natural wood grain. If three consecutive curing defects are detected, the sweep focusing module narrows the sweep range and verifies the identified area with a second sweep, focusing the acoustic and vibration sweep to 200 Hz ± 50 Hz.

[0091] When the signal-to-noise ratio (SNR) of the acoustic signal is less than 30dB, a Hanning window is used for short-time Fourier transform to reduce the impact of noise. When the SNR of the acoustic signal is greater than or equal to 30dB, a rectangular window is used to preserve the downlink signal components. Ultrasonic energy weighting assigns energy weights based on the frequency band pattern. When the high and medium frequency bands dominate, the 1MHz echo weight is 0.7-0.8, the 200kHz echo weight is 0.15-0.25, and the 50kHz echo weight is 0.05. Deviation is based on a texture reference library, calculating the degree of deviation of the tested board to classify defects. The Euclidean distance threshold T1 for pine wood is 0.35; the Euclidean distance threshold T1 for birch wood is 0.28; and the Euclidean distance threshold T1 for mixed boards is 0.32±0.03. If the deviation is greater than T1 and the 1mHz echo attenuation rate is greater than 45dB / m, it is a minor debonding defect. For debonding less than 0.5mm, 1.5mHz enhanced scanning is implemented; if the deviation l is greater than T1 and the phase extension is greater than 120% of the baseline value, it is a curing defect and a second scan is performed on the judgment area for verification.

[0092] When the closed loop is triggered, feedback is initiated within 50ms after a single test is completed. The robot adjusts the next test parameters based on the feedback results, such as: the curing defect density is greater than 5 / m 2 When the signal-to-noise ratio is 30-35dB, the contact pressure is 5N±0.5N; when the signal-to-noise ratio is 25-30dB, the contact pressure is 7N±0.8N; when the signal-to-noise ratio is less than 25dB, the contact pressure is increased stepwise to 10N, up to 3 times; when the detection speed is greater than 0.5m / s, the anti-phase vibration cancellation algorithm is enabled, and piezoelectric ceramic dampers are installed at the joints of the manipulator to attenuate the amplitude greater than or equal to 60%.

[0093] This method achieves precise internal defect detection for plywood, addressing the inapplicability of conventional methods for plywood with complex textures and multiple layers. The entire detection process is driven by closed-loop triggering and adaptive parameter rules, enabling dynamic optimization and ensuring consistent detection results for plywood made of different materials and processes.

[0094] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A plywood defect detection system, characterized in that: include: Multimodal sensor array, supported by a ring base, a central processing unit connected to the sensor array via a pre-processing circuit; Dynamic feedback control unit, integrated into the central processing unit, used to adjust sensor parameters in real time; A robotic arm carrying a multimodal sensor array to achieve three-dimensional scanning; Among them, the multimodal sensor array includes: The acoustic-vibration coupling sensors, which are evenly spaced on the outer ring, are used to collect vibration signals by attaching to the surface of the plywood; The multi-frequency ultrasonic transmitter-receiver pairs arranged symmetrically on the inner ring are used to transmit penetrating ultrasonic pulses.

2. The system according to claim 1, characterized in that , the spatial position of the acoustic vibration coupling sensor and the ultrasonic transmitter-receiver pair satisfies: The acoustic vibration sensor is 60mm away from the center of the ring base, and the ultrasonic transmitter and receiver are 30mm away from the center. The dynamic feedback control unit includes: Frequency band reorganization module, which adjusts the ultrasonic frequency band combination according to the defect probability distribution; The sweep frequency focusing module dynamically narrows the acoustic vibration sweep frequency range according to the defect type.

3. A plywood defect detection method based on any one of claims 1-2, characterized in that: include: S1: Parameter-adaptive multimodal signal acquisition, with real-time adjustment of the acoustic vibration sweep range and ultrasonic frequency band combination through a dynamic feedback control unit; S2: Time-frequency domain feature decoupling analysis, phase attenuation modeling and energy distribution construction of the acquired signal, and generation of fusion feature tensor; S3: Defect classification and dynamic feedback optimization, perform texture benchmark comparison and defect judgment, and feed back the judgment results to steps S1 and S2 to form a closed loop.

4. The method according to claim 3, characterized in that The step S1 comprises: S101: Receive the defect type distribution fed back in step S3; S102: If the probability of debonding defects exceeds a first threshold, activating a high-frequency dominant ultrasonic mode; S103: If the probability of uneven curing defects exceeds a second threshold, the focused acoustic vibration is swept within a narrow band range of 40 Hz to 800 Hz.

5. The method according to claim 3, characterized in that The step S2 comprises: S201: Dynamically select a window function based on the signal-to-noise ratio of the acoustic vibration signal: when the signal-to-noise ratio is lower than a set value, enable the Hanning window; S202: Perform weighted energy calculation on the multi-frequency ultrasonic echoes, and the weight ratio is dynamically adjusted according to the frequency band instruction of step S1.

6. The method according to claim 3, characterized in that The step S3 comprises: S301: Calculate the Euclidean distance between the fused feature tensor and the texture feature library, and output the deviation; S302: When the deviation is greater than the third threshold and the phase attenuation characteristic exceeds the fourth threshold, it is determined that the adhesive layer is unevenly cured; S303: When the deviation is greater than the third threshold and the high-frequency echo attenuation rate exceeds the fifth threshold, it is determined to be a minor debonding.

7. The method according to claim 6, characterized in that The step S3 further comprises: S304: Reversely input the fused feature tensor of the misjudged sample into the convolutional neural network of step S2, and update the convolution kernel weight online; S305: Automatically expand the texture features of the normal area into the texture feature library.

8. The method according to claim 3, characterized in that , the ultrasonic frequency band optimization instruction generated in step S3 is fed back to the ultrasonic frequency band selection module in step S1 in real time; The window function selection strategy generated in step S3 is fed back to the acoustic vibration signal modeling module in step S2 in real time.

9. The method according to claim 3, characterized in that Closed-loop feedback is triggered immediately after each single test is completed; the acoustic vibration sweep range is dynamically narrowed according to the distribution density of solidification defects in continuous testing.

10. The method according to claim 3, characterized in that The robotic arm adjusts the sensor pressing force in real time according to the signal-to-noise ratio data fed back in step S2; When the signal-to-noise ratio is lower than the warning value, the contact pressure of the acoustic vibration sensor on the plywood surface is increased.

Images