A method and system for detecting and locating foaming defects of a sandwich foam board

Abstract

The present invention relates to the field of nondestructive testing technology, and specifically discloses a method and system for detecting and locating foaming defects of a sandwich foam board, including: S1: equipment preparation and environmental control; S2: data acquisition, using an infrared thermal imager to fully scan the sandwich board and record temperature distribution data; S3: normalizing infrared thermal imaging and electromagnetic induction data, extracting key features such as temperature gradient from infrared thermal imaging data, and extracting key features such as electromagnetic induction intensity change from electromagnetic induction data; S4: establishing a sandwich foam board defect recognition and classification model, and using the data of the sandwich foam board with marked defects to train a neural network model; S5: inputting the preprocessed data into the trained sandwich foam board defect recognition and classification model. By combining advanced infrared thermal imagers with electromagnetic induction technology, and using deep learning models for intelligent recognition and classification, not only the detection accuracy is improved, but also good adaptability and real-time performance are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of nondestructive testing, and in particular to a method and system for detecting and locating foaming defects of a sandwich foam board. Background Art

[0002] Sandwich foam panels, especially polyurethane sandwich foam panels, are widely used in many fields such as construction and cold chain logistics due to their light weight, high strength and good thermal insulation performance. However, during the production process, due to uneven foaming, void-type defects are easily generated, which directly affect the thermal insulation effect and structural safety of sandwich foam panels. Traditional detection methods, such as the knock echo method, rely on the experience of workers, have large errors, and are inefficient. Therefore, the development of new detection technologies is crucial to improve the quality control and safety of sandwich foam panels.

[0003] In the prior art, an ultrasonic guided wave excitation probe, an ultrasonic guided wave receiving probe, a motor and a cross guide slide are used to fit a distance-amplitude change curve through the received ultrasonic guided wave A-scan signal, and the location of the foaming defect is determined by comparing the change rate of the detection amplitude curve with the fitting amplitude curve;

[0004] However, the prior art has the following problems:

[0005] 1. When the sandwich foam board is thick or multi-layered, the ultrasonic guided wave technology will have great limitations. In addition, the detection ability of ultrasonic guided waves will decrease for deep or obscured defects;

[0006] 2. Ultrasonic guided wave technology cannot accurately identify the type of defects when identifying defects. In addition, the accuracy of detection will be reduced for defects of small size and irregular shape;

[0007] 3. For sandwich foam panels of different sizes and shapes, the positioning equipment has poor adaptability, which is not conducive to rapid deployment and movement on site;

[0008] 4. Ultrasonic guided wave technology will generate greater noise pollution when used. Summary of the invention

[0009] The embodiment of the present invention provides a method and system for detecting and locating foaming defects of a sandwich foam board, so as to overcome the above-mentioned technical problems existing in the prior art.

[0010] A method for detecting and locating foaming defects of a sandwich foam board comprises the following steps:

[0011] S1: Equipment preparation and environmental control;

[0012] Prepare infrared thermal imagers and electromagnetic induction instruments to ensure that the resolution and sensitivity of the equipment meet the detection requirements;

[0013] Control the testing environment to ensure stable ambient temperature and reduce the impact of ambient temperature fluctuations on test results;

[0014] Clean the surface of the sandwich foam board to remove dust and oil, and ensure that there is no ferromagnetic material on the surface of the sandwich foam board to avoid interference with electromagnetic induction signals;

[0015] S2: Data acquisition, use an infrared thermal imager to fully scan the sandwich panel and record the temperature distribution data;

[0016] Place the electromagnetic induction instrument on the surface of the sandwich panel, move it along the panel surface, and record the changes of the electromagnetic induction signal;

[0017] S3: Normalize the infrared thermal imaging and electromagnetic induction data to eliminate the influence of different equipment and environment, extract key features such as temperature gradient from the infrared thermal imaging data, and extract key features such as electromagnetic induction intensity change from the electromagnetic induction data;

[0018] S4: Establish a defect recognition and classification model for sandwich foam boards, and use the data of sandwich foam boards with marked defects to train the neural network model so that it can recognize and classify defects;

[0019] S5: Input the preprocessed data into the trained sandwich foam board defect recognition and classification model, and output the type and precise location of the defect.

[0020] Preferably, in S2, the following is specifically included:

[0021] Use an infrared thermal imager to fully scan the sandwich panel and record the temperature distribution data; the infrared thermal imager moves along the predetermined scanning path to ensure that the entire surface of the sandwich foam board is covered; the scanning speed is controlled at 0.1-0.5 m / s to ensure the stability and accuracy of the collected data, and the working band of the infrared camera is 7.5-13μm;

[0022] Place the electromagnetic induction instrument on the surface of the sandwich panel, move along the panel surface, and record the changes of the electromagnetic induction signal. The electromagnetic induction instrument moves along the predetermined scanning path to ensure that the entire surface of the sandwich foam board is covered. The scanning speed is controlled at 0.1-0.5 m / s to ensure the stability and accuracy of the collected data.

[0023] According to the thickness and defect type of the sandwich foam board, multiple scans are performed to improve the comprehensiveness and accuracy of the detection.

[0024] Preferably, in S2, scanning path planning is also included, and the scanning paths of the infrared thermal imager and the electromagnetic induction instrument are planned according to the size and shape of the sandwich foam board to ensure that the entire surface of the sandwich foam board is covered, which specifically includes the following steps:

[0025] Initial planning: Generate the initial scanning path according to the size and shape of the sandwich foam board;

[0026] Dynamic adjustment: During the scanning process, the scanning path is dynamically adjusted based on the real-time feedback of infrared thermal imaging data and electromagnetic induction data to ensure that all key areas are covered;

[0027] Path optimization: Use ant colony algorithm to optimize the scanning path and reduce scanning time. The mathematical model of ant colony algorithm is:

[0028] ;

[0029] in, is the incremental pheromone on path (i, j);

[0030] Q is a constant, is the total path length of ant k.

[0031] Preferably, in S3, a smoothing filter is used to filter the collected temperature data and electromagnetic induction signal to reduce noise interference;

[0032] Among them, key features such as temperature gradient, temperature difference and temperature fluctuation are extracted from infrared thermal imaging data. The temperature gradient calculation formula is:

[0033] ;

[0034] Where T is the temperature, x and y are the spatial coordinates;

[0035] and Respectively represent the partial derivatives of temperature T in the x and y directions;

[0036] Extract key features such as electromagnetic induction intensity change and electromagnetic field distribution from electromagnetic induction data; the calculation formula for electromagnetic induction intensity change is:

[0037] ;

[0038] Among them, B(t) is the electromagnetic induction intensity at the current time point, is the electromagnetic induction intensity at the previous time point, For the time interval.

[0039] Preferably, in S4, a convolutional neural network (CNN) is used to extract high-level features from the preprocessed data; the data is feature extracted and classified through convolutional layers, pooling layers and fully connected layers, and the long short-term memory network (LSTM) is used to capture the features of time series data to enhance the recognition ability of defect types and locations; the gradient vanishing and gradient exploding problems in long sequence data are solved through a gating mechanism, and the extracted features are classified using a support vector machine (SVM) to improve the classification accuracy; the optimal classification hyperplane is found by maximizing the classification interval;

[0040] Among them, the convolutional neural network and the long short-term memory network are combined to form a hybrid model. The convolutional neural network is used to extract spatial features, and then the extracted features are input into the long short-term memory network to capture time series features and improve the detection ability of dynamically changing defects. The output of the convolutional neural network is used as the input of the long short-term memory network, and feature fusion is performed through a gating mechanism;

[0041] A multi-scale convolution layer is introduced into the convolutional neural network, and multi-scale features are extracted by convolution kernels of different sizes to improve the detection ability of small-sized and irregular-shaped defects. Specifically, convolution kernels of sizes 3×3, 5×5 and 7×7 are used to extract fine-grained, medium-grained and coarse-grained features, respectively.

[0042] Preferably, in S4, an attention mechanism is introduced into the convolutional neural network, and the weight of the feature map is dynamically adjusted by calculating the importance of each feature map to improve the detection capability of key features; specifically, the Softmax function is used to calculate the weight of each feature map, and the formula is:

[0043] ;

[0044] in, Represents the eigenvalue of the i-th feature map;

[0045] Represents the weight of the i-th feature map;

[0046] is the total number of feature maps.

[0047] Preferably, the mathematical model of CNN includes a convolutional layer, a pooling layer and a fully connected layer; the output of the convolutional layer is:

[0048] ;

[0049] Where, X is the input data;

[0050] W is the convolution kernel;

[0051] b is the bias;

[0052] * is the convolution operation;

[0053] is the activation function;

[0054] The mathematical model of LSTM includes input gate, forget gate and output gate; the calculation formula of input gate is:

[0055] ;

[0056] in, is the output of the input gate, is the hidden state at the previous moment, is the input at the current moment, is the weight matrix, is bias;

[0057] The mathematical model of SVM is:

[0058] ;

[0059] Constraints:

[0060] Among them, W is the weight vector, b is the bias, is the feature vector, For labels, is the slack variable and C is the penalty parameter.

[0061] Correspondingly, the present invention also provides a detection and positioning system for foaming defects of a sandwich foam board, comprising an infrared thermal imager and an electromagnetic induction instrument, wherein the infrared thermal imager comprises an infrared camera, a data storage module, a signal processing module, a wireless communication module and a power module, the infrared camera is used to capture the temperature distribution image of the surface of the sandwich foam board, the data storage module is used to store the collected temperature data, the signal processing module is used to perform preliminary processing on the collected temperature data, and extract key features such as temperature gradient, the wireless communication module is used to transmit the processed data to a central processing unit, and the power module is used to provide a stable working power supply for the infrared thermal imager;

[0062] The electromagnetic induction instrument includes an electromagnetic induction probe, a signal amplifier, a data acquisition module, a wireless communication module and a power module; the electromagnetic induction probe is used to detect the electromagnetic induction signal on the surface of the sandwich foam board, the signal amplifier is used to amplify the electromagnetic induction signal to improve the detection accuracy, the data acquisition module is used to collect and store the electromagnetic induction signal, the wireless communication module is used to transmit the collected data to the central processing unit, and the power module is used to provide a stable working power supply for the electromagnetic induction instrument.

[0063] Preferably, the central processing unit includes a data receiving module, a data fusion module, a model training module, a model reasoning module, a display module, a storage module, a cloud platform connection module, a human-computer interaction module, an automatic alarm module and a data encryption module; the data receiving module is used to receive data transmitted by the infrared thermal imager and the electromagnetic induction instrument, the data fusion module is used to fuse the infrared thermal imaging data and the electromagnetic induction data to improve the detection accuracy, the model training module is used to train the defect recognition and classification model using the marked defect data, the model reasoning module is used to input the preprocessed data into the trained model, and output the type and precise location of the defect, the display module is used to display the detection results, the storage module is used to store the detection results and training data, the cloud platform connection module is used to upload the detection results and training data to the cloud platform to realize remote storage and analysis of data, the human-computer interaction module is used to receive the user's operating instructions, feedback the detection results, and provide a user-friendly operation interface, the automatic alarm module is used to automatically alarm when a serious defect is detected, and remind the operator to deal with it in time, and the data encryption module is used to encrypt the detection results and training data to ensure the security and confidentiality of the data.

[0064] Preferably, a foaming defect detection and positioning system for a sandwich foam board further includes a temperature regulating device to achieve a stable detection environment temperature;

[0065] Shading equipment ensures that the light in the detection environment is moderate to avoid the influence of excessive or weak light on the imaging of the infrared thermal imager;

[0066] Vibration isolation equipment ensures that the vibration in the detection environment is moderate and avoids the influence of vibration on the electromagnetic induction signal.

[0067] Beneficial Effects

[0068] 1. By combining advanced infrared thermal imagers with electromagnetic induction technology and using deep learning models for intelligent identification and classification, it not only improves the detection accuracy, but also has good adaptability and real-time performance, providing an efficient and reliable solution for the quality control of sandwich foam panels.

[0069] 2. Through the comprehensive application of infrared thermal imagers and electromagnetic induction instruments, as well as the optimization strategy of multiple scans, not only the accuracy and comprehensiveness of foaming defect detection are improved, but also the stability and adaptability of the entire system are enhanced, and it can be effectively applied to quality inspection in large-scale production.

[0070] 3. Through the combination of initial planning, dynamic adjustment and path optimization, this method can not only effectively improve the accuracy and comprehensiveness of sandwich foam board defect detection, but also significantly improve the detection efficiency, reduce labor costs, and make the entire detection process more automated and intelligent.

[0071] 4. The hybrid model combining convolutional neural network and long short-term memory network can significantly improve the detection capability of the foaming defect detection system of sandwich foam panels. It not only improves the recognition ability of dynamic and multi-scale defects, but also improves the classification accuracy and system robustness, thus providing strong support for industrial automation quality control.

[0072] 5. The convolutional neural network with the introduction of the attention mechanism not only improves the detection capability of key defect features in the foaming defect detection of sandwich foam boards by dynamically adjusting the weight of the feature map, but also improves the accuracy, robustness and computational efficiency of the overall system. This enables this method to more accurately identify and locate foaming defects in actual industrial applications, providing more reliable quality control support.

[0073] 6. The noise pollution generated by the use of ultrasonic guided wave technology is avoided by adopting a detection and positioning system and method for foaming defects of sandwich foam panels. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0075] Figure 1 The invention is a flow chart of the steps of a method for detecting and locating foaming defects of a sandwich foam board;

[0076] Figure 2 It is a system framework for detecting and locating foam defects of sandwich foam panels. Figure 1 ;

[0077] Figure 3 It is a system framework for detecting and locating foam defects of sandwich foam panels. Figure 2 . DETAILED DESCRIPTION

[0078] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0079] See also Figure 1 The embodiment of the present invention provides a method for detecting and locating foaming defects of a sandwich foam board, comprising the following steps:

[0080] S1: Equipment preparation and environmental control; prepare infrared thermal imagers and electromagnetic induction instruments to ensure that the resolution and sensitivity of the equipment meet the detection requirements; control the detection environment to ensure that the ambient temperature is stable and reduce the impact of ambient temperature fluctuations on the detection results; clean the surface of the sandwich foam board to remove dust and oil, and ensure that there is no ferromagnetic material on the surface of the sandwich foam board to avoid interference with the electromagnetic induction signal;

[0081] S2: Data acquisition: use an infrared thermal imager to fully scan the sandwich panel and record the temperature distribution data; place an electromagnetic induction instrument on the surface of the sandwich panel, move it along the panel surface, and record the changes in the electromagnetic induction signal;

[0082] S3: Normalize the infrared thermal imaging and electromagnetic induction data to eliminate the influence of different equipment and environment, extract key features such as temperature gradient from the infrared thermal imaging data, and extract key features such as electromagnetic induction intensity change from the electromagnetic induction data;

[0083] S4: Establish a defect recognition and classification model for sandwich foam boards, and use the data of sandwich foam boards with marked defects to train the neural network model so that it can recognize and classify defects;

[0084] S5: Input the preprocessed data into the trained sandwich foam board defect recognition and classification model, and output the type and precise location of the defect.

[0085] In one possible implementation, first ensure that the equipment used - infrared thermal imager and electromagnetic induction instrument have sufficient resolution and sensitivity to meet the requirements for detecting defects in sandwich foam boards. The resolution and sensitivity of the equipment directly affect the accuracy and sensitivity of defect detection. Among them, the infrared thermal imager can provide a temperature distribution map on the surface of the board, and the electromagnetic induction instrument can detect electromagnetic changes related to defects. Through dual detection of infrared thermal imager and electromagnetic induction instrument, a more comprehensive defect type can be covered, and subtle changes in defects can be identified more accurately.

[0086] Secondly, the goal of environmental control is to ensure that the ambient temperature is stable and reduce the detection errors caused by ambient temperature fluctuations. This is especially important for infrared thermal imagers, which rely on temperature differences to identify defects. By controlling the ambient temperature, the stability and reliability of the detection results can be significantly improved. Furthermore, cleaning the surface of the sandwich foam board and removing surface dust and oil stains is not only to ensure the normal operation of the equipment, but also to avoid these substances from interfering with the electromagnetic induction signal.

[0087] Furthermore, in the data collection stage, the infrared thermal imager and electromagnetic induction instrument scan the sandwich foam board respectively to collect two types of data: temperature distribution data and electromagnetic induction data. During the infrared thermal imager scanning process, the temperature distribution map can be obtained by comprehensively scanning the surface of the sandwich board. The abnormal temperature area may correspond to foaming defects (such as uneven foaming, voids, etc.). The electromagnetic induction instrument moves along the surface of the sandwich board and records the changes in the electromagnetic signal of the board surface. Abnormal changes in the electromagnetic signal may indicate the presence of ferromagnetic substances or defects in the sandwich foam board.

[0088] After data collection, the data collected by the infrared thermal imager and the electromagnetic induction instrument need to be normalized to eliminate the impact of equipment performance differences and environmental changes on the data. The purpose of normalization is to make the data from different devices have the same metric, which is convenient for further analysis and processing. After that, key features such as temperature gradient are extracted from the infrared thermal imager data, and features such as changes in electromagnetic induction intensity are extracted from the electromagnetic induction instrument data. These key features are very important for subsequent defect identification and classification.

[0089] Furthermore, the neural network model is trained through the collected data and sample data with marked defects. The neural network can learn the laws and characteristics of foaming defects from a large amount of training data and establish a recognition and classification model for sandwich foam board defects. The core of this process is to train a model that can automatically identify and classify defects through a large amount of calibration data (including defect type and location). Through deep learning, the model can find more complex defect patterns in the data and further improve the detection accuracy.

[0090] Furthermore, by inputting the pre-processed data into the trained model, the model can output the type and exact location of the defect. Through the recognition of the neural network model, the defect area in the sandwich foam board can be accurately located and the defect type can be classified, such as whether it is a bubble, uneven foaming or other foaming defects. This process enables the inspection personnel to quickly and accurately judge and analyze the defects in the board, providing an important basis for subsequent production or maintenance.

[0091] In an embodiment of the present invention, in S2, it specifically includes the following: using an infrared thermal imager to comprehensively scan the sandwich panel and record temperature distribution data, the infrared thermal imager moves along a predetermined scanning path to ensure that the entire surface of the sandwich foam board is covered, the scanning speed is controlled at 0.1-0.5 m / s, and the stability and accuracy of the collected data are ensured. The working band of the infrared camera is 7.5-13 μm; placing an electromagnetic induction instrument on the surface of the sandwich panel, moving along the panel surface, and recording changes in the electromagnetic induction signal, the electromagnetic induction instrument moves along a predetermined scanning path to ensure that the entire surface of the sandwich foam board is covered, and the scanning speed is controlled at 0.1-0.5 m / s, and the stability and accuracy of the collected data are ensured; according to the thickness and defect type of the sandwich foam board, multiple scans are performed to improve the comprehensiveness and accuracy of the detection.

[0092] In one possible implementation, an infrared thermal imager is first used to perform a comprehensive scan of the sandwich foam board. The infrared thermal imager moves along a predetermined scanning path, and the scanning speed is controlled at 0.1-0.5 m / s. This speed setting ensures the stability of the scanning process and the accuracy of data acquisition. The infrared thermal imager mainly captures foaming defects by detecting the temperature distribution on the surface of the board. Abnormal changes in temperature may reflect uneven foaming, voids or other defects. For example, uneven foaming on the surface of a foam board may cause the temperature to rise or fall in a local area. These abnormal temperature differences provide key information for subsequent defect identification. The working band of the infrared thermal imager is 7.5-13μm, which is particularly suitable for detecting temperature changes in the medium and low temperature range, and can effectively identify subtle changes in surface temperature.

[0093] Furthermore, an electromagnetic induction instrument is used to scan the surface of the sandwich panel and record the changes in the electromagnetic induction signal. The scanning process of the electromagnetic induction instrument is also carried out along a predetermined path, and the scanning speed is controlled between 0.1-0.5 m / s to ensure the stability and accuracy of the collected signal. The electromagnetic induction instrument can detect electromagnetic changes caused by defects in the sandwich foam board. For example, due to uneven foaming or the presence of foreign matter such as metal materials, the electromagnetic signal will deviate. The combination of electromagnetic induction data and temperature distribution data can provide more dimensional information about defects and improve the accuracy and comprehensiveness of detection.

[0094] Furthermore, multiple scans are performed according to the thickness and defect type of the sandwich foam board to improve the comprehensiveness and accuracy of the detection. Due to the complex structure of the sandwich foam board, a single scan may not cover all possible defect areas or defects of sufficient depth. Therefore, multiple scans, especially for areas with thicker boards or less obvious defects, can capture potential defects more comprehensively. By scanning at different angles or depths, the coverage of the detection and the richness of the detection data are increased, thereby ensuring high accuracy of the detection.

[0095] In the embodiment of the present invention, S2 also includes scanning path planning, and plans the scanning paths of the infrared thermal imager and the electromagnetic induction instrument according to the size and shape of the sandwich foam board to ensure that the entire surface of the sandwich foam board is covered, which specifically includes the following steps: initial planning: generating an initial scanning path according to the size and shape of the sandwich foam board; dynamic adjustment: during the scanning process, dynamically adjusting the scanning path according to the real-time feedback of the infrared thermal imaging data and the electromagnetic induction data to ensure that all key areas are covered; path optimization: using the ant colony algorithm to optimize the scanning path and reduce the scanning time. The mathematical model of the ant colony algorithm is: ;in, is the incremental pheromone on the path (i, j); Q is a constant, is the total path length of ant k.

[0096] In one possible implementation, initial planning is the first step in the design of the entire scanning path, which mainly generates a scanning path covering the entire surface based on the size and shape of the sandwich foam board. In this step, considering the specific size of the sandwich foam board (such as length, width, thickness, etc.) and the possible geometric features on the surface (such as rounded corners, unevenness), a scanning route will be preset through an algorithm to ensure that the scanning equipment can cover every area of ​​the board to prevent omissions. At the same time, the path planning needs to take into account the performance characteristics of the scanning equipment, such as the sensing range of infrared thermal imagers and electromagnetic induction devices, so as to reasonably arrange the scanning path.

[0097] Furthermore, based on the initially planned scanning path, the real-time data feedback from the infrared thermal imager and electromagnetic induction instrument will affect the adjustment of the scanning path. When the system finds abnormalities in certain areas during the scanning process (such as large temperature deviations or significant changes in electromagnetic signals), the detection algorithm will automatically identify these areas and instruct the scanning equipment to adjust the path to ensure that these potential defective areas are fully covered and further inspected. The introduction of dynamic adjustment makes the scanning process more intelligent, and can flexibly change the original route according to real-time data to avoid missing any possible defective areas, further improving the comprehensiveness and accuracy of defect detection.

[0098] Furthermore, the scanning path is further optimized by using the ant colony algorithm. The ant colony algorithm is an optimization algorithm that simulates the foraging behavior of ants in nature and is suitable for solving problems such as path planning. In this process, the ant algorithm gradually finds a more efficient scanning path by simulating the actions of multiple "ants" on the scanning path, thereby reducing unnecessary repeated scanning and improving work efficiency. Through iterative optimization, the ant algorithm can find the shortest or optimal path among multiple path options, thereby reducing scanning time and improving detection efficiency. Through this optimization, the scanning path is streamlined and can quickly and effectively cover the entire plate while ensuring sufficient detection of potential defect areas.

[0099] In the embodiment of the present invention, in S3, a smoothing filter is used to filter the collected temperature data and electromagnetic induction signal to reduce noise interference; wherein key features such as temperature gradient, temperature difference and temperature fluctuation are extracted from the infrared thermal imaging data, and the temperature gradient calculation formula is: ; Where T is the temperature, x and y are the spatial coordinates; and Respectively represent the partial derivatives of temperature T in the x and y directions; extract key features such as electromagnetic induction intensity change and electromagnetic field distribution from the electromagnetic induction data; the calculation formula for electromagnetic induction intensity change is:

[0100] ; Where B(t) is the electromagnetic induction intensity at the current time point, is the electromagnetic induction intensity at the previous time point, For the time interval.

[0101] In a possible implementation, the temperature gradient reflects the change in the temperature around a certain location and is an important indicator for detecting foaming defects. During the inspection of sandwich foam boards, the temperature data captured by the infrared thermal imager needs to be used to calculate the temperature gradient to reveal the change in the surface temperature of the board. The calculated temperature gradient can reflect the law of temperature change on the surface of the foam board, thereby revealing potential defect areas in the foaming process.

[0102] Furthermore, the electromagnetic induction data can reflect the changes in the internal material structure of the foam board. By measuring the changes in the intensity of the electromagnetic field, it can be further determined whether the sandwich foam board has uneven foaming or defects. The calculation formula for the change in electromagnetic induction intensity is: ; Where B(t) is the electromagnetic induction intensity at the current time point, is the electromagnetic induction intensity at the previous time point, is the time interval. The electromagnetic induction intensity changes It reflects the changes in the electromagnetic response of the foam board at different positions, thereby revealing areas with defects or structural changes. Similarly, the electromagnetic induction signal will also be interfered by factors such as environmental noise and equipment instability, so it is necessary to process the electromagnetic induction signal through a smoothing filter. The smoothing filter can eliminate these high-frequency noises, making the measurement results of the electromagnetic induction intensity changes more stable and accurate, ensuring that potential defects in the foam board can be effectively identified.

[0103] In an embodiment of the present invention, in S4, a convolutional neural network (CNN) is used to extract high-level features from preprocessed data; through convolutional layers, pooling layers and fully connected layers, feature extraction and classification of data are performed, and a long short-term memory network (LSTM) is used to capture the features of time series data to enhance the ability to identify defect types and locations; through a gating mechanism, the gradient vanishing and gradient exploding problems in long sequence data are solved, and a support vector machine (SVM) is used to classify the extracted features to improve classification accuracy, and the optimal classification hyperplane is found by maximizing the classification interval; wherein, the convolutional neural network and the long short-term memory network are combined to form a multi-layered classification model; and the convolutional neural network and the long short-term memory network are combined to form a multi-layered classification model. The convolutional neural network is combined with the long short-term memory network to form a hybrid model. The convolutional neural network is used to extract spatial features, and then the extracted features are input into the long short-term memory network to capture time series features and improve the detection ability of dynamically changing defects. The output of the convolutional neural network is used as the input of the long short-term memory network, and feature fusion is performed through a gating mechanism. A multi-scale convolutional layer is introduced into the convolutional neural network, and multi-scale features are extracted through convolution kernels of different sizes to improve the detection ability of small-sized and irregular-shaped defects. Specifically, convolution kernels of sizes 3×3, 5×5 and 7×7 are used to extract fine-grained, medium-grained and coarse-grained features, respectively.

[0104] In a possible implementation, in step S4, a convolutional neural network (CNN) and a long short-term memory network (LSTM) are combined to form a hybrid model to extract and classify high-level features of the preprocessed data, thereby improving the accuracy and robustness of foaming defect detection of sandwich foam panels.

[0105] Specifically, in the hybrid model, the role of the convolutional neural network is to extract spatial features from the input data (such as temperature images or electromagnetic induction images), and the convolution layer of the CNN is responsible for automatically learning local features (such as the edges and textures of defects) from the data. Through the combination of multiple layers of convolution, pooling, and fully connected layers, CNN gradually abstracts more advanced features, and finally uses the extracted features as the input of LSTM. The long short-term memory network is used to model the time series of spatial features extracted by the convolutional neural network. Since the temperature changes or electromagnetic induction signal changes in defect detection are usually dynamic, LSTM can learn and remember the features of long-term dependencies in the time series, helping the detector to accurately identify the evolution of defects in the dynamic process. The time series features captured by LSTM help to improve the sensitivity and accuracy of defect changes. When the two are combined, the convolutional neural network is responsible for capturing spatial structural features, while LSTM is responsible for capturing patterns in the time series, ensuring that defects with large changes at different time points can also be effectively identified.

[0106] Furthermore, in order to enhance the detection capability of small-sized and irregular-shaped defects, a multi-scale convolution layer is introduced into the convolutional neural network. By using convolution kernels of sizes 3×3, 5×5, and 7×7, CNN is able to extract features from multiple scales, thus better coping with defects of different sizes and shapes.

[0107] Among them, fine-grained features (3×3 convolution kernel): are suitable for capturing small-sized, local defect features, such as tiny cracks or small uneven foaming areas.

[0108] Medium-sized features (5×5 convolution kernel): suitable for extracting medium-sized defect features, such as uneven foaming or local depressions in a large area.

[0109] Coarse-grained features (7×7 convolution kernel): can capture larger-scale defects, such as obvious cracks or large-scale blistering defects.

[0110] This multi-scale convolution kernel design enables the convolutional neural network to not only identify small defects, but also detect a wider range of structural problems, thereby improving the comprehensiveness and robustness of defect detection.

[0111] Furthermore, after feature extraction, the long short-term memory network helps capture the changing features in the time series, and the final classification process is completed by the support vector machine (SVM). SVM finds the optimal classification hyperplane by maximizing the classification interval, so that defects of different categories can be accurately distinguished. In this way, combined with the spatial and temporal features extracted by CNN and LSTM, SVM can better classify defects in foam boards and ensure high-precision detection and positioning.

[0112] In the embodiment of the present invention, in S4, an attention mechanism is introduced into the convolutional neural network, and the weight of the feature map is dynamically adjusted by calculating the importance of each feature map to improve the detection capability of key features. Specifically, the Softmax function is used to calculate the weight of each feature map, and the formula is:

[0113] ;in, Represents the eigenvalue of the i-th feature map; Represents the weight of the i-th feature map; is the total number of feature maps.

[0114] In one possible implementation, an attention mechanism is introduced into a convolutional neural network (CNN) to further optimize the weighting process of feature maps and improve the accuracy of detection and positioning. Specifically, the core idea of ​​the attention mechanism is to dynamically adjust the weight of each feature map in the network by calculating the importance of each feature map. Its role is similar to that of humans focusing on important parts and ignoring irrelevant or unimportant parts during perception. In a convolutional neural network, feature maps reflect feature information in different regions, while some regions may contain defect information and other regions may be background or irrelevant regions. By introducing the attention mechanism, the network can automatically evaluate the relative importance of each feature map, thereby selectively enhancing the response to key features. At the same time, through the Softmax function, the weights of all feature maps will be normalized into a probability distribution, so that the weight of each feature map can be adjusted according to its importance in feature extraction. Based on these weights, the network can automatically focus on those feature maps containing key defect information and suppress unimportant feature maps.

[0115] The introduction of the attention mechanism enables the network to automatically evaluate the importance of each feature map, thereby effectively improving the ability to detect key features of foaming defects. For complex foaming defect images, some areas may be critical to the judgment of defects, while other areas may interfere with the detection results. The attention mechanism helps the network focus on important areas and avoid information overload.

[0116] Furthermore, after the introduction of the attention mechanism, CNN can more accurately extract key features containing defect information. This mechanism, combined with the convolutional layer and pooling layer of CNN, enhances the network's ability to identify foaming defect features. After being weighted by the attention mechanism, the feature map of the convolutional layer can highlight important features related to defects, suppress the influence of background noise or irrelevant areas, and avoid interference from redundant information.

[0117] For example, when detecting foam defects in sandwich foam panels, problems such as uneven foaming or local temperature fluctuations may occur, and the characteristics of these defects may appear in different forms in different image regions. Through the attention mechanism, the network can focus on the feature maps of these key areas and dynamically adjust the weight of each feature map according to its importance, ensuring that important information is amplified and unimportant features are weakened.

[0118] In the embodiment of the present invention, the mathematical model of CNN includes a convolutional layer, a pooling layer and a fully connected layer; the output of the convolutional layer is: ; Where X is the input data; W is the convolution kernel; b is the bias; * is the convolution operation; is the activation function; the mathematical model of LSTM includes input gate, forget gate and output gate; the calculation formula of input gate is:

[0119] ;in, is the output of the input gate, is the hidden state at the previous moment, is the input at the current moment, is the weight matrix, is the bias; the mathematical model of SVM is: ; Constraints: ; Where W is the weight vector, b is the bias, is the feature vector, For labels, is the slack variable and C is the penalty parameter.

[0120] In one possible implementation, CNN first extracts feature maps with spatial information from the input foam board image. These feature maps contain key information in the image, such as the shape, color, and texture of the foam. The CNN model is usually composed of convolutional layers, pooling layers, and fully connected layers, among which the convolutional layer is its core component and is responsible for extracting features from the input data. Specifically, the output calculation formula of the convolutional layer is: ; Where X is the input data; W is the convolution kernel; b is the bias; * is the convolution operation; is the activation function; these feature maps contain the spatial features in the input data. Through a series of convolution operations, CNN can extract relevant information about blister defects from the original image. The pooling layer is used to reduce dimensionality, reduce computational complexity, and enhance feature stability.

[0121] Furthermore, in some cases, foaming defects may show time-varying characteristics, and the LSTM network is used to perform time-series modeling on the feature maps extracted by CNN to capture the changing trends of defects at different time points. For example, if an abnormality occurs during a certain period of time in the foaming process, LSTM can help capture this information. LSTM is a common neural network for processing time series data, and its core is the input gate, forget gate, and output gate. These gates control the flow of information between time steps, allowing LSTM to effectively capture long-term dependencies. In foaming defect detection, LSTM can help process and analyze defect information with time series characteristics, such as changes in images during the detection process or the development trend of defects. By calculating the output of the input gate, forget gate, and output gate at each moment, LSTM can effectively model the sequence data, thereby accurately predicting and locating the time changes of foaming defects.

[0122] Furthermore, through the features extracted and modeled by CNN and LSTM, SVM makes the final classification judgment to determine whether the foam board has defects. If there are defects, SVM can further locate and calibrate the location and type of the defective area.

[0123] SVM is a powerful classification algorithm used to segment data into different categories. In foaming defect detection, SVM is used to classify the extracted features and determine whether there are defects. The mathematical model of SVM is: ; Constraints: ; Where W is the weight vector, b is the bias, is the feature vector, For labels, is the slack variable, and C is the penalty parameter. SVM solves an optimization problem to find an optimal hyperplane to distinguish different categories of data (defects and non-defects).

[0124] See also Figure 2Based on the same inventive concept, an embodiment of the present invention provides a detection and positioning system for foaming defects of a sandwich foam board, including an infrared thermal imager and an electromagnetic induction instrument, wherein the infrared thermal imager includes an infrared camera, a data storage module, a signal processing module, a wireless communication module and a power supply module, the infrared camera is used to capture the temperature distribution image of the surface of the sandwich foam board, the data storage module is used to store the collected temperature data, the signal processing module is used to perform preliminary processing on the collected temperature data, and extract key features such as temperature gradient, the wireless communication module is used to transmit the processed data to a central processing unit, and the power supply module is used to provide a stable working power supply for the infrared thermal imager; the electromagnetic induction instrument includes an electromagnetic induction probe, a signal amplifier, a data acquisition module, a wireless communication module and a power supply module; the electromagnetic induction probe is used to detect the electromagnetic induction signal on the surface of the sandwich foam board, the signal amplifier is used to amplify the electromagnetic induction signal and improve the detection accuracy, the data acquisition module is used to collect and store the electromagnetic induction signal, the wireless communication module is used to transmit the collected data to the central processing unit, and the power supply module is used to provide a stable working power supply for the electromagnetic induction instrument.

[0125] In a possible embodiment, in the infrared thermal imager, the infrared camera is used to capture the temperature distribution image of the surface of the sandwich foam board. The infrared camera can monitor the heat changes on the surface of the foam board in real time, and the temperature distribution image provides intuitive information of potential defects. Usually, foaming defects may cause local heat changes. Therefore, the area of ​​uneven heat can be found through the infrared image, so as to locate possible defects. The data storage module is used to store the collected temperature data to provide data support for subsequent analysis and processing. These data can be stored at different time points, and their change trends can be analyzed by subsequent processing modules. The signal processing module is used to perform preliminary processing on the collected temperature data and extract key features such as temperature gradient and thermal anomaly. By analyzing the gradient change of temperature, the system can determine whether there is a thermal anomaly area, which is very critical for the preliminary positioning of foaming defects. The wireless communication module is used to wirelessly transmit the processed data to the central processor for further analysis and processing. Through the wireless communication module, the system can transmit data in real time, which is convenient for the rapid response of the central processor. The power supply module is used to provide stable power supply support for the entire infrared thermal imager to ensure the normal operation of the equipment.

[0126] Furthermore, in the electromagnetic induction instrument, the electromagnetic induction probe is used to detect the electromagnetic induction signal on the surface of the sandwich foam board. By detecting the electromagnetic properties of the surface material of the foam board through the electromagnetic induction probe, the system can obtain other information about foaming defects, such as material unevenness, pore changes, etc. The signal amplifier is used to amplify the electromagnetic induction signal to improve the detection accuracy. Since the electromagnetic signal may be very weak, the role of the signal amplifier is to enhance the strength of the signal to ensure that potential defect signals can be accurately captured. The data acquisition module is responsible for collecting and storing electromagnetic induction signals. The electromagnetic induction signal provides different information about the internal state of the material, and these signals can be effectively collected through the data acquisition module. The wireless communication module is used to wirelessly transmit the collected data to the central processor to ensure the real-time and response speed of the system. Through the wireless communication module, the system can upload the collected data of the electromagnetic induction instrument in a timely manner, which is convenient for the central processor to perform unified processing. The power module provides stable power support for the electromagnetic induction instrument to ensure that the equipment continues to work reliably in different detection scenarios.

[0127] The infrared thermal imager and electromagnetic induction instrument are connected to the central processor through the wireless communication module to form an efficient and coordinated detection system. The infrared thermal imager focuses on thermal anomaly detection and locates the defects of the foam board through image and temperature gradient analysis. The electromagnetic induction instrument assists in analyzing the nature and distribution of defects by detecting the electromagnetic signals of the material. The collaborative work of these two modules can not only improve the comprehensiveness of detection, but also improve the accuracy of positioning.

[0128] During operation, the infrared thermal imager first provides a temperature distribution image, and the central processor performs a preliminary analysis based on characteristics such as temperature gradient to determine whether there is a defect. If a suspicious area is found, the electromagnetic induction instrument will further provide electromagnetic signal data to verify the location and nature of the defect. Finally, the central processor comprehensively analyzes and determines the exact location and type of the defect based on the two different types of data.

[0129] In an embodiment of the present invention, the central processing unit includes a data receiving module, a data fusion module, a model training module, a model reasoning module, a display module, a storage module, a cloud platform connection module, a human-computer interaction module, an automatic alarm module and a data encryption module; the data receiving module is used to receive data transmitted by an infrared thermal imager and an electromagnetic induction instrument, the data fusion module is used to fuse infrared thermal imaging data and electromagnetic induction data to improve detection accuracy, the model training module is used to train defect recognition and classification models using labeled defect data, the model reasoning module is used to input preprocessed data into a trained model, and output the type and precise location of the defect, the display module is used to display the detection results, the storage module is used to store the detection results and training data, the cloud platform connection module is used to upload the detection results and training data to the cloud platform to realize remote storage and analysis of data, the human-computer interaction module is used to receive user operation instructions, feedback detection results, and provide a user-friendly operation interface, the automatic alarm module is used to automatically alarm when serious defects are detected, and remind operators to deal with them in time, and the data encryption module is used to encrypt the detection results and training data to ensure the security and confidentiality of the data.

[0130] In a possible implementation, the central processing unit undertakes the core data processing and decision-making tasks. Its main components include: data receiving module, data fusion module, model training module, model reasoning module, display module, storage module, cloud platform connection module, human-computer interaction module, automatic alarm module and data encryption module. These modules work together efficiently to ensure the real-time, accuracy and security of the system.

[0131] These modules ensure the smooth operation of the entire system through efficient interconnection and collaborative work. Specifically, the data receiving module first receives data from the infrared thermal imager and the electromagnetic induction instrument, and passes it to the data fusion module. After fusion, the data is sent to the model reasoning module, and the model identifies and locates the defect type according to the trained parameters. The final result is presented to the user by the display module. The data and results are saved by the storage module at the same time, and uploaded to the cloud platform through the cloud platform connection module for remote storage and analysis.

[0132] The human-computer interaction module plays a bridging role in this process. Users interact with the system through this module, and operation instructions and feedback results are transmitted in real time in the system. If the system detects a serious defect, the automatic alarm module will trigger an alarm in time to ensure that the operator takes prompt action. In addition, all data will be encrypted through the data encryption module to ensure the security of the test results and training data.

[0133] See also Figure 3In an embodiment of the present invention, a detection and positioning system for foaming defects of a sandwich foam board also includes a temperature regulating device to achieve stable temperature of the detection environment; a shading device to ensure that the light in the detection environment is moderate to avoid the influence of excessive or weak light on the imaging of the infrared thermal imager; and a vibration isolation device to ensure that the vibration of the detection environment is moderate to avoid the influence of vibration on the electromagnetic induction signal.

[0134] In a possible implementation, the temperature regulating device is used to ensure the temperature stability of the detection environment. This is because temperature changes will directly affect the imaging accuracy of the infrared thermal imager and the data collection of the electromagnetic induction instrument. Temperature instability may cause deviations in the infrared image and affect the location and classification of defects. The temperature regulating device maintains a constant temperature in the detection area by heating or cooling to ensure the accuracy of the detection data.

[0135] Furthermore, the shading device is used to adjust the light in the detection environment to ensure that the light in the detection environment is moderate. Infrared thermal imagers are very sensitive to ambient light. Strong light may cause overexposure of infrared imaging images, while weak light may result in poor image quality and affect detection accuracy. Therefore, the shading device can effectively control the light source, avoid interference from external light sources, and provide stable infrared imaging conditions.

[0136] Furthermore, vibration isolation equipment is used to reduce or eliminate vibration interference in the detection environment, especially the impact on electromagnetic induction instruments. Electromagnetic induction instruments are very sensitive to vibrations in the environment, which can cause signal noise and affect the accurate detection of internal defects in materials. Vibration isolation equipment effectively isolates external mechanical vibrations through special shock-absorbing devices (such as shock-absorbing platforms, vibration-absorbing pads, etc.), ensuring that electromagnetic induction instruments can operate stably.

[0137] These auxiliary devices form a close synergy with the core detection module of the system. The temperature control device ensures a suitable ambient temperature, thereby stabilizing the working state of the infrared thermal imager and the electromagnetic induction instrument; the shading device ensures the quality of infrared imaging by adjusting the light and prevents strong or weak light from interfering with the detection; the vibration isolation device reduces the impact of environmental vibration on the electromagnetic induction instrument and ensures the accuracy of the signal.

[0138] The optional implementation modes of the embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above implementation modes. Within the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical scheme of the embodiments of the present invention, and these simple modifications all belong to the protection scope of the embodiments of the present invention.

[0139] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe various possible combinations.

[0140] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a single-chip microcomputer, a chip or a processor to execute all or part of the steps of the method described in each embodiment of the present application. In addition, various different implementation methods of the embodiments of the present invention can also be combined arbitrarily, as long as they do not violate the ideas of the embodiments of the present invention, they should also be regarded as the contents disclosed in the embodiments of the present invention.

Claims

1. A method for detecting and locating foaming defects of a sandwich foam board, characterized in that: The steps include: S1: Equipment preparation and environmental control; Prepare infrared thermal imagers and electromagnetic induction instruments to ensure that the resolution and sensitivity of the equipment meet the detection requirements; Control the testing environment to ensure stable ambient temperature and reduce the impact of ambient temperature fluctuations on test results; Clean the surface of the sandwich foam board to remove dust and oil, and ensure that there is no ferromagnetic material on the surface of the sandwich foam board to avoid interference with electromagnetic induction signals; S2: Data acquisition, use an infrared thermal imager to fully scan the sandwich panel and record the temperature distribution data; Place the electromagnetic induction instrument on the surface of the sandwich panel, move it along the panel surface, and record the changes of the electromagnetic induction signal; S3: Normalize the infrared thermal imaging and electromagnetic induction data to eliminate the influence of different equipment and environment, extract key features related to temperature gradient from the infrared thermal imaging data, and extract key features related to electromagnetic induction intensity change from the electromagnetic induction data; S4: Establish a defect recognition and classification model for sandwich foam boards, and use the data of sandwich foam boards with marked defects to train the neural network model so that it can recognize and classify defects; S5: input the preprocessed data into the trained sandwich foam board defect recognition and classification model, and output the type and precise location of the defect; In S4, a convolutional neural network (CNN) is used to extract high-level features from the preprocessed data; Through the convolution layer, pooling layer and fully connected layer, the data is feature extracted and classified. The long short-term memory network LSTM is used to capture the characteristics of time series data and enhance the ability to identify defect types and locations. The gating mechanism is used to solve the gradient vanishing and gradient exploding problems in long sequence data. The support vector machine SVM is used to classify the extracted features to improve the classification accuracy. The optimal classification hyperplane is found by maximizing the classification interval. Among them, the convolutional neural network and the long short-term memory network are combined to form a hybrid model. The convolutional neural network is used to extract spatial features, and then the extracted features are input into the long short-term memory network to capture time series features and improve the detection ability of dynamically changing defects. The output of the convolutional neural network is used as the input of the long short-term memory network, and feature fusion is performed through a gating mechanism; A multi-scale convolution layer is introduced into the convolutional neural network to extract multi-scale features through convolution kernels of different sizes, thereby improving the detection capability of small-sized and irregular-shaped defects. Specifically, convolution kernels of sizes 3×3, 5×5, and 7×7 are used to extract fine-grained, medium-grained, and coarse-grained features, respectively. In S4, the attention mechanism is introduced into the convolutional neural network. By calculating the importance of each feature map, the weight of the feature map is dynamically adjusted to improve the detection ability of key features. Specifically, the Softmax function is used to calculate the weight of each feature map, and the formula is: ； in, Represents the eigenvalue of the i-th feature map; Represents the weight of the i-th feature map; is the total number of feature maps; The mathematical model of CNN includes convolutional layer, pooling layer and fully connected layer; the output of the convolutional layer is: ； Where, X is the input data; W is the convolution kernel; b is the bias; * is the convolution operation; is the activation function; The mathematical model of LSTM includes input gate, forget gate and output gate; the calculation formula of input gate is: ； in, is the output of the input gate, is the hidden state at the previous moment, is the input at the current moment, is the weight matrix, is bias; The mathematical model of SVM is: ； Constraints: ; Among them, w is the weight vector, b is the bias, is the feature vector, For labels, is the slack variable and C is the penalty parameter.

2. A method for detecting and locating foaming defects of a sandwich foam board according to claim 1, characterized in that: In S2, the specifics include: Use an infrared thermal imager to fully scan the sandwich panel and record the temperature distribution data. The infrared thermal imager moves along the predetermined scanning path to ensure that the entire surface of the sandwich foam board is covered. The scanning speed is controlled at 0.1-0.5 m / s to ensure the stability and accuracy of the collected data. The working band of the infrared camera is 7.5-13μm; Place the electromagnetic induction instrument on the surface of the sandwich panel, move along the panel surface, and record the changes of the electromagnetic induction signal. The electromagnetic induction instrument moves along the predetermined scanning path to ensure that the entire surface of the sandwich foam board is covered. The scanning speed is controlled at 0.1-0.5 m / s to ensure the stability and accuracy of the collected data. According to the thickness and defect type of the sandwich foam board, multiple scans are performed to improve the comprehensiveness and accuracy of the detection.

3. The method for detecting and locating foaming defects of a sandwich foam board according to claim 2, characterized in that: S2 also includes scanning path planning. According to the size and shape of the sandwich foam board, the scanning paths of the infrared thermal imager and the electromagnetic induction instrument are planned to ensure that the entire surface of the sandwich foam board is covered. Specifically, the following steps are included: Initial planning: Generate the initial scanning path according to the size and shape of the sandwich foam board; Dynamic adjustment: During the scanning process, the scanning path is dynamically adjusted based on the real-time feedback of infrared thermal imaging data and electromagnetic induction data to ensure that all key areas are covered; Path optimization: Use ant colony algorithm to optimize the scanning path and reduce scanning time. The mathematical model of ant colony algorithm is: ； in, is the incremental pheromone on path (i, j); Q is a constant, is the total path length of ant k.

4. The method for detecting and locating foaming defects of a sandwich foam board according to claim 1, characterized in that: In S3, a smoothing filter is used to filter the collected temperature data and electromagnetic induction signals to reduce noise interference; Among them, the key features related to temperature gradient, temperature difference and temperature fluctuation are extracted from infrared thermal imaging data. The temperature gradient calculation formula is: ； Where T is the temperature, x and y are the spatial coordinates; and Respectively represent the partial derivatives of temperature T in the x and y directions; Extract the key features related to the change of electromagnetic induction intensity and electromagnetic field distribution from the electromagnetic induction data; the calculation formula for the change of electromagnetic induction intensity is: ； Among them, B(t) is the electromagnetic induction intensity at the current time point, is the electromagnetic induction intensity at the previous time point, For the time interval.

5. A foaming defect detection and positioning system for a sandwich foam board, used for running the foaming defect detection and positioning method for a sandwich foam board according to any one of claims 1 to 4, characterized in that: It includes an infrared thermal imager and an electromagnetic induction instrument, wherein the infrared thermal imager includes an infrared camera, a data storage module, a signal processing module, a wireless communication module and a power supply module, the infrared camera is used to capture the temperature distribution image of the surface of the sandwich foam board, the data storage module is used to store the collected temperature data, the signal processing module is used to perform preliminary processing on the collected temperature data and extract key features related to the temperature gradient, the wireless communication module is used to transmit the processed data to the central processing unit, and the power supply module is used to provide a stable working power supply for the infrared thermal imager; The electromagnetic induction instrument includes an electromagnetic induction probe, a signal amplifier, a data acquisition module, a wireless communication module and a power module; the electromagnetic induction probe is used to detect the electromagnetic induction signal on the surface of the sandwich foam board, the signal amplifier is used to amplify the electromagnetic induction signal to improve the detection accuracy, the data acquisition module is used to collect and store the electromagnetic induction signal, the wireless communication module is used to transmit the collected data to the central processing unit, and the power module is used to provide a stable working power supply for the electromagnetic induction instrument; The central processing unit includes a data receiving module, a data fusion module, a model training module, a model reasoning module, a display module, a storage module, a cloud platform connection module, a human-computer interaction module, an automatic alarm module and a data encryption module; the data receiving module is used to receive data transmitted by the infrared thermal imager and the electromagnetic induction instrument, the data fusion module is used to fuse the infrared thermal imaging data and the electromagnetic induction data to improve the detection accuracy, the model training module is used to train the defect recognition and classification model using the marked defect data, the model reasoning module is used to input the preprocessed data into the trained model, and output the type and precise location of the defect, the display module is used to display the detection results, the storage module is used to store the detection results and training data, the cloud platform connection module is used to upload the detection results and training data to the cloud platform to realize remote storage and analysis of data, the human-computer interaction module is used to receive the user's operation instructions, feedback the detection results, and provide a user-friendly operation interface, the automatic alarm module is used to automatically alarm when a serious defect is detected, and remind the operator to deal with it in time, and the data encryption module is used to encrypt the detection results and training data to ensure the security and confidentiality of the data.

6. A foaming defect detection and positioning system for a sandwich foam board according to claim 5, characterized in that: A detection and positioning system for foaming defects of a sandwich foam board, further comprising a temperature regulating device to achieve a stable detection environment temperature; Shading equipment ensures that the light in the detection environment is moderate to avoid the influence of excessive or weak light on the imaging of the infrared thermal imager; Vibration isolation equipment ensures that the vibration in the detection environment is moderate and avoids the influence of vibration on the electromagnetic induction signal.

Images