Furniture plate structure performance detection system and detection method

Abstract

The invention discloses a furniture plate structure performance detection system and detection method, and the system comprises a multi-source collection module which collects plate structure data and environment related data, extracts a first related parameter from the collected plate structure parameters, and extracts a second related parameter from the related environment parameters; the association acquisition module is used for performing cross-modal fusion analysis on the first association parameter and the second association parameter, acquiring an optimized association index, and acquiring trend data from the optimized association index; the data correction module is used for collecting plate characteristic data, optimizing a cross correlation coefficient between the plate characteristic data and the trend data, carrying out correlation analysis based on the optimized cross correlation coefficient to determine an attenuation mode, carrying out attenuation compensation on the collected real-time plate characteristic data based on the attenuation mode, obtaining corrected plate characteristic data, and outputting the corrected plate characteristic data. And it is ensured that the corrected strength data are closer to real performance in an actual environment.

Description

Technical Field

[0001] The present invention relates to the technical field of data detection, and particularly relates to a furniture board structure performance detection system and a detection method. Background Art

[0002] The development of furniture board structure performance detection systems and detection methods is the result of the combined action of growing market demand and technological progress. With the improvement of people's living standards, the requirements for furniture quality have become increasingly strict. As a key component of furniture, the structure performance of furniture boards directly determines the quality, safety, and service life of furniture, prompting the continuous development of related detection technologies. At the same time, innovations in materials science have provided new detection objects and challenges for the development of detection technologies. The prominent impact of environmental factors on board performance has also exposed the limitations of traditional detection methods. The rapid development of technologies such as artificial intelligence, sensors, and data processing has provided strong support for the innovative breakthrough of detection technologies.

[0003] Currently, existing detection systems usually use fixed algorithm models to process detection data. However, the performance of furniture boards changes with the environment, and fixed algorithms lack consideration of environmental factors. In an environment with high humidity, solid wood boards are prone to moisture absorption, resulting in an increase in moisture content, which causes dimensional expansion and a decrease in strength. For example, during the plum rain season in the south, the moisture content of red oak boards may rise from the normal 12% to 20%. When using a fixed algorithm for detection, without adjusting the evaluation criteria for board strength and dimensional stability according to the change in moisture content and still detecting according to the parameters in a dry environment, it will overestimate the actual performance of the board, causing problems such as deformation and cracking in the furniture made in subsequent use. Therefore, a furniture board structure performance detection system and a detection method are proposed here to solve the above problems. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above object, the present invention proposes the following technical solutions:

[0005] A furniture board structure performance detection system, comprising:

[0006] A multi-source acquisition module: acquiring board structure data and environment-related data, extracting a first correlation parameter from the acquired board structure parameters, and extracting a second correlation parameter from the related environment parameters;

[0007] A correlation acquisition module: performing cross-modal fusion analysis on the first correlation parameter and the second correlation parameter to obtain an optimized correlation index, and obtaining trend data from the optimized correlation index;

[0008] Data correction module: Collect the data of the board characteristics, optimize the cross-correlation coefficient between the data of the board characteristics and the trend data, determine the attenuation mode based on the optimized cross-correlation coefficient through correlation analysis, and perform attenuation compensation on the collected real-time board characteristic data based on the attenuation mode to obtain the corrected board characteristic data;

[0009] Detection and output module: Obtain the detection result of the furniture board based on the corrected board characteristic data.

[0010] The process of obtaining the board structure data is as follows:

[0011] Use an industrial CT scanner to perform a full-range scan on the board to obtain the three-dimensional structure image data of the board , and at the same time use a density measurement device to obtain the density values at different positions of the board , and use a three-dimensional laser scanner to obtain the geometric shape data of the board surface ;

[0012] The method for collecting the environment-related data is as follows:

[0013] Arrange a distributed environmental sensor network to collect the environmental temperature data and the relative humidity data .

[0014] The process of obtaining the first correlation parameter is as follows:

[0015] Through an image segmentation algorithm, separate the defective areas from the three-dimensional structure image data to obtain the defect parameters ;

[0016] Extract the flatness of the board surface from the geometric shape data , the contour information data of the edge , and obtain the optimized form index parameters through the mapping relationship between the flatness of the board surface and the contour information data of the edge ;

[0017] Obtain the average density from the density values at different positions ;

[0018] The first correlation parameter is expressed as ;

[0019] The process of obtaining the second correlation parameter is as follows:

[0020] Extract the temperature change range within a specific time period , and the humidity change rate

[0021] Record a period of time The maximum value of the internal environment temperature and the minimum value The difference between the two is the temperature change range ;

[0022] By continuously monitoring the relative humidity for a period of time calculate the humidity change rate ;

[0023] The second set of correlation parameters is expressed as .

[0024] The process of obtaining the optimized morphological index parameters is as follows:

[0025] Convert the flatness of the plate surface and the contour information data of the edge into point cloud data ;

[0026] Through a model, incorporate an attention mechanism module into the middle layer of the model to obtain a fused network model;

[0027] Input the point cloud data into the PointNet++ fused network model to obtain local feature vectors ;

[0028] The network model fuses the processed local features to obtain global morphological features ;

[0029] Then, obtain topological features through topological data analysis of the extracted global feature vectors ;

[0030] According to the extracted global morphological features and topological features , establish a mapping relationship between the global morphological features and topological features;

[0031] Map the eigenvalue representing the edge contour continuity in the topological features to the smoothness index of the edge;

[0032] Map the components related to surface undulation in the global feature vector to the flatness index of the surface;

[0033] The smoothness index of the edge and the flatness index of the surface together form sub-indices ;

[0034] According to each sub-index obtained by mapping , determine the weights of different sub - indicators according to the analytic hierarchy process, and then obtain the final optimized form index parameters through weighted summation. .

[0035] The process of obtaining the optimized correlation index is as follows:

[0036] Perform feature - encoding representation on the first correlation parameter and the second correlation parameter .

[0037] For the defect parameter in the first correlation parameter , use a convolutional neural network to extract spatial feature data to obtain the first feature vector ;

[0038] For the numerical - type parameter optimized form index parameter and the average density , perform normalization processing to obtain the second feature vector , directly splice the first feature vector with the second feature vector to obtain the first comprehensive feature representation ;

[0039] For the temperature change range and the humidity change rate in the second correlation parameter , directly perform normalization linear conversion processing to obtain the second comprehensive feature representation ;

[0040] Take and as the input of the multi - head attention mechanism, and obtain the optimized correlation index through the output of the multi - head attention mechanism.

[0041] The process of extracting the trend data is as follows:

[0042] Input the optimized correlation index into a capsule network model;

[0043] The optimized correlation index data obtained after preliminary mapping is assigned to different capsules according to the capsule rules in the capsule network model;

[0044] After the assignment, different capsules output specific multi - dimensional vectors ;

[0045] For the multi - dimensional vector output by the th capsule at the th time point, it is represented as ;

[0046] Statistically analyze the modulus lengths and trend directions corresponding to the optimized correlation metrics for all capsules;

[0047] Organize the modulus lengths of each capsule at different time points into a matrix , where , ;

[0048] At adjacent time points and , let the change rate of the output vector modulus length of the -th capsule be :

[0049] When , it indicates that the features captured by the -th capsule are on the increase trend. When , it indicates that the features are on the decay trend. Organize the change rates of all capsules at different time points into a matrix , where , ;

[0050] Finally, form a data sequence , where represents the importance matrix, represents the trend direction matrix;

[0051] This data sequence is the trend data.

[0052] The sheet material characteristic data includes the tensile strength , compressive strength , and flexural strength data .

[0053] The process for obtaining the decay mode is as follows:

[0054] Obtain the relevant target characteristic index c of the sheet material characteristic data;

[0055] Calculate the dynamic weight factor for each characteristic data according to the target characteristic index c. The formula is expressed as:

[0056]

[0057] where is the mean square error of the neural network model for the target characteristic index c, is the weight coefficient;

[0058] Obtain the Pearson correlation coefficient between the sheet material characteristic data and the trend data:

[0059] Multiply the dynamic weight factor Assign the Pearson correlation coefficient Obtain the optimized cross-correlation coefficient ;

[0060] Set a correlation threshold ;

[0061] When , it is considered that there is a strong correlation between the sheet property data and the trend data;

[0062] When , it is considered that there is a weak correlation between the sheet property data and the trend data;

[0063] Calculate the difference of the trend data at adjacent time points , when at multiple consecutive time points , then it is considered that the trend data is in a decaying state;

[0064] For the sheet property data with a decaying trend data and a strong correlation ;

[0065] If and the trend data decays, then the sheet property data shows an increasing situation;

[0066] If and the trend data decays, then the sheet property data shows a decaying situation;

[0067] Determine the decay mode of the sheet according to the decay situation of the sheet property data.

[0068] The process of obtaining the corrected sheet property data is as follows:

[0069] For the sheet property data determined to have a decay mode, establish a linear decay compensation model ;

[0070] Let the decay relationship between the sheet property data and the trend data be: , where is the decay coefficient, indicating that any type of data in the sheet property data has decay;

[0071] Collect the sheet property data and the corresponding trend data, and form a series of sample points , and divide the collected series of data into a training set and a validation set as the training data of the linear compensation model;

[0072] Substitute the change amount of the trend data into the established linear compensation relationship to calculate the predicted change amount of the sheet property data ;

[0073] The predicted change amount is compared with the actual change amount of the sheet material characteristic data in the validation set and the mean square error method is used to minimize the loss function;

[0074] After minimizing the loss function, a trained linear compensation model is obtained, and the predicted attenuation value is calculated according to the output result of the linear compensation model.

[0075] A method for detecting the structural performance of furniture sheets, the implementation steps include:

[0076] S1: Collect sheet structure data and environment-related data, extract the first correlation parameter from the collected sheet structure parameters, and extract the second correlation parameter from the relevant environment parameters;

[0077] S2: Perform cross-modal fusion analysis on the first correlation parameter and the second correlation parameter to obtain an optimized correlation index, and obtain trend data from the optimized correlation index;

[0078] S3: Collect sheet material characteristic data, optimize the cross-correlation coefficient between the sheet material characteristic data and the trend data, perform correlation analysis based on the optimized cross-correlation coefficient to determine the attenuation mode, and perform attenuation compensation on the collected real-time sheet material characteristic data based on the attenuation mode to obtain corrected sheet material characteristic data;

[0079] S4: Obtain the detection result of the furniture sheet based on the corrected sheet material characteristic data.

[0080] The present invention has the following beneficial effects:

[0081] In the present invention, first, the system collects environmental temperature and humidity data in real time and performs cross-modal fusion analysis with the sheet structure data. The trend data is extracted through a capsule network and a multi-head attention mechanism, and the correlation analysis is optimized by combining a dynamic weight factor, realizing the dynamic adjustment of detection parameters with environmental changes, avoiding overestimating performance, and solving the problem that the fixed algorithm cannot adapt to the decrease in strength after the sheet is affected by moisture;

[0082] Secondly, multi-sensors such as industrial CT and 3D laser scanners are used to collect the three-dimensional structure, density distribution and surface geometry data of the sheet, and the temperature and humidity time series data are obtained by combining a distributed environmental sensor network. At the same time, the global morphological features are extracted through a PointNet++ fusion network model, and the edge smoothness and surface flatness are quantified by using topological data analysis. The sub-indicators are dynamically weighted by combining the analytic hierarchy process, significantly improving the correlation analysis accuracy of structural defects and environmental impacts;

[0083] Finally, based on the Pearson correlation coefficient and the decay state of the trend data, the system identifies the decay mode of the sheet material characteristic data (tensile, compressive, and flexural strengths), and corrects the real-time data through a linear compensation model. For example, during the detection in the plum rain season, if the humidity change rate increases and the flexural strength is strongly negatively correlated with the trend data , the system will automatically enhance the compensation to ensure that the corrected strength data is closer to the true performance in the actual environment. Description of the Drawings

[0084] Figure 1 It is a system block diagram of a furniture sheet material structure performance detection system and a detection method proposed by the present invention;

[0085] Figure 2 It is a method step diagram of a furniture sheet material structure performance detection system and a detection method proposed by the present invention. Detailed Embodiments

[0086] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0087] Embodiment 1

[0088] As Figure 1 shown, a furniture sheet material structure performance detection system proposed by the present invention includes:

[0089] Multi-source acquisition module: acquires sheet material structure data and environment-related data, extracts the first correlation parameter from the acquired sheet material structure parameters, and extracts the second correlation parameter from the relevant environment parameters;

[0090] Process of obtaining sheet material structure data:

[0091] Use an industrial CT scanner to perform a full-range scan on the sheet material to obtain three-dimensional structure image data of the sheet material , and at the same time use a density measurement device to obtain the density values at different positions of the sheet material , and use a three-dimensional laser scanner to obtain the geometric shape data of the surface of the sheet material ;

[0092] Process of obtaining the first correlation parameter:

[0093] Through an image segmentation algorithm, the three-dimensional structure image data Separate the defective areas (such as identifying defects, cavities, cracks, etc. inside the sheet), and obtain defect parameters ;

[0094] Extract the flatness of the sheet surface from the geometric shape data , and the contour information data of the edge , and obtain the optimized shape index parameters through the mapping relationship between the flatness of the sheet surface and the contour information data of the edge ;

[0095] Obtain the average density from the density values at different positions ;

[0096] The first correlation parameter is expressed as ;

[0097] Specifically, the specific process of obtaining the optimized shape index parameters through the flatness of the sheet surface and the contour information data of the edge is as follows:

[0098] Convert the flatness of the sheet surface and the contour information data of the edge into point cloud data ;

[0099] Through a PointNet++ model, incorporate an attention mechanism module into the middle layer of the model to obtain a PointNet++ fusion network model;

[0100] Specifically, through the PointNet++ fusion network model, the traditional PointNet++ model is good at processing unordered point cloud data and can extract local and global features. On this basis, an attention mechanism module is incorporated into the middle layer to enable the network to automatically focus on the key areas closely related to the flatness and edge contour of the sheet when processing point cloud data;

[0101] Input the point cloud data into the PointNet++ fusion network model. The fusion network model first preliminarily extracts the features of each point, uses grouping and feature extraction operations to extract the local features of the point cloud at different scales, and aggregates the features within the neighborhood of each point to obtain local feature vectors ;

[0102] The network model fuses the processed local features to obtain the global shape features that can represent the overall shape of the sheet ;

[0103] Then, topological features are obtained from the extracted global feature vectors through topological data analysis ;

[0104] According to the extracted global morphological features and topological features , a mapping relationship between the global morphological features and topological features is established;

[0105] The eigenvalue representing the continuity of the edge contour in the topological features is mapped to the smoothness index of the edge;

[0106] The components related to the surface undulation in the global feature vectors are mapped to the flatness index of the surface;

[0107] The smoothness index of the edge and the flatness index of the surface together form sub-indices ;

[0108] According to each sub-index obtained by mapping , the weights of different sub-indices are determined according to the analytic hierarchy process, and then the final optimized morphological index parameters are obtained by weighted summation , which is expressed by the formula:

[0109]

[0110] where, is the th sub-index obtained by mapping, is the number of sub-indices, is the weight of the th sub-index;

[0111] The method for collecting environment-related data is as follows:

[0112] Deploy a distributed environmental sensor network to collect environmental temperature data and relative humidity data ;

[0113] The method for obtaining the second correlation parameter is as follows:

[0114] Extract the temperature change range within a specific time period , the humidity change rate ;

[0115] Record the maximum value and minimum value of the environmental temperature within a period of time , and the difference between the two is the temperature change range ;

[0116] Specifically, the temperature change range has an important impact on the performance of the board. For example, a large temperature change may cause the board to expand and contract thermally, resulting in internal stress and affecting its structural stability.

[0117] By measuring the relative humidity continuously for a period of time and calculating the humidity change rate ;

[0118] Specifically, the humidity change rate reflects the speed of environmental humidity change. A rapid humidity change may cause the moisture content of the board to change rapidly, leading to problems such as board deformation and cracking.

[0119] These parameters constitute the second associated parameter set .

[0120] Association acquisition module: Perform cross-modal fusion analysis on the first associated parameter and the second associated parameter to obtain an optimized association index, and obtain trend data from the optimized association index.

[0121] The process of obtaining the optimized association index is as follows:

[0122] Perform feature encoding representation on the first associated parameter and the second associated parameter ;

[0123] For the defect parameter in the first associated parameter , extract spatial feature data using a convolutional neural network to obtain the first feature vector ;

[0124] For the optimized morphological index parameter and the average density of numerical parameters, perform normalization processing to obtain the second feature vector , directly splice the first feature vector with the second feature vector to obtain the first comprehensive feature representation ;

[0125] Perform direct normalization linear transformation processing on the temperature change range and the humidity change rate in the second associated parameter to obtain the second comprehensive feature representation ;

[0126] Specifically, this step is to convert parameters in different forms into a feature vector form suitable for subsequent attention mechanism processing.

[0127] Take and As the input of the multi-head attention mechanism, an optimized correlation metric is obtained through the output of the multi-head attention mechanism ;

[0128] The process of extracting trend data is as follows:

[0129] Input the optimized correlation metric data obtained through the multi-head attention transfer mechanism into a capsule network model;

[0130] In the capsule network model, the original optimized correlation metric data is mapped to a new feature space so that the latent features in the data can be better represented and processed;

[0131] The optimized correlation metric data obtained after the preliminary mapping is assigned to different capsules according to the capsule rules in the capsule network model;

[0132] Specifically, the capsule network model is based on the sensitivity and capture ability of each capsule to different types of metric data. For example, features related to the internal texture structure of the board will be assigned to capsules specifically responsible for processing structural features; while features related to changes in environmental humidity and temperature will be assigned to capsules for corresponding environmental factor processing;

[0133] After the assignment, different capsules will output specific multi-dimensional vectors ;

[0134] For the multi-dimensional vector output by the th capsule at the th time point, it is denoted as ;

[0135] Observe the magnitude and trend of the multi-dimensional vector. The magnitude represents the intensity of the features captured by the capsule, that is, the degree of importance, and the trend of the vector reflects whether the pattern of the features is increasing or decaying;

[0136] The magnitude of the multi-dimensional vector is denoted as: , where represents the th component (i.e., the corresponding optimized correlation metric) in the multi-dimensional vector ;

[0137] The trend is reflected by the change of the magnitude over time. At adjacent time points and , compare the magnitudes of the optimized correlation metrics corresponding to the th capsule. If the optimized correlation metric corresponding to time point is greater than the optimized correlation metric of , it means that the optimized correlation metric in the th capsule is in a decaying trend; If the optimized correlation metric corresponding to time point

[0138] Statistically analyze the importance (magnitude) and trend corresponding to the optimized correlation index for all capsules to form a data sequence , where represents the importance matrix, represents the trend matrix, and this data sequence is the trend data;

[0139] The specific process of obtaining the data sequence is as follows:

[0140] Arrange the magnitudes of each capsule at different time points into a matrix , where , so that the change of the characteristic intensity of different capsules over time can be visually observed;

[0141] At adjacent time points and , let the change rate of the output vector magnitude of the th capsule be :

[0142]

[0143] When , it means that the characteristics captured by the th capsule show an increasing trend; when , it means that the characteristics show a decaying trend. Arrange the change rates of all capsules at different time points into a matrix , where .

[0144] Data correction module: Collect the data of the board characteristics, optimize the cross-correlation coefficient between the data of the board characteristics and the trend data, perform correlation analysis based on the optimized cross-correlation coefficient to determine the attenuation mode, and perform attenuation compensation on the collected real-time data of the board characteristics based on the attenuation mode to obtain the corrected data of the board characteristics;

[0145] The data of the board characteristics include the tensile strength , the compressive strength , and the flexural strength data ;

[0146] Specifically, these three strength data intuitively show the bearing capacity of the board under different stress modes. In an environment with higher humidity, the strength of solid wood boards will decrease after being affected by moisture. By collecting the data of tensile strength, compressive strength, and flexural strength in real time, the change of the actual strength performance of the board can be directly reflected;

[0147] Obtain the attenuation mode;

[0148] The process of obtaining the attenuation mode is as follows:

[0149] First, optimize the cross-correlation coefficient between the sheet material characteristic data and the trend data;

[0150] Modify the traditional correlation coefficient calculation method through dynamic weighting of the target characteristic index between the sheet material characteristic data and the environment to obtain the optimized cross-correlation coefficient between the sheet material characteristic data and the trend data;

[0151] Obtain the relevant target characteristic index of the sheet material characteristic data:

[0152] Let the collected sheet material characteristic data be ;

[0153] Analyze the correlation between different characteristic data and environmental factors, and preliminarily judge the degree of association between them , and then use a neural network model to take the environmental data and the characteristic data of the sheet material as the input layer, and the target characteristic index as the output layer. Through the degree of association Train the network model with historical data to learn the mapping relationship between them, and the model finally outputs the target characteristic index c;

[0154] Calculate the dynamic weight factor for each characteristic data according to the target characteristic index c , and the formula is expressed as:

[0155]

[0156] Among them, is the mean square error of the neural network model for the target characteristic index c, is the weight coefficient;

[0157] Obtain the Pearson correlation coefficient between the sheet material characteristic data and the trend data :

[0158] Assign the dynamic weight factor to the Pearson correlation coefficient to obtain the optimized cross-correlation coefficient ;

[0159] Set a correlation threshold , which is used to judge the strength of the correlation. When , it is considered that there is a strong correlation between the sheet material characteristic data and the trend data;

[0160] When , it is considered that there is a weak correlation between the sheet material characteristic data and the trend data;

[0161] Continuously monitor the change of the trend data , and judge whether it shows a decay trend. By calculating the difference of the trend data at adjacent time points , if at multiple consecutive time points , it is considered that the trend data is in a decaying state;

[0162] For the trend data in the decaying state and having a strong correlation ( ), further analyze the change situation of the sheet material characteristic data. If and the trend data decays, then the sheet material characteristic data shows an enhancing situation;

[0163] If and the trend data decays, then the sheet material characteristic data shows a decaying situation;

[0164] Determine the decay mode of the sheet material according to the decay situation of the sheet material characteristic data;

[0165] Further traverse the tensile strength , compressive strength , and flexural strength data of the sheet material characteristic data for the correlation with the trend data. If and the trend data continuously decays, it represents that there is a decay mode in the tensile strength, denoted as , indicating that "compressive strength - trend data has decay";

[0166] For the sheet material characteristic data determined to have a decay mode, establish a linear decay compensation model ;

[0167] Let the decay relationship between the sheet material characteristic data and the trend data be: , where is the decay coefficient, indicating that there is decay in any type of data in the sheet material characteristic data;

[0168] Collect the sheet material characteristic data and the corresponding trend data within a random time span to form a series of sample points , and divide the collected series of data into a training set and a validation set as the training data for the linear compensation model;

[0169] According to the ratio of 7:3, use the first 70% of the data as the training set for fitting, and the last 30% of the data as the validation set;

[0170] Use the validation set data, substitute the change amount of the trend data into the established linear compensation relationship to calculate the predicted change amount of the sheet material characteristic data, and then compare it with the actual change amount of the sheet material characteristic data in the validation set, and use the mean square error method to minimize the loss function;

[0171] After minimizing the loss function, a trained linear compensation model is obtained. Based on the output result of the linear compensation model, the predicted attenuation value is calculated, and then the real-time collected sheet data is corrected, which is expressed as;

[0172] Specifically, the predicted attenuation value calculated according to the output result of the linear compensation model is obtained through the trained linear compensation model, which is used to reflect the expected reduction of the sheet characteristic data due to the change of the trend data;

[0173] Specifically, in the established linear compensation relationship, the change amount of the real-time collected trend data is substituted into the model, and the change amount of the sheet characteristic data calculated is the predicted attenuation value, which reflects the estimation of the attenuation degree of the sheet characteristic data under the current trend data change based on the model trained with historical data.

[0174] Detection output module: Based on the corrected sheet characteristic data, obtain the detection result of the furniture sheet;

[0175] Among the corrected sheet characteristic data, the inherent properties of the sheet itself can be reflected by the sheet structure data, while the environment-related data reflects the external conditions where the sheet is located. Only by mastering these two aspects of data at the same time can we comprehensively understand the performance state of the sheet in actual use. For example, for solid wood sheets in different temperature and humidity environments, their internal structures will be affected, thereby changing their performance and affecting the detection accuracy;

[0176] Since the performance of the sheet will change due to various factors such as the environment, and the original sheet structure data cannot accurately reflect its current true performance, by performing attenuation compensation on the real-time collected sheet characteristic data based on the attenuation mode, these data can be corrected to make them more in line with the actual situation. After correcting the collected strength data and directly performing conventional detection on it, a more accurate detection result of the furniture sheet can be obtained.

[0177] Embodiment 2

[0178] As Figure 2 shown, a method for detecting the structural performance of a furniture sheet proposed by the present invention includes:

[0179] S1: Collect sheet structure data and environment-related data, extract the first correlation parameter from the collected sheet structure parameters, and extract the second correlation parameter from the relevant environment parameters;

[0180] S2: Perform cross-modal fusion analysis on the first correlation parameter and the second correlation parameter to obtain an optimized correlation index, and obtain trend data from the optimized correlation index;

[0181] S3: Collect the data of the board characteristics, optimize the cross-correlation coefficient between the board characteristics data and the trend data, conduct a correlation analysis based on the optimized cross-correlation coefficient to determine the attenuation mode, and perform attenuation compensation on the collected real-time board characteristics data based on the attenuation mode to obtain the corrected board characteristics data;

[0182] S4: Obtain the test result of the furniture board based on the corrected board characteristics data.

[0183] In the application, several formulas involved are calculated by taking their numerical values after dimensionless. The establishment of the formulas is obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. Some coefficients or weights in the formulas are set by those skilled in the art according to the actual situation, so no more details will be given here.

[0184] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution.

[0185] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A furniture board structural performance detection system, characterized in that: include: Multi-source acquisition module: collects plate structure data and environment-related data, extracts the first associated parameter from the collected plate structure parameters, and extracts the second associated parameter from the related environment parameters; Correlation acquisition module: performs cross-modal fusion analysis on the first correlation parameter and the second correlation parameter to obtain optimized correlation indicators, and obtains trend data from the optimized correlation indicators; Data correction module: collects plate characteristic data, optimizes the correlation coefficient between the plate characteristic data and trend data, performs correlation analysis based on the optimized correlation coefficient to determine the attenuation mode, performs attenuation compensation on the collected real-time plate characteristic data based on the attenuation mode, and obtains the corrected plate characteristic data; Detection output module: obtains the detection results of furniture boards based on the corrected board characteristic data.

2. A furniture board structure performance detection system according to claim 1, characterized in that: The plate structure data acquisition process is as follows: Use an industrial CT scanner to scan the plate in all directions to obtain the three-dimensional structural image data of the plate , and use density measuring equipment to obtain the density values ​​of different positions of the board , using 3D laser scanner to obtain the geometric shape data of the plate surface ; The environment-related data collection method is: Deploy a distributed environmental sensor network to collect ambient temperature data and relative humidity data .

3. A furniture board structure performance detection system according to claim 1, characterized in that: The first associated parameter acquisition process is: Through the image segmentation algorithm, the three-dimensional structure image data Separate the defective area and obtain the defect parameters ; From geometry data Extract the flatness of the board surface , edge contour information data , through the mapping relationship between the flatness of the plate surface and the contour information data of the edge, the optimized morphological index parameters are obtained ; Density values ​​from different locations Get the average density ; The first associated parameter is expressed as ; The second associated parameter acquisition process is: Extract a specific time period Temperature range within , humidity change rate ; Record a period of time Maximum internal ambient temperature and minimum value The difference between the two is the temperature range ; By relative humidity For a period of time Continuously monitor and calculate humidity change rate ; The second set of associated parameters is represented as .

4. A furniture board structure performance detection system according to claim 3, characterized in that: The optimized morphological indicator parameters The acquisition process is: The flatness of the board surface , edge contour information data Convert to point cloud data ; Through a Model, integrate the attention mechanism module into the middle layer of the model to obtain Fusion network model; Point cloud data Input into the PointNet++ fusion network model to obtain the local feature vector ; The network model fuses the processed local features to obtain the global morphological features. ; Then, the extracted global feature vector is analyzed through topological data to obtain the topological features. ; According to the extracted global morphological features And topological features , establish the mapping relationship between global morphological features and topological features; Mapping the feature value representing the continuity of edge contour in the topological feature to the smoothness index of the edge; Mapping the components of the global eigenvector related to surface relief into surface flatness indicators; The edge smoothness index and the surface flatness index together form a sub-index ; According to the sub-indicators obtained by mapping , determine the weights of different sub-indicators according to the hierarchical analysis method, and then obtain the final optimized morphological index parameters through weighted summation .

5. A furniture board structure performance detection system according to claim 1, characterized in that: The process of obtaining the optimization correlation index is as follows: For the first associated parameter and the second associated parameter Perform feature encoding representation; For the first associated parameter Defect parameters in , a convolutional neural network is used to extract spatial feature data and obtain the first feature vector ; Optimizing morphological indicator parameters for numerical parameters and the average density , normalize it to get the second eigenvector , directly take the first eigenvector Concatenate with the second feature vector to get the first comprehensive feature representation ; For the second associated parameter Temperature range and humidity change rate Directly perform normalized linear transformation to obtain the second comprehensive feature representation ; Will and As the input of the multi-head attention mechanism, the optimization correlation index is obtained through the output of the multi-head attention mechanism .

6. A furniture board structure performance detection system according to claim 1, characterized in that: The process of extracting the trend data is as follows: Will optimize the associated indicators Input into a capsule network model; The optimized correlation index data obtained after preliminary mapping is allocated to different capsules according to the capsule rules in the capsule network model; After the allocation is completed, different capsules output specific multidimensional vectors ; For The capsule in The multidimensional vector output at each time point is expressed as ; Count the modulus lengths and trend trends of the optimized correlation indicators in all capsules; Arrange the modulus of each capsule at different time points into a matrix ,in , ; At adjacent time points and , set The rate of change of the modulus of the capsule output vector : when When The features captured by the capsules tend to increase. When , it indicates that the feature is attenuating, and the change rates of all capsules at different time points are organized into a matrix ,in , ; Finally, a data sequence is formed ,in, represents the importance matrix, Represents the trend matrix; This data series is called trend data.

7. A furniture board structure performance detection system according to claim 1, characterized in that: The plate characteristic data includes tensile strength , compressive strength , flexural strength data .

8. A furniture board structure performance detection system according to claim 1, characterized in that: The acquisition process of the attenuation mode is: Obtain relevant target characteristic index c of plate characteristic data; Calculate the dynamic weight factor for each characteristic data according to the target characteristic index c , the formula is: in, is the mean square error of the neural network model for the target characteristic index c, is the weight coefficient; Get the Pearson correlation coefficient between the plate characteristic data and the trend data : Dynamic weight factor Assign Pearson correlation coefficient Get the optimized cross-correlation coefficient ; Set a relevance threshold ; when When , it is considered that there is a strong correlation between the plate characteristic data and the trend data; when When , it is considered that there is a weak correlation between the plate characteristic data and the trend data; Calculate the difference between trend data at adjacent time points , when at multiple consecutive time points , then the trend data is considered to be in a decaying state; For trend data that is in a decaying state and has a strong correlation Plate characteristic data; like And the trend data decays, then the plate characteristic data increases; like And if the trend data decays, then the plate characteristic data decays; According to the attenuation of the plate characteristic data, the attenuation mode of the plate is determined.

9. A furniture board structure performance detection system according to claim 1, characterized in that: The process of obtaining the corrected plate characteristic data is as follows: For the plate characteristic data with attenuation mode, a linear attenuation compensation model is established. ; Assume plate characteristic data With trend data The attenuation relationship between them is: ,in, is the attenuation coefficient, indicating that there is attenuation in any type of data in the plate characteristic data; Collect plate characteristic data and corresponding trend data to form a series of sample points , the collected series of data are divided into a training set and a validation set as training data for the linear compensation model; Substitute the change in trend data into the established linear compensation relationship Calculate the predicted change in the plate characteristic data ; The predicted change The change in the actual plate characteristic data in the validation set Compare and minimize the loss function using the mean square error method; After minimizing the loss function, the trained linear compensation model is obtained, and the predicted attenuation value is calculated according to the output result of the linear compensation model.

10. A method for detecting structural performance of furniture panels, using the detection system according to any one of claims 1 to 9, characterized in that: The implementation steps include: S1: Collecting plate structure data and environment-related data, extracting first associated parameters from the collected plate structure parameters, and extracting second associated parameters from related environment parameters; S2: performing cross-modal fusion analysis on the first correlation parameter and the second correlation parameter to obtain an optimized correlation index, and obtaining trend data from the optimized correlation index; S3: collecting plate characteristic data, optimizing the correlation coefficient between the plate characteristic data and the trend data, performing correlation analysis based on the optimized correlation coefficient to determine the attenuation mode, performing attenuation compensation on the collected real-time plate characteristic data based on the attenuation mode, and obtaining the corrected plate characteristic data; S4: Based on the corrected board characteristic data, obtain the test results of the furniture board.

Images