Method and device for detecting production quality of fire-resistant window

Abstract

The invention discloses a fire-resistant window production quality detection method and device, belongs to the technical field of fire-resistant window quality detection, and is used for solving the technical problems that the current fire-resistant window product quality detection method can damage a detection sample and cannot detect products one by one, the risk of potential quality hazards is increased, and certain product loss is caused. The method comprises the following steps: acquiring spectrograms and sound wave detection signals of a plurality of fire-resistant window samples, and extracting spectral features of the spectrograms to obtain first quality features; extracting a signal feature of the sound wave detection signal to obtain a second quality feature; fusing the first quality feature and the second quality feature to obtain a corresponding sample quality feature; determining a model training data set based on the sample quality characteristics and the corresponding fire resistance test results; training the fire-resistant window quality detection model through the model training data set; and acquiring a spectrogram and a sound wave detection signal of each fire-resistant window product in real time, and inputting the spectrogram and the sound wave detection signal into the fire-resistant window quality detection model for detection.

Description

Technical Field

[0001] The present invention relates to the technical field of fire-resistant window quality detection, and in particular to a method and device for detecting the production quality of fire-resistant windows. Background Art

[0002] The development of fireproof windows has a relatively long history in my country, and there is a relatively mature system for the research and specification of fireproof windows' fire resistance performance, but fireproof windows are mainly used in indoor partition walls such as fire walls and fireproof partition walls. With the increasing number of high-rise buildings, in order to prevent the rapid spread of fire in high-rise buildings, there are more and more studies on the fire resistance of external partition walls and windows of buildings, and fire resistance integrity requirements are proposed for external walls and windows. The purpose is to alleviate the pressure on emergency rescue capabilities in domestic high-rise buildings by strengthening their own passive fire defense capabilities.

[0003] Since the research on fire-resistant windows has just started, the current quality inspection of fire-resistant windows still mainly relies on manual fire resistance tests, such as physical methods such as combustion tests. This method is not only time-consuming and labor-intensive, requiring a large amount of manpower and testing resources, but also often damages the samples during the inspection process. The tested samples can no longer be sold as normal products, resulting in the inability to conduct quality inspections on each fire-resistant window product one by one, and only sampling inspections can be adopted. This limitation not only increases the risk of quality hazards and makes it difficult to ensure the quality and safety of all products, but also increases certain product losses, causing certain economic losses to manufacturers. Therefore, it is particularly important to develop a non-destructive, comprehensive, accurate and reliable fire-resistant window quality inspection method. Summary of the invention

[0004] The embodiments of the present invention provide a method and device for detecting the quality of fire-resistant windows, which are used to solve the following technical problems: the current quality detection method for fire-resistant window products will damage the detection samples, and it is impossible to detect the products one by one, which increases the risk of quality hazards and causes certain product losses.

[0005] The embodiment of the present invention adopts the following technical solutions:

[0006] On the one hand, an embodiment of the present invention provides a method for detecting the production quality of fire-resistant windows, the method comprising: collecting spectrograms and acoustic wave detection signals of a number of fire-resistant window samples, and performing a combustion test to obtain a fire resistance test result;

[0007] Extracting a spectral feature of the spectrum graph to obtain a first quality feature;

[0008] Extracting a signal feature of the acoustic wave detection signal to obtain a second quality feature;

[0009] The first quality feature and the second quality feature of the same fire-resistant window sample are merged to obtain the corresponding sample quality feature;

[0010] Determining a model training data set based on the sample quality characteristics and corresponding fire resistance test results;

[0011] Constructing a fire-resistant window quality detection model, and training the fire-resistant window quality detection model through the model training data set;

[0012] On the constructed fire-resistant window quality inspection production line, the spectrum and acoustic wave detection signals of each fire-resistant window product are collected in real time, and input into the fire-resistant window quality inspection model for inspection to obtain the quality assessment results of each fire-resistant window product.

[0013] In a feasible implementation, the spectra and acoustic wave detection signals of several fire-resistant window samples are collected, and a combustion test is performed to obtain the fire resistance test results, which specifically include:

[0014] In the conventional device for quality inspection of fire-resistant windows through combustion test, a spectrum transceiver and a sound wave transceiver are added; the spectrum transceiver and the sound wave transceiver can work simultaneously and are arranged in the link before the combustion test device;

[0015] Before each fire-resistant window sample is subjected to a combustion test, a spectrum diagram of multiple key detection points of the fire-resistant window sample is obtained through the spectrum transceiver; wherein the spectrum transceiver includes a continuous spectrum emitting device and an absorption spectrum collecting device;

[0016] Before each fire-resistant window sample is subjected to a combustion test, the sound wave detection signals of multiple key detection points of the fire-resistant window sample are obtained through the sound wave transceiver; wherein the sound wave transceiver includes an ultrasonic transmitting device and an ultrasonic receiving device;

[0017] After each fire-resistant window sample is subjected to a combustion test and a fire-resistant test result is obtained, the fire-resistant test result is associated and stored with a corresponding spectrum graph and an acoustic wave detection signal;

[0018] After a preset amount of sample data is collected, data collection is stopped and the spectrum transceiver and the sound wave transceiver are removed.

[0019] In a feasible implementation manner, extracting the spectral feature of the spectrum graph to obtain the first quality feature specifically includes:

[0020] Performing basic analysis on each spectrum graph to obtain basic spectrum characteristics; wherein the basic spectrum characteristics at least include: redshift / blueshift characteristics, peak position characteristics, peak width characteristics, peak intensity characteristics, peak shape characteristics, and wavelength range characteristics;

[0021] Through the DB4 wavelet basis in the Daubeehies wavelet function, each spectrum is transformed n times by wavelet transform to obtain the discrete detail signal of the spectrum.

[0022] Performing wavelet decomposition on the discrete detail signal to obtain wavelet component features;

[0023] The basic spectrum feature and the wavelet component feature constitute the first quality feature.

[0024] In a feasible implementation manner, extracting the signal feature of the acoustic wave detection signal to obtain the second quality feature specifically includes:

[0025] Perform Fourier transform on each acoustic wave detection signal to obtain initial frequency domain features;

[0026] Performing inverse Fourier transform on the initial frequency domain features to obtain a harmonic wavelet function;

[0027] Decomposing the acoustic wave detection signal by frequency band by using the harmonic wavelet function to obtain a discrete signal;

[0028] The discrete signal is input into a pre-trained convolutional neural network for further feature extraction to obtain the second quality feature.

[0029] In a feasible implementation, the first quality feature and the second quality feature of the same fire-resistant window sample are merged to obtain the corresponding sample quality feature, specifically including:

[0030] assigning initial contribution weight values ​​to the first quality feature and the second quality feature respectively;

[0031] Based on the initial contribution weight value, determining the fusion ratio of the two feature channels, and performing feature fusion based on the fusion ratio to obtain the initial sample quality feature;

[0032] The fire-resistant window quality detection model is trained by the initial sample quality features and the corresponding fire resistance test results, and the learning effect of the fire-resistant window quality detection model on the two parts of features and the contribution of the two parts of features to the recognition accuracy are analyzed during the training process;

[0033] According to the learning effect and contribution, the initial contribution weight value is adjusted until the recognition accuracy of the model is higher than a preset threshold. According to the contribution weight value at this time, the first quality feature and the second quality feature of all fire-resistant window samples are fused to obtain the final sample quality feature.

[0034] In a feasible implementation manner, based on the sample quality characteristics and the corresponding fire resistance test results, determining a model training data set specifically includes:

[0035] Extract the fire resistance test data and fire resistance scores of each key detection point from the fire resistance test results;

[0036] According to the fire resistance scores, divide the key detection points into positive detection points and negative detection points; the positive detection points are the detection points with qualified fire resistance, and the negative detection points are the detection points with unqualified fire resistance;

[0037] After associating the sample quality characteristics of the same positive detection point with the fire resistance test data and the fire resistance scores, store them in the dataset to form a positive sample dataset;

[0038] After associating the sample quality characteristics of the same negative detection point with the fire resistance test data and the fire resistance scores, store them in the dataset to form a negative sample dataset.

[0039] In a feasible implementation manner, construct a fire-resistant window quality detection model, and train the fire-resistant window quality detection model through the model training dataset, specifically including:

[0040] Combine the GhostNet network and the YOLOv4 network, and introduce an attention mechanism to construct the fire-resistant window quality detection model; the input data of the fire-resistant window quality detection model is a spectrogram and an acoustic wave detection signal, and the output result is the quality evaluation result of each key detection point;

[0041] Train the attention feature fusion network and the feature recognition module in the fire-resistant window quality detection model through the positive sample dataset and the negative sample dataset respectively, and perform parameter tuning based on the loss function until the model converges.

[0042] In a feasible implementation manner, combine the GhostNet network and the YOLOv4 network, and introduce an attention mechanism to construct the fire-resistant window quality detection model, specifically including:

[0043] Based on the Chost model in the GhostNet network and a preset feature extraction algorithm, construct a feature extraction module; the feature extraction module includes a first feature extraction module and a second feature extraction module, and the two feature extraction modules operate independently to respectively extract the first quality feature and the second quality feature;

[0044] Based on a preset feature fusion algorithm, construct a feature fusion module; the output of the overall feature extraction module is the input of the feature fusion module;

[0045] Combine the CBAM attention mechanism, the spatial pyramid pooling layer and the attention path aggregation network to construct an attention feature fusion network;

[0046] Based on the YOLO-Head module of the YOLOv4 model, a feature recognition module is constructed, and the Focal Loss loss function is integrated into the feature recognition module;

[0047] The feature extraction module, the feature fusion module, the input module, the attention feature fusion network and the feature recognition module are connected in sequence to form the fire-resistant window quality detection model.

[0048] In a feasible implementation, the spectrum and acoustic wave detection signals of each fire-resistant window product are collected in real time on the constructed fire-resistant window quality inspection line, and input into the fire-resistant window quality inspection model for inspection to obtain the quality assessment results of each fire-resistant window product, specifically including:

[0049] Based on the spectrum transceiver and the sound wave transceiver, a non-destructive fire-resistant window quality inspection line is built; the preset position of the fire-resistant window quality inspection line is equipped with a spectrum transceiver and a sound wave transceiver that are relatively arranged; the spectrum transceiver and the sound wave transceiver work simultaneously;

[0050] Through the spectrum transceiver, a real-time spectrum diagram of the fire-resistant window product to be inspected is obtained; through the sound wave transceiver, a real-time sound wave detection signal of the fire-resistant window product to be inspected is obtained;

[0051] The real-time spectrum graph and the real-time acoustic wave detection signal are input into the fire-resistant window quality detection model to obtain an output quality assessment result; wherein the quality assessment result at least includes: fire resistance prediction data and fire resistance prediction score of each key detection point in the current fire-resistant window product to be inspected;

[0052] According to the quality assessment result, the current fire-resistant window product to be inspected is transferred to the corresponding assembly line branch, and the quality inspection of the next fire-resistant window product to be inspected is carried out; the assembly line branches at least include a quality inspection qualified branch and a quality inspection unqualified branch.

[0053] On the other hand, an embodiment of the present invention further provides a fire-resistant window production quality detection device, the device comprising:

[0054] A model training module is used to collect spectrograms and acoustic wave detection signals of several fire-resistant window samples, and perform combustion tests to obtain fire resistance test results; extract spectral features of the spectrogram to obtain a first quality feature; extract signal features of the acoustic wave detection signal to obtain a second quality feature; fuse the first quality feature and the second quality feature of the same fire-resistant window sample to obtain a corresponding sample quality feature; determine a model training data set based on the sample quality feature and the corresponding fire resistance test results; construct a fire-resistant window quality detection model, and train the fire-resistant window quality detection model through the model training data set;

[0055] The quality assessment module is used to collect the spectrum and acoustic wave detection signals of each fire-resistant window product in real time on the constructed fire-resistant window quality inspection line, and extract the corresponding sample quality characteristics, input them into the fire-resistant window quality inspection model for inspection, and obtain the quality assessment results of each fire-resistant window product.

[0056] Compared with the prior art, the fire-resistant window production quality detection method and device provided by the embodiment of the present invention have the following beneficial effects:

[0057] The present invention constructs a non-destructive fire-resistant window quality detection method by combining non-contact detection methods such as spectral detection and acoustic wave detection with a neural network model. A large amount of detection data is collected in the traditional quality inspection process, and features related to the fire resistance results are extracted. Then, the improved neural network model is used to deeply learn the relationship between these data features and the fire resistance performance of the fire-resistant window, so that in the actual quality inspection process, the fire resistance performance of the fire-resistant window can be predicted with high precision through physical properties such as spectral features and acoustic wave features. The present invention does not need to perform sampling combustion detection on the fire-resistant window. With the new quality inspection line proposed by the present invention, it can realize the quality inspection of the fire-resistant window products one by one. The quality inspection of a product can be completed in a very short time, and it can be automatically classified into qualified products and unqualified products, which greatly improves the detection efficiency. The present invention not only increases the comprehensiveness of the quality inspection of the fire-resistant window, but also reduces the detection loss. It is obvious that it has extremely high practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0059] Figure 1 A flow chart of a method for detecting the quality of fire-resistant window production provided by an embodiment of the present invention;

[0060] Figure 2 A schematic structural diagram of a fire-resistant window production quality detection device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0062] The embodiment of the present invention provides a method for detecting the production quality of fire-resistant windows. Figure 1 As shown, the fire-resistant window production quality detection method specifically includes steps S101-S106:

[0063] S101, collecting spectra and acoustic wave detection signals of several fire-resistant window samples, and performing combustion tests to obtain fire resistance test results.

[0064] Specifically, in the conventional device for quality inspection of fire-resistant windows through combustion test, a spectrum transceiver and an acoustic wave transceiver are added. The spectrum transceiver and the acoustic wave transceiver can work simultaneously and are arranged in the link before the combustion test device.

[0065] Furthermore, before each fire-resistant window sample is subjected to a combustion test, a spectrum diagram of multiple key detection points of the fire-resistant window sample is obtained through a spectrum transceiver, wherein the spectrum transceiver includes a continuous spectrum emitting device and an absorption spectrum collecting device.

[0066] Before each fire-resistant window sample is subjected to a combustion test, the sound wave detection signals of multiple key detection points of the fire-resistant window sample are obtained through a sound wave transceiver, wherein the sound wave transceiver includes an ultrasonic transmitter and an ultrasonic receiver.

[0067] The above two data collection processes can be carried out simultaneously without affecting each other.

[0068] Furthermore, after each fire-resistant window sample is subjected to a combustion test and a fire resistance test result is obtained, the fire resistance test result is associated and stored with the corresponding spectrum graph and the acoustic wave detection signal. After a preset number of sample data are collected, data collection is stopped and the spectrum transceiver and the acoustic wave transceiver are removed.

[0069] As a feasible implementation method, in the existing quality inspection process of fire-resistant windows, for the quality inspection samples, spectrum test and sound wave test are carried out before the combustion test. A continuous spectrum is emitted to the sample by a continuous spectrum emission device, and the aluminum alloy material of the sample and the like absorb a part of the spectrum and is collected by a relatively arranged absorption spectrum collection device. In addition, an ultrasonic wave is emitted by an ultrasonic emission device, and the ultrasonic wave is received by a relatively arranged ultrasonic receiving device after penetrating the sample. The continuous spectrum and the ultrasonic wave will undergo certain changes after passing through the sample. The present invention analyzes the relationship between the characteristics of this change and the fire resistance performance, so that the neural network model learns this correlation, thereby being able to measure the fire resistance performance of the fire-resistant window through non-contact and non-destructive physical quantities.

[0070] Furthermore, when collecting data sets, the present invention does not conduct combustion experiments on a large number of fire-resistant windows to obtain data, but installs a device to collect data during the normal quality inspection process of fire-resistant windows, which does not cause excessive sample waste and does not hinder or affect the normal quality inspection process. After collecting enough relevant data, the collection device can be removed to restore the original state.

[0071] S102, extracting the spectrum feature of the spectrum graph to obtain a first quality feature; extracting the signal feature of the sound wave detection signal to obtain a second quality feature.

[0072] Specifically, basic analysis is performed on each spectrum graph to obtain basic spectrum characteristics; wherein the basic spectrum characteristics at least include: redshift / blueshift characteristics, peak position characteristics, peak width characteristics, peak intensity characteristics, peak shape characteristics, and wavelength range characteristics.

[0073] Furthermore, by using the DB4 wavelet basis in the Daubeehies wavelet function, each spectrum is subjected to n wavelet transforms to obtain discrete detail signals of the spectrum. The discrete detail signals are subjected to wavelet decomposition to obtain wavelet component features. The basic spectrum features and the wavelet component features constitute the first quality features.

[0074] Further, each acoustic wave detection signal is subjected to Fourier transformation to obtain an initial frequency domain feature. The initial frequency domain feature is subjected to inverse Fourier transformation to obtain a harmonic wavelet function. The acoustic wave detection signal is subjected to frequency band decomposition through the harmonic wavelet function to obtain a discrete signal. The discrete signal is input into a pre-trained convolutional neural network for further feature extraction to obtain a second quality feature.

[0075] S103, merging the first quality feature and the second quality feature of the same fire-resistant window sample to obtain a corresponding sample quality feature.

[0076] Specifically, initial contribution weight values ​​are assigned to the first quality feature and the second quality feature respectively. Based on the initial contribution weight values, the fusion ratio of the two feature channels is determined, and feature fusion is performed based on the fusion ratio to obtain the initial sample quality feature.

[0077] Furthermore, the fire-resistant window quality detection model is trained through the initial sample quality features and the corresponding fire resistance test results, and the learning effect of the fire-resistant window quality detection model on the two parts of features and the contribution of the two parts of features to the recognition accuracy are analyzed during the training process.

[0078] Furthermore, according to the learning effect and contribution, the initial contribution weight value is adjusted until the recognition accuracy of the model is higher than the preset threshold. According to the contribution weight value at this time, the first quality characteristics and the second quality characteristics of all fire-resistant window samples are fused to obtain the final sample quality characteristics.

[0079] As a feasible implementation method, the initial contribution weight values ​​can all be set to 1. In the subsequent training process, by analyzing the contribution and correlation between the two parts of the features and the test results, the contribution weights are dynamically adjusted until the detection accuracy reaches a higher preset value. Then the contribution weights of the two parts of the features are obtained and applied to the feature fusion of all samples to obtain the final sample quality features. In this way, the first quality feature and the second quality feature are fused according to the contribution value to the test results, and the two parts of the features are maximized.

[0080] S104. Determine a model training data set based on sample quality characteristics and corresponding fire resistance test results.

[0081] Specifically, the fire resistance test data and fire resistance score of each key detection point are extracted from the fire resistance test results. The fire resistance score is a score given by the tester according to the fire resistance performance during the combustion test.

[0082] Furthermore, according to the fire resistance score, the key detection points are divided into positive detection points and negative detection points. Positive detection points are detection points that meet the fire resistance standards, and negative detection points are detection points that do not meet the fire resistance standards.

[0083] Furthermore, the sample quality features of the same positive detection point are associated with the fire resistance test data and the fire resistance score, and then stored in the data set to form a positive sample data set. The sample quality features of the same negative detection point are associated with the fire resistance test data and the fire resistance score, and then stored in the data set to form a negative sample data set.

[0084] S105, constructing a fire-resistant window quality detection model, and training the fire-resistant window quality detection model through a model training data set.

[0085] Specifically, the GhostNet network is combined with the YOLOv4 network, and the attention mechanism is introduced to build a fire-resistant window quality detection model. The input data of the fire-resistant window quality detection model are the spectrum and the sound wave detection signal, and the output result is the quality evaluation result of each key detection point.

[0086] Furthermore, the attention feature fusion network and feature recognition module in the fire-resistant window quality detection model are trained respectively through positive sample data sets and negative sample data sets, and the parameters are tuned based on the loss function until the model converges.

[0087] As a feasible implementation method, the fire-resistant window quality detection model constructed by the present invention is an improvement on the YOLOv4 network model, specifically including:

[0088] Based on the Chost model in the GhostNet network and the preset feature extraction algorithm, a feature extraction module is constructed; the feature extraction module includes a first feature extraction module and a second feature extraction module, and the two feature extraction modules operate independently to extract the first quality feature and the second quality feature respectively. Then, based on the preset feature fusion algorithm, a feature fusion module is constructed; the output of the overall feature extraction module is the input of the feature fusion module. Then, the CBAM attention mechanism, the spatial pyramid pooling layer and the attention path aggregation network are combined to construct an attention feature fusion network. Furthermore, based on the YOLO-Head module of the YOLOv4 model, a feature recognition module is constructed, and the Focal Loss loss function is integrated into the feature recognition module.

[0089] Finally, the feature extraction module, feature fusion module, input module, attention feature fusion network and feature recognition module are connected in sequence to form a fire-resistant window quality detection model.

[0090] S106. Collect the spectrum and acoustic wave detection signal of each fire-resistant window product in real time on the constructed fire-resistant window quality inspection production line, input them into the fire-resistant window quality inspection model for inspection, and obtain the quality evaluation results of each fire-resistant window product.

[0091] Specifically, based on the spectral transceiver and the acoustic wave transceiver, a non-destructive fire-resistant window quality inspection line is built; the preset positions of the fire-resistant window quality inspection line are installed with relatively set spectral transceivers and relatively set acoustic wave transceivers; the spectral transceiver and the acoustic wave transceiver work simultaneously.

[0092] Furthermore, a real-time spectrum diagram of the fire-resistant window product to be inspected is obtained through a spectrum transceiver; and a real-time sound wave detection signal of the fire-resistant window product to be inspected is obtained through a sound wave transceiver.

[0093] Furthermore, the real-time spectrum graph and the real-time acoustic wave detection signal are input into the fire-resistant window quality detection model to obtain the output quality assessment result; wherein, the quality assessment result at least includes: the fire resistance prediction data and fire resistance prediction score of each key detection point in the current fire-resistant window product to be inspected.

[0094] Furthermore, according to the quality assessment result, the current fire-resistant window product to be inspected is transferred to the corresponding assembly line branch, and the quality inspection of the next fire-resistant window product to be inspected is carried out; the assembly line branches at least include a quality inspection qualified branch and a quality inspection unqualified branch.

[0095] As a feasible implementation method, the present invention designs a quality inspection line that matches the provided fire-resistant window production quality inspection method. The fire-resistant window products pass through the data acquisition device in the line in turn. After the data is collected, it is transmitted to the fire-resistant window quality inspection model for quality inspection. The processing time is very short. It can detect whether the fire-resistant window is qualified within a few seconds or even milliseconds. If qualified, the fire-resistant window product is transported to the qualified branch line for the next step of labeling, packaging, etc. If unqualified, it is transported to the unqualified branch line for the next step of processing.

[0096] Since the present invention performs fire resistance tests on several key test points on the fire-resistant window, after detecting unqualified products, it is possible to determine whether the product is suitable for normal use after repair or modification based on the type and location of the key test points where the unqualified results appear in the product. If so, further repairs can be performed without destroying and rebuilding, thus saving a portion of production resources. For example, if the fire resistance performance of several test points on the glass of the fire-resistant window is unqualified, the product can be repaired by replacing the glass, and then quality inspection can be conducted again after repair. If it is qualified, it can be sold normally.

[0097] In addition, the embodiment of the present invention also provides a fire-resistant window production quality detection device, such as Figure 2 As shown, the fire-resistant window production quality detection device 200 specifically includes:

[0098] The model training module 210 is used to collect spectrograms and acoustic wave detection signals of several fire-resistant window samples, and perform combustion tests to obtain fire resistance test results; extract spectral features of the spectrogram to obtain a first quality feature; extract signal features of the acoustic wave detection signal to obtain a second quality feature; fuse the first quality feature and the second quality feature of the same fire-resistant window sample to obtain a corresponding sample quality feature; determine a model training data set based on the sample quality feature and the corresponding fire resistance test results; construct a fire-resistant window quality detection model, and train the fire-resistant window quality detection model through the model training data set;

[0099] A quality assessment module 220, which is used to collect spectrograms and acoustic detection signals of each fire-resistant window product in real time on the established quality inspection production line of fire-resistant windows, extract corresponding sample quality characteristics, and input them into the fire-resistant window quality detection model for detection to obtain the quality assessment results of each fire-resistant window product.

[0100] As a feasible implementation manner, the device further includes a quality inspection production line module 230, which includes a spectrum transceiver device, an acoustic wave transceiver device, a main conveyor belt, and a conveyor belt branch. It is used to collect spectrograms and acoustic wave signals of fire-resistant window products and perform quality inspections.

[0101] Each embodiment in the present invention is described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0102] The above describes specific embodiments of the present invention. Additionally, the processes depicted in the drawings do not necessarily require the specific order or consecutive order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0103] The above are only the embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, various changes and modifications can be made to the embodiments of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the embodiments of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for detecting the production quality of fire-resistant windows, characterized in that: The method comprises: Collect spectra and acoustic wave detection signals of several fire-resistant window samples, and conduct combustion tests to obtain fire resistance test results; Extracting a spectral feature of the spectrum graph to obtain a first quality feature; Extracting a signal feature of the acoustic wave detection signal to obtain a second quality feature; The first quality feature and the second quality feature of the same fire-resistant window sample are merged to obtain the corresponding sample quality feature; Determining a model training data set based on the sample quality characteristics and corresponding fire resistance test results; Constructing a fire-resistant window quality detection model, and training the fire-resistant window quality detection model through the model training data set; On the constructed fire-resistant window quality inspection production line, the spectrum and acoustic wave detection signals of each fire-resistant window product are collected in real time, and input into the fire-resistant window quality inspection model for inspection to obtain the quality assessment results of each fire-resistant window product.

2. A method for detecting the production quality of fire-resistant windows according to claim 1, characterized in that: Collect spectra and acoustic wave detection signals of several fire-resistant window samples, and conduct combustion tests to obtain fire resistance test results, including: In the conventional device for quality inspection of fire-resistant windows through combustion test, a spectrum transceiver and a sound wave transceiver are added; the spectrum transceiver and the sound wave transceiver can work simultaneously and are arranged in the link before the combustion test device; Before each fire-resistant window sample is subjected to a combustion test, a spectrum diagram of multiple key detection points of the fire-resistant window sample is obtained through the spectrum transceiver; wherein the spectrum transceiver includes a continuous spectrum emitting device and an absorption spectrum collecting device; Before each fire-resistant window sample is subjected to a combustion test, the sound wave detection signals of multiple key detection points of the fire-resistant window sample are obtained through the sound wave transceiver; wherein the sound wave transceiver includes an ultrasonic transmitting device and an ultrasonic receiving device; After each fire-resistant window sample is subjected to a combustion test and a fire-resistant test result is obtained, the fire-resistant test result is associated and stored with a corresponding spectrum graph and an acoustic wave detection signal; After a preset amount of sample data is collected, data collection is stopped and the spectrum transceiver and the sound wave transceiver are removed.

3. A method for detecting the production quality of fire-resistant windows according to claim 1, characterized in that: Extracting the spectral feature of the spectrum graph to obtain the first quality feature specifically includes: Performing basic analysis on each spectrum graph to obtain basic spectrum characteristics; wherein the basic spectrum characteristics at least include: redshift / blueshift characteristics, peak position characteristics, peak width characteristics, peak intensity characteristics, peak shape characteristics, and wavelength range characteristics; Through the DB4 wavelet basis in the Daubeehies wavelet function, each spectrum is transformed n times by wavelet transform to obtain the discrete detail signal of the spectrum. Performing wavelet decomposition on the discrete detail signal to obtain wavelet component features; The basic spectrum feature and the wavelet component feature constitute the first quality feature.

4. A method for detecting the production quality of fire-resistant windows according to claim 1, characterized in that: Extracting the signal feature of the acoustic wave detection signal to obtain the second quality feature specifically includes: Perform Fourier transform on each acoustic wave detection signal to obtain initial frequency domain features; Performing inverse Fourier transform on the initial frequency domain features to obtain a harmonic wavelet function; Decomposing the acoustic wave detection signal by frequency band by using the harmonic wavelet function to obtain a discrete signal; The discrete signal is input into a pre-trained convolutional neural network for further feature extraction to obtain the second quality feature.

5. A method for detecting the production quality of fire-resistant windows according to claim 1, characterized in that: The first quality feature and the second quality feature of the same fire-resistant window sample are merged to obtain the corresponding sample quality features, including: assigning initial contribution weight values ​​to the first quality feature and the second quality feature respectively; Based on the initial contribution weight value, determining the fusion ratio of the two feature channels, and performing feature fusion based on the fusion ratio to obtain the initial sample quality feature; The fire-resistant window quality detection model is trained by the initial sample quality features and the corresponding fire resistance test results, and the learning effect of the fire-resistant window quality detection model on the two parts of features and the contribution of the two parts of features to the recognition accuracy are analyzed during the training process; According to the learning effect and contribution, the initial contribution weight value is adjusted until the recognition accuracy of the model is higher than a preset threshold. According to the contribution weight value at this time, the first quality feature and the second quality feature of all fire-resistant window samples are fused to obtain the final sample quality feature.

6. A method for detecting the production quality of fire-resistant windows according to claim 1, characterized in that: Based on the sample quality characteristics and the corresponding fire resistance test results, a model training data set is determined, specifically including: Extracting fire resistance test data and fire resistance scores of each key detection point from the fire resistance test results; According to the fire resistance score, the key detection points are divided into positive detection points and negative detection points; the positive detection points are detection points where the fire resistance meets the standard, and the negative detection points are detection points where the fire resistance does not meet the standard; After associating the sample quality characteristics of the same positive detection point with the fire resistance test data and the fire resistance score, the samples are stored in a data set to form a positive sample data set; After associating the sample quality features of the same negative detection point with the fire resistance test data and the fire resistance score, the samples are stored in a data set to form a negative sample data set.

7. A method for detecting the production quality of fire-resistant windows according to claim 1, characterized in that: Constructing a fire-resistant window quality detection model, and training the fire-resistant window quality detection model through the model training data set, specifically including: The GhostNet network is combined with the YOLOv4 network, and the attention mechanism is introduced to construct the fire-resistant window quality detection model; the input data of the fire-resistant window quality detection model is the spectrum graph and the sound wave detection signal, and the output result is the quality evaluation result of each key detection point; The attention feature fusion network and the feature recognition module in the fire-resistant window quality detection model are trained respectively through positive sample data sets and negative sample data sets, and the parameters are tuned based on the loss function until the model converges.

8. A method for detecting the production quality of fire-resistant windows according to claim 7, characterized in that: The GhostNet network is combined with the YOLOv4 network, and the attention mechanism is introduced to construct the fire-resistant window quality detection model, which specifically includes: Based on the Chost model in the GhostNet network and the preset feature extraction algorithm, a feature extraction module is constructed; the feature extraction module includes a first feature extraction module and a second feature extraction module, and the two feature extraction modules operate independently to extract the first quality feature and the second quality feature respectively; Based on the preset feature fusion algorithm, a feature fusion module is constructed; the output of the overall feature extraction module is the input of the feature fusion module; Combine the CBAM attention mechanism, spatial pyramid pooling layer and attention path aggregation network to build an attention feature fusion network; Based on the YOLO-Head module of the YOLOv4 model, a feature recognition module is constructed, and the FocalLoss loss function is integrated into the feature recognition module; The feature extraction module, the feature fusion module, the input module, the attention feature fusion network and the feature recognition module are connected in sequence to form the fire-resistant window quality detection model.

9. A method for detecting the production quality of fire-resistant windows according to claim 1, characterized in that: The spectrum and acoustic wave detection signals of each fire-resistant window product are collected in real time on the fire-resistant window quality inspection line, and input into the fire-resistant window quality inspection model for inspection to obtain the quality assessment results of each fire-resistant window product, including: Based on the spectrum transceiver and the sound wave transceiver, a non-destructive fire-resistant window quality inspection line is built; the preset position of the fire-resistant window quality inspection line is equipped with a spectrum transceiver and a sound wave transceiver that are relatively arranged; the spectrum transceiver and the sound wave transceiver work simultaneously; Through the spectrum transceiver, a real-time spectrum diagram of the fire-resistant window product to be inspected is obtained; through the sound wave transceiver, a real-time sound wave detection signal of the fire-resistant window product to be inspected is obtained; The real-time spectrum graph and the real-time acoustic wave detection signal are input into the fire-resistant window quality detection model to obtain an output quality assessment result; wherein the quality assessment result at least includes: fire resistance prediction data and fire resistance prediction score of each key detection point in the current fire-resistant window product to be inspected; According to the quality assessment result, the current fire-resistant window product to be inspected is transferred to the corresponding assembly line branch, and the quality inspection of the next fire-resistant window product to be inspected is carried out; the assembly line branches at least include a quality inspection qualified branch and a quality inspection unqualified branch.

10. A fire-resistant window production quality detection device, characterized in that: The device comprises: A model training module is used to collect spectrograms and acoustic wave detection signals of several fire-resistant window samples, and perform combustion tests to obtain fire resistance test results; extract spectral features of the spectrogram to obtain a first quality feature; extract signal features of the acoustic wave detection signal to obtain a second quality feature; fuse the first quality feature and the second quality feature of the same fire-resistant window sample to obtain a corresponding sample quality feature; determine a model training data set based on the sample quality feature and the corresponding fire resistance test results; construct a fire-resistant window quality detection model, and train the fire-resistant window quality detection model through the model training data set; The quality assessment module is used to collect the spectrum and acoustic wave detection signals of each fire-resistant window product in real time on the constructed fire-resistant window quality inspection line, and extract the corresponding sample quality characteristics, input them into the fire-resistant window quality inspection model for inspection, and obtain the quality assessment results of each fire-resistant window product.

Images