CN118406394A - Preparation method of fireproof flame-retardant material - Google Patents

Abstract

The present invention discloses a method for preparing a fireproof and flame retardant material. The method firstly prepares an inorganic coating by a sol-gel method, then attaches the inorganic coating to the surface of a substrate to form a fireproof and flame retardant material, and then performs a quality inspection on the fireproof and flame retardant material. In this way, the quality inspection of the fireproof and flame retardant material can be achieved.

Description

A method for preparing fire retardant material

Technical Field

The present application relates to the field of flame retardant materials, and more specifically, to a method for preparing a fireproof and flame retardant material.

Background technique

With the development of society and people's increasing requirements for safety performance, the demand for fire-retardant materials in the fields of construction, transportation, electronic equipment, etc. is growing. These fire-retardant materials can effectively prevent the occurrence and spread of fires and protect people's lives and property safety.

The preparation of fire-retardant materials involves a variety of chemical and physical processes. A common method is to add flame retardants, which are compounds that can produce chemical reactions at high temperatures, thereby inhibiting the growth of flames or preventing their spread. For example, halogen compounds are a widely used class of flame retardants that release halogen atoms during combustion. These atoms can capture free radicals required for combustion, thereby slowing the spread of flames. In addition to chemical additives, physical methods are also used to prepare flame-retardant materials, such as providing protection through surface treatment or coating of materials.

However, there are often some problems with traditional fire-retardant materials and their preparation processes. For example, some materials on the market have poor flame retardant effects and cannot effectively prevent the occurrence and spread of fires. Some other materials have good flame retardant effects, but their preparation processes are complicated and costly, and are not suitable for large-scale production. Therefore, an optimized method for preparing fire-retardant materials is desired.

Summary of the invention

In view of this, the present application proposes a method for preparing a fire-retardant material.

According to one aspect of the present application, a method for preparing a fire-proof and flame-retardant material is provided, which comprises: preparing an inorganic coating by a sol-gel method; attaching the inorganic coating to the surface of a substrate to form a fire-proof and flame-retardant material; and performing quality inspection on the fire-proof and flame-retardant material.

In the above-mentioned method for preparing the fire-retardant material, the inorganic coating is attached to the surface of the substrate to form the fire-retardant material, including: obtaining an attachment state image captured by a camera; performing image feature extraction and image reconstruction on the attachment state image to obtain a reconstructed attachment state image; calculating a pixel-by-pixel difference image between the attachment state image and the reconstructed attachment state image; and determining the quality inspection result of the attachment based on the pixel-by-pixel difference image.

In the above-mentioned method for preparing fire-retardant and flame-retardant materials, image feature extraction and image reconstruction are performed on the attachment state image to obtain a reconstructed attachment state image, including: feature encoding the attachment state image to obtain an attachment state image coding feature map; feature visualization of the attachment state image coding feature map to obtain a significant attachment state image coding feature map; optimizing the significant attachment state image coding feature map to obtain an optimized significant attachment state image coding feature map; and passing the optimized significant attachment state image coding feature map through an image reconstructor based on a decoder part to obtain the reconstructed attachment state image.

In the above-mentioned method for preparing fire-retardant materials, feature encoding is performed on the attachment state image to obtain an attachment state image encoding feature map, including: brightness compensation is performed on the attachment state image to obtain an attachment state image after brightness compensation; and feature extraction is performed on the attachment state image after brightness compensation using a deep learning network model to obtain the attachment state image encoding feature map.

In the above-mentioned method for preparing the fire-retardant material, performing brightness compensation on the attachment state image to obtain the brightness compensated attachment state image comprises: processing the attachment state image according to the following brightness compensation formula to obtain the brightness compensated attachment state image; wherein the brightness compensation formula is: ;in, is the pixel value at each position of the image after the attachment state image is converted from the RGB space to the HSV space, A, B, C and D are adjustment parameters with different values, It is the pixel value at each position of the attached state image after brightness compensation.

In the above-mentioned method for preparing fire-retardant materials, a deep learning network model is used to extract features of the attachment state image after brightness compensation to obtain the attachment state image coding feature map, including: passing the attachment state image after brightness compensation through an encoder part based on a convolutional neural network model to obtain the attachment state image coding feature map.

In the above-mentioned method for preparing fire-retardant materials, the attachment state image coding feature map is feature visualized to obtain a significant attachment state image coding feature map, including: passing the attachment state image coding feature map through a feature saliency device based on a convolution kernel attention mechanism to obtain the significant attachment state image coding feature map.

In the above-mentioned method for preparing fire-retardant materials, the attachment state image coding feature map is passed through a feature saliency device based on a convolution kernel attention mechanism to obtain the salient attachment state image coding feature map, including: extracting the multi-scale nonlinear convolution global representation features of the attachment state image coding feature map to obtain a first nonlinear convolution feature map, a second nonlinear convolution feature map, a third nonlinear convolution feature map, a fourth nonlinear convolution feature map and a concentrated global representation feature vector; and extracting the vector elements of the channels corresponding to the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map in the concentrated global representation feature vector as weights for attention application to obtain the salient attachment state image coding feature map.

In the above-mentioned method for preparing fire-retardant material, extracting the multi-scale nonlinear convolution global characterization features of the attachment state image coding feature map to obtain a first nonlinear convolution feature map, a second nonlinear convolution feature map, a third nonlinear convolution feature map, a fourth nonlinear convolution feature map and a concentrated global characterization feature vector, including: performing a nonlinear convolution operation on the attachment state image coding feature map through a multi-scale convolution kernel group to obtain the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map; processing the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map with the following global representation formula to obtain a global characterization feature vector; wherein the global representation formula is: ;in, , , and are respectively the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map, is a multi-scale nonlinear convolution fusion feature map, Indicates cascade processing, represents the first eigenvalues, Represents the multi-scale nonlinear convolution fusion feature map The feature matrix is processed by global mean pooling. and They are respectively the first The height and width of the feature matrix, is the first In the feature matrix The feature value at the position; and activating the global characterization feature vector to obtain the concentrated global characterization feature vector.

In the above-mentioned method for preparing fire-retardant materials, the attached quality inspection result is determined based on the pixel-by-pixel differential image, including: passing the pixel-by-pixel differential image through a classifier-based quality inspector to obtain the quality inspection result, and the quality inspection result is used to indicate whether it is qualified.

In the present application, an inorganic coating is first prepared by a sol-gel method, then the inorganic coating is attached to the surface of a substrate to form a fire-retardant material, and then the fire-retardant material is quality inspected. In this way, the quality inspection of the fire-retardant material can be achieved.

Other features and aspects of the present application will become apparent from the following detailed description of the present application with reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the present application and, together with the description, serve to explain the principles of the present application.

FIG1 shows a flow chart of a method for preparing a fire-proof and flame-retardant material according to an embodiment of the present application.

FIG. 2 shows a flow chart of sub-step S120 of the method for preparing a fire-proof and flame-retardant material according to an embodiment of the present application.

FIG3 shows a flow chart of sub-step S122 of the method for preparing a fire-proof and flame-retardant material according to an embodiment of the present application.

FIG. 4 shows a flowchart of sub-step S1222 of the method for preparing a fire-proof and flame-retardant material according to an embodiment of the present application.

FIG5 shows a block diagram of a system for preparing a fire-proof and flame-retardant material according to an embodiment of the present application.

FIG. 6 shows an application scenario diagram of a method for preparing a fire-proof and flame-retardant material according to an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and/or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprises" and "includes" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

In addition, in order to better illustrate the present application, numerous specific details are provided in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

Considering that the preparation of inorganic coatings by the sol-gel method can form a thermally stable protective layer, this protective layer is not easy to decompose at high temperatures and can effectively prevent flames from contacting the surface of the material. Specifically, the sol-gel method is a method for preparing inorganic materials by converting a sol (colloidal solution) into a gel (semi-solid). In this process, a precursor of a metal or metal oxide is dissolved in a solvent to form a sol. Then, by adding a catalyst or changing the pH value of the solution, the precursors in the sol undergo hydrolysis and polycondensation reactions to form a gel network. Since these inorganic coatings are usually composed of highly crystalline inorganic materials, they have excellent thermal stability after being prepared by the sol-gel method, are not easy to decompose at high temperatures, and can prevent flames from contacting the surface of the substrate.

Based on this, the present application provides a method for preparing a fireproof and flame-retardant material. FIG1 shows a flow chart of the method for preparing a fireproof and flame-retardant material according to an embodiment of the present application. The specific steps, as shown in FIG1, include: S110, preparing an inorganic coating by a sol-gel method; S120, attaching the inorganic coating to the surface of a substrate to form a fireproof and flame-retardant material; and, S130, performing a quality inspection on the fireproof and flame-retardant material. In the embodiment of the present application, the inorganic coating includes but is not limited to aluminum oxide (Al ₂ O ₃ ), zirconium oxide (ZrO ₂ ), silicate (SiO ₂ ) and titanate (TiO ₂ ).

Among them, in the actual preparation process of the fire-retardant material, more specifically in step S120, it is usually necessary to perform quality inspection on the adhesion effect. The reason is that the adhesion effect directly affects the performance and stability of the fire-retardant material, which is of great significance for ensuring the fire resistance and service life of the material. Specifically, the adhesion effect directly affects the bonding strength between the inorganic coating and the substrate. If the adhesion effect is not good, the coating is easy to peel off or fall off, which will reduce the flame retardant properties of the fire-retardant material and make it unable to effectively resist flames and high temperatures. At present, the commonly used quality inspection method is to manually use a grid instrument to scratch the junction of the coating and the substrate, and then evaluate the adhesion strength by observing the peeling of the coating under the scratch. This method often relies on manual operation and subjective judgment, and the results may be affected by the operator's technical level and subjective factors, and lack objectivity. At the same time, this destructive test may cause damage to the sample during the measurement process, and subsequent analysis or use cannot be performed. Therefore, an optimized solution is expected.

In view of the above technical problems, the technical concept of the present application is as follows: based on an unsupervised anomaly detection mechanism, feature extraction is performed on the attachment state image in combination with image processing technology, the attachment state feature information obtained through feature extraction is processed through image reconstruction based on a deep learning algorithm to obtain a reconstructed attachment state image, the reconstructed attachment state image and the actual attachment state image are differentiated pixel by pixel, and the quality inspection of the attachment effect is intelligently realized based on the degree of difference between the two. In particular, in the process of image reconstruction, since a large number of attachment state images marked as qualified are used for image reconstruction in the training stage, when the attachment state image actually detected contains unqualified products, the reconstructed image will lose a large amount of abnormal information, resulting in a large difference between the reconstructed attachment state image and the actual attachment state image, thereby accurately judging the quality inspection results. That is, this unsupervised anomaly detection mechanism can optimize the quality inspection process of the attachment effect.

Based on this, as shown in FIG. 2 , the inorganic coating is attached to the surface of the substrate to form a fire-retardant material, including: S121, acquiring an attachment state image captured by a camera; S122, performing image feature extraction and image reconstruction on the attachment state image to obtain a reconstructed attachment state image; S123, calculating a pixel-by-pixel difference image between the attachment state image and the reconstructed attachment state image; and, S124, determining an attachment quality inspection result based on the pixel-by-pixel difference image.

It should be understood that in step S121, a camera is used to capture an image of the inorganic coating attached to the surface of the substrate, and the image shows the adhesion state of the coating, including peeling, blistering, cracking or other defects; in step S122, features representing the adhesion state of the coating, such as color, texture and shape, are extracted from the original image, and the extracted features are reconstructed into a new image using machine learning or computer vision technology, which is called a reconstructed adhesion state image; in step S123, a pixel-by-pixel difference image highlights the differences between the original image and the reconstructed image, which may indicate coating adhesion defects; in step S124, the difference image is analyzed to identify coating adhesion defects, and the coating adhesion quality inspection result is determined based on the severity and number of defects, and the quality inspection result may be qualified, unqualified or requires rework.

Specifically, in the technical solution of the present application, the specific encoding process of attaching the inorganic coating to the surface of the substrate to form a fire-retardant material includes: first, obtaining an attachment state image captured by a camera. Here, using a camera to capture an attachment state image is a non-destructive detection method that will not cause damage to the detected object. At the same time, the image captured by the camera can provide objective data and is an intuitive information carrier to display and reflect the state of attachment, which is convenient for further processing and analysis of subsequent models. Then, the attachment state image is brightness compensated to obtain a brightness compensated attachment state image. That is, the contrast of the attachment state image is enhanced by the processing means of brightness compensation, so that the details in the attachment state image are more clearly visible.

Next, the attachment state image after brightness compensation is passed through the encoder part based on the convolutional neural network model to obtain the encoding feature map of the attachment state image. Among them, the convolutional neural network (CNN) is a deep learning model specifically used to process data with a similar grid structure, such as images and two-dimensional matrix data. The core idea of CNN is to extract abstract features in the image layer by layer through convolutional layers and pooling layers. That is, in the technical solution of the present application, the convolutional neural network model is used to capture the implicit neighborhood association features and attachment state representation in the attachment state image after brightness compensation.

Accordingly, as shown in FIG3 , image feature extraction and image reconstruction are performed on the attachment state image to obtain a reconstructed attachment state image, including: S1221, feature encoding the attachment state image to obtain an attachment state image coding feature map; S1222, feature visualization of the attachment state image coding feature map to obtain a significant attachment state image coding feature map; S1223, optimizing the significant attachment state image coding feature map to obtain an optimized significant attachment state image coding feature map; and, S1224, passing the optimized significant attachment state image coding feature map through an image reconstructor based on a decoder part to obtain the reconstructed attachment state image.

Among them, in step S1221, feature encoding is performed on the attachment state image to obtain an attachment state image encoding feature map, including: brightness compensation is performed on the attachment state image to obtain a brightness compensated attachment state image; and feature extraction is performed on the brightness compensated attachment state image using a deep learning network model to obtain the attachment state image encoding feature map.

In a specific example, performing brightness compensation on the attachment state image to obtain the brightness compensated attachment state image includes: processing the attachment state image using the following brightness compensation formula to obtain the brightness compensated attachment state image; wherein the brightness compensation formula is: ;in, is the pixel value at each position of the image after the attachment state image is converted from the RGB space to the HSV space, A, B, C and D are adjustment parameters with different values, It is the pixel value at each position of the attached state image after brightness compensation.

In a specific example, a deep learning network model is used to extract features of the attachment state image after brightness compensation to obtain a coding feature map of the attachment state image, including: passing the attachment state image after brightness compensation through an encoder part based on a convolutional neural network model to obtain the coding feature map of the attachment state image.

It is worth mentioning that a convolutional neural network (CNN) is a deep learning model that is specifically designed to process data with a grid-like structure, such as images. Convolutional neural networks use special layers called convolutional layers to extract features from images. Convolutional layers contain a set of learnable filters that slide over the image to detect specific patterns and features. Among them, the convolution operation involves element-wise multiplication of the filter with a portion of the image and then summing the results. The filter slides over the image and performs a convolution operation at each position. The result of the convolution operation is a new feature map, where each pixel value represents the feature intensity of the corresponding area in the original image. By stacking multiple convolutional layers, convolutional neural networks can extract increasingly complex features in images, where early layers detect low-level features such as edges and textures, while later layers detect high-level features such as objects and faces. Convolutional neural networks can automatically learn features from data without manual feature engineering; convolutional neural networks are invariant to translation and rotation in images, which means that features can be detected regardless of their position or orientation in the image.

Furthermore, the feature map of the attached state image coding is passed through a feature saliency device based on a convolution kernel attention mechanism to obtain a salient attached state image coding feature map. The feature saliency device based on a convolution kernel attention mechanism performs a nonlinear convolution operation on the attached state image coding feature map through a multi-scale convolution kernel group to extract the associated features in different local spatial neighborhoods in the attached state image coding feature map, thereby improving the expressiveness and distinguishability of the features. Afterwards, the importance of both the channel and the convolution kernel is considered at the same time, and the weight information is used to guide the network to assign different attentions to convolution kernels of different sizes. In this way, the feature distribution captured by the convolution kernel of useful features is extracted, and the target information of the convolution kernel of useless background features or noise is ignored.

Among them, in step S1222, the attachment state image coding feature map is feature-manifested to obtain a salient attachment state image coding feature map, including: passing the attachment state image coding feature map through a feature salientator based on a convolution kernel attention mechanism to obtain the salient attachment state image coding feature map.

It is worth mentioning that the convolution kernel attention mechanism is a technique for highlighting important convolution kernels in convolutional neural networks (CNNs). The convolution kernel attention mechanism assigns a weight to each convolution kernel, which represents the importance of the convolution kernel in feature extraction. The weight is calculated by calculating the activation degree of the convolution kernel output feature map, and then the weight is multiplied by the convolution kernel output feature map to highlight important features while suppressing unimportant features. The convolution kernel attention mechanism is used to: identify the most important features and convolution kernels for model prediction; select the most important features for a specific task; and reduce the size of the model by removing unimportant convolution kernels. It should be understood that by focusing on important features, the convolution kernel attention mechanism can improve the accuracy and robustness of the model; the convolution kernel attention mechanism allows visualization of how the model focuses on different regions and features in the image; and by removing unimportant convolution kernels, the convolution kernel attention mechanism can reduce the computational cost of the model. When performing feature visualization on the feature map encoded by the adhesion state image, the convolution kernel attention mechanism can identify features related to coating adhesion defects, such as peeling, blistering, or cracking. By highlighting these features, the convolutional neural network model can detect and classify attachment defects more accurately.

More specifically, as shown in Figure 4, in step S1222, the attachment state image coding feature map is passed through a feature saliency device based on a convolution kernel attention mechanism to obtain the salient attachment state image coding feature map, including: S12221, extracting the multi-scale nonlinear convolution global representation features of the attachment state image coding feature map to obtain a first nonlinear convolution feature map, a second nonlinear convolution feature map, a third nonlinear convolution feature map, a fourth nonlinear convolution feature map and a concentrated global representation feature vector; and, S12222, extracting the vector elements of the channels corresponding to the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map in the concentrated global representation feature vector as weights for attention application to obtain the salient attachment state image coding feature map.

In a specific example, in step S12221, extracting the multi-scale nonlinear convolution global representation features of the attachment state image coding feature map to obtain a first nonlinear convolution feature map, a second nonlinear convolution feature map, a third nonlinear convolution feature map, a fourth nonlinear convolution feature map and a concentrated global representation feature vector, including: performing a nonlinear convolution operation on the attachment state image coding feature map through a multi-scale convolution kernel group to obtain the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map; processing the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map according to the following global representation formula to obtain a global representation feature vector; wherein the global representation formula is: ;in, , , and are respectively the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map, is a multi-scale nonlinear convolution fusion feature map, Indicates cascade processing, represents the first eigenvalues, Represents the multi-scale nonlinear convolution fusion feature map The feature matrix is processed by global mean pooling. and They are respectively the first The height and width of the feature matrix, is the first In the feature matrix The feature value at the position; and activating the global characterization feature vector to obtain the concentrated global characterization feature vector.

Subsequently, the salient attachment state image coding feature map is passed through an image reconstructor based on a decoder part to obtain a reconstructed attachment state image. Here, the salient attachment state image coding feature map after feature extraction and local attention information enhancement can be converted back into an image form through the image reconstructor based on the decoder part to achieve reconstruction of the original image. And further calculate the pixel-by-pixel difference image between the attachment state image and the reconstructed attachment state image to quantify the degree of pixel difference between the attachment state image and the reconstructed attachment state image. It should be understood that since the decoder does not have the ability to reconstruct abnormal images, if the attachment state image contains unqualified products, the reconstructed attachment state image obtained after reconstruction will lose the information of the abnormal part. Therefore, the pixel-by-pixel difference image describes and depicts the difference information between the attachment state image and the product marked as qualified to a certain extent. Subsequently, the pixel-by-pixel difference image is passed through a quality inspector based on a classifier to obtain a quality inspection result, and the quality inspection result is used to indicate whether it is qualified.

Accordingly, in step S124, the attached quality inspection result is determined based on the pixel-by-pixel differential image, including: passing the pixel-by-pixel differential image through a classifier-based quality inspector to obtain the quality inspection result, and the quality inspection result is used to indicate whether it is qualified.

Specifically, the pixel-by-pixel difference image is passed through a classifier-based quality inspector to obtain the quality inspection result, and the quality inspection result is used to indicate whether it is qualified, including: expanding the pixel-by-pixel difference image into a classification feature vector according to a row vector or a column vector; using the fully connected layer of the classifier-based quality inspector to fully connect encode the classification feature vector to obtain an encoded classification feature vector; and inputting the encoded classification feature vector into the Softmax classification function of the classifier-based quality inspector to obtain the quality inspection result.

It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression, SVM, etc. are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification, but this is prone to errors and inefficient. Commonly used multi-classification methods include the Softmax classification function.

In the above technical scheme, the attachment state image coding feature map expresses the image semantic features of the attachment state image after brightness compensation, and after the attachment state image coding feature map passes through the feature saliency device based on the convolution kernel attention mechanism, the image semantic feature representation of the significant attachment state image coding feature map obtained will further strengthen the spatial distribution of local image semantic features on the basis of the attachment state image coding feature map. However, while this improves the significance of the local image semantic feature representation, it will also lead to the enhancement of the discreteness of the overall feature distribution of the image semantic feature representation of the significant attachment state image coding feature map.

In this way, when the significant attachment state image coding feature map is passed through an image reconstructor based on the decoder part to obtain a reconstructed attachment state image, the significant attachment state image coding feature map will have a multi-sub-dimensional distribution feature representation dimensional complexity based on multiple discrete local sub-dimensions due to the enhanced discreteness of the overall feature distribution of the significant attachment state image coding feature map relative to the semantic feature distribution of the source image. Therefore, it is also expected to improve the overall feature expression effect under the complex feature representation dimension during the model iteration process.

Based on this, in each iteration of the model, the present application encodes the characteristic map of the salient attachment state image. Optimization is performed, specifically comprising: encoding the characteristic map of the salient attachment state image Each eigenvalue of the salient attachment state image encoding feature map The reciprocal of the maximum eigenvalue of Perform dot multiplication, and then add the dot multiplication result to the salient attachment state image encoding feature map The ratio of the eigenvalue mean and the eigenvalue standard deviation, that is, Perform point subtraction and point addition, take the absolute value of the point subtraction result and then perform logarithm calculation with base 2, and then perform further point addition calculation with the point addition result and the dot product result of the weight hyperparameter.

That is, by including the mean and standard value The interactive representation of the probability statistical features represented by and The short sequence of distribution interaction representations is used as the salient attachment state image encoding feature map The submanifold potential motif under the complex manifold network is used to encode the characteristic map of the salient attachment state image. The latent motif feature information pattern and feature distribution pattern of the global structure inference unit are used to reconstruct the salient attachment state image encoding feature map in the form of a global structure latent motif dictionary based on the connection. The complex manifold structure of the model is used to improve the model's ability to understand the generation and evolution of the manifold structure corresponding to the associated features of the complex feature representation dimension during the iteration process, and to improve the feature expression effect of the complex feature representation dimension based on multi-dimensional distribution during the model iteration process, thereby improving the image quality of the decoded image and improving the accuracy of the quality inspection results.

Accordingly, the significant attachment state image coding feature map is optimized to obtain an optimized significant attachment state image coding feature map, including: performing a dot multiplication of each eigenvalue of the significant attachment state image coding feature map and the inverse of the maximum eigenvalue of the significant attachment state image coding feature map by the significant attachment state image coding image semantic interaction representation map; dividing the eigenvalue mean of the significant attachment state image coding feature map by the eigenvalue standard deviation of the significant attachment state image coding feature map to obtain a statistical interaction value corresponding to the significant attachment state image coding feature map; and Each eigenvalue of the image semantic interaction representation map is subtracted from the statistical interaction value and then the absolute value is taken, and then a logarithm calculation with base 2 is performed to obtain a significant attachment state image encoding image semantic interaction information representation map; each eigenvalue of the significant attachment state image encoding image semantic interaction representation map is added to the statistical interaction value and then multiplied by a predetermined weight hyperparameter to obtain a significant attachment state image encoding image semantic interaction pattern representation map; the significant attachment state image encoding image semantic interaction information representation map and the significant attachment state image encoding image semantic interaction pattern representation map are added to obtain an optimized significant attachment state image encoding feature map.

In summary, based on the method for preparing the fire-proof and flame-retardant material according to the embodiment of the present application, it is possible to realize the detection-free fire-proof and flame-retardant material.

FIG5 shows a block diagram of a system 100 for preparing a fireproof and flame retardant material according to an embodiment of the present application. As shown in FIG5, the system 100 for preparing a fireproof and flame retardant material according to an embodiment of the present application comprises: an inorganic coating preparation module 110 for preparing an inorganic coating by a sol-gel method; an attachment module 120 for attaching the inorganic coating to the surface of a substrate to form a fireproof and flame retardant material; and a quality inspection module 130 for performing quality inspection on the fireproof and flame retardant material.

Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned fire-proof and flame-retardant material preparation system 100 have been described in detail in the description of the fire-proof and flame-retardant material preparation method with reference to Figures 1 to 4 above, and therefore, its repeated description will be omitted.

As described above, the system 100 for preparing fire-retardant materials according to the embodiment of the present application can be implemented in various wireless terminals, such as a server having a preparation algorithm for fire-retardant materials. In a possible implementation, the system 100 for preparing fire-retardant materials according to the embodiment of the present application can be integrated into a wireless terminal as a software module and/or a hardware module. For example, the system 100 for preparing fire-retardant materials can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the system 100 for preparing fire-retardant materials can also be one of the many hardware modules of the wireless terminal.

Alternatively, in another example, the fire retardant material preparation system 100 and the wireless terminal may also be separate devices, and the fire retardant material preparation system 100 may be connected to the wireless terminal via a wired and/or wireless network and transmit interactive information in accordance with an agreed data format.

Figure 6 shows an application scenario diagram of the method for preparing a fire-retardant material according to an embodiment of the present application. As shown in Figure 6, in this application scenario, first, an attachment state image captured by a camera (for example, D shown in Figure 6) is obtained, and then the attachment state image is input into a server (for example, S shown in Figure 6) in which a preparation algorithm for fire-retardant materials is deployed, wherein the server can process the attachment state image using the preparation algorithm for fire-retardant materials to obtain a quality inspection result indicating whether the material is qualified.

In an exemplary embodiment, there is also provided a non-volatile computer-readable storage medium, such as a memory including computer program instructions, which can be executed by a processing component of an apparatus to perform the above method.

The present application may be a system, a method and/or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present application.

Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the function involved. It should also be noted that each square box in the block diagram and/or flow chart, and the combination of the square boxes in the block diagram and/or flow chart can be realized by a dedicated hardware-based system that performs the specified function or action, or can be realized by a combination of special-purpose hardware and computer instructions.

The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims (10)

1. A method of preparing a fire-resistant material, comprising: preparing an inorganic coating by a sol-gel method; attaching the inorganic coating to a surface of a substrate to form a fire-retardant material; and performing quality inspection on the fireproof flame-retardant material.

2. The method of preparing a fire resistant material according to claim 1, wherein attaching the inorganic coating to a surface of a substrate to form the fire resistant material comprises: acquiring an attachment state image acquired by a camera; extracting image features and reconstructing images of the attached state images to obtain reconstructed attached state images; calculating a pixel-by-pixel differential image between the attachment state image and the reconstructed attachment state image; and determining an attached quality inspection result based on the pixel-by-pixel differential image.

3. The method of producing a fire-retardant material according to claim 2, wherein performing image feature extraction and image reconstruction on the adhesion state image to obtain a reconstructed adhesion state image comprises: performing feature coding on the attached state image to obtain an attached state image coding feature map; performing feature visualization on the attached state image coding feature map to obtain a visualized attached state image coding feature map; optimizing the salified attached state image coding feature map to obtain an optimized salified attached state image coding feature map; and passing the optimized salified attached state image coding feature map through an image reconstructor based on a decoder portion to obtain the reconstructed attached state image.

4. A method of producing a fire resistant material according to claim 3, wherein feature encoding the adhesion state image to obtain an adhesion state image encoded feature map comprises: performing brightness compensation on the attached state image to obtain a brightness-compensated attached state image; and extracting the characteristics of the brightness compensated attached state image by using a deep learning network model to obtain the attached state image coding characteristic map.

5. The method of producing a fire-retardant material according to claim 4, wherein the brightness-compensating the adhesion-state image to obtain a brightness-compensated adhesion-state image comprises: processing the attached state image with the following brightness compensation formula to obtain an attached state image after brightness compensation; wherein, the brightness compensation formula is: ; wherein, For the pixel values of each position of the image after the attachment state image is converted from RGB space to HSV space, A, B, C and D are the adjusting parameters with different values,And pixel values of the positions of the attached state image after the brightness compensation are obtained.

6. The method for preparing a fire-retardant material according to claim 5, wherein the feature extraction of the brightness-compensated adhesion state image by using a deep learning network model to obtain the adhesion state image coding feature map comprises: and passing the brightness compensated attachment state image through an encoder part based on a convolutional neural network model to obtain the attachment state image coding feature map.

7. The method of producing a fire-retardant material according to claim 6, wherein characterizing the adhesion state image coding feature map to obtain a pronounced adhesion state image coding feature map comprises: the attached state image coding feature map is passed through a feature salizer based on a convolution kernel attention mechanism to obtain the salified attached state image coding feature map.

8. The method of producing a fire resistant material according to claim 7, wherein passing the adhesion state image coding feature map through a feature salizer based on a convolution kernel attention mechanism to obtain the salified adhesion state image coding feature map comprises: extracting multi-scale nonlinear convolution global characterization features of the attached state image coding feature map to obtain a first nonlinear convolution feature map, a second nonlinear convolution feature map, a third nonlinear convolution feature map, a fourth nonlinear convolution feature map and a concentrated global characterization feature vector; and extracting vector elements of channels corresponding to the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map in the condensed global characterization feature vector as weights to apply attention so as to obtain the salient attachment state image coding feature map.

9. The method of preparing a fire-resistant material of claim 8, wherein extracting the multi-scale nonlinear convolution global characterization feature of the adhesion state image encoding feature map to obtain a first nonlinear convolution feature map, a second nonlinear convolution feature map, a third nonlinear convolution feature map, a fourth nonlinear convolution feature map, and a condensed global characterization feature vector comprises: performing nonlinear convolution operation on the attached state image coding feature map through a multi-scale convolution kernel group to obtain the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map; processing the first nonlinear convolution feature map, the second nonlinear convolution feature map, the third nonlinear convolution feature map and the fourth nonlinear convolution feature map with the following global expression formula to obtain a global characterization feature vector; wherein the global expression formula is: ; wherein, 、、AndThe first nonlinear convolution characteristic map, the second nonlinear convolution characteristic map, the third nonlinear convolution characteristic map, and the fourth nonlinear convolution characteristic map,The feature map is fused for multi-scale nonlinear convolution,A cascade of processes is represented which is a cascade of processes,Representing the first of the global characterization feature vectorsThe value of the characteristic is a value of,Representing the third dimension in the multi-scale nonlinear convolution fusion characteristic diagramThe feature matrices are subjected to global averaging treatment,AndRespectively the first of the multi-scale nonlinear convolution fusion characteristic diagramsThe height and width of the individual feature matrices,Fusing the third feature map for the multi-scale nonlinear convolutionIn a characteristic matrixA feature value at the location; and activating the global characterization feature vector to obtain the concentrated global characterization feature vector.

10. The method of preparing a fire resistant material according to claim 9, wherein determining an attached quality inspection result based on the pixel-by-pixel differential image comprises: and passing the pixel-by-pixel differential image through a classifier-based quality detector to obtain the quality detection result, wherein the quality detection result is used for indicating whether the quality detection result is qualified or not.