Ultra-high performance concrete slab anti-explosion response prediction method and system based on machine learning

Abstract

The invention provides an ultra-high performance concrete slab anti-explosion response prediction method and system based on machine learning, and relates to the technical blasting effect prediction field, and the method comprises the steps: obtaining explosion experiment data, and analyzing the explosion experiment data to generate explosion response data; constructing a blasting data adversarial network, training by using the blasting experiment data and blasting response data adversarial network, and generating a virtual blasting experiment data set by using the trained blasting data adversarial network; constructing a blasting response prediction model, and training the model by using a virtual blasting experiment data set; predicting blasting response data by using the blasting response prediction model; and performing comprehensive analysis on the pre-blasting data and the blasting response data to obtain a comprehensive blasting response coefficient. Through virtual data, the diversity and coverage range of data samples are enhanced, and the prediction precision of the model is improved. And a comprehensive index of anti-explosion response is constructed, so that the anti-explosion performance can be reflected more comprehensively.

Description

Technical Field

[0001] The present invention relates to the technical field of blasting effect prediction, and in particular to a method and system for predicting the anti-blast response of an ultra-high performance concrete slab based on machine learning. Background Art

[0002] Ultra-high performance concrete is increasingly used in defense engineering and safety protection fields due to its excellent strength, toughness and impact resistance. However, when faced with explosive impact loads, the explosion resistance of ultra-high performance concrete panels is complex, and its dynamic response involves a variety of physical, mechanical and damage mechanisms within the material. At present, traditional research on explosion resistance mostly relies on experimental testing and finite element simulation technology, but experimental testing is expensive, has a long cycle, and is difficult to fully cover multi-dimensional parameter combinations; although finite element simulation can partially make up for the limitations of experiments, its accuracy and reliability are strongly dependent on material models and parameter settings, and there are certain limitations.

[0003] With the development of machine learning technology, it has become a trend to apply it to the prediction of material dynamic response. However, there is currently no systematic method that can combine multi-dimensional experimental data with generative adversarial networks to achieve the construction of a high-precision ultra-high performance concrete slab explosion-resistant response prediction model to fill the problem of insufficient experimental data.

[0004] In the prior art, the publication number CN117610407A discloses a method and device for predicting the impact response of a steel-concrete composite slab based on machine learning, by constructing a first data set according to factors affecting the impact response; filling the first data set; training an improved Gaussian process regression algorithm model with the first data set; and calculating the accuracy and visualizing the results of the improved Gaussian process regression algorithm model. The prior art predicts the impact response capability of steel-concrete composite slabs of different structures in multiple scenarios by adding the features of objects that may fall in the city and the features of corrugated steel structures, thereby reducing the limitations of the range and increasing the upper limit of the impact of the steel-concrete composite slab; by using Bayesian optimization to optimize the hyperparameters of the Gaussian process regression model used for prediction, the hyperparameter combination with the minimum loss function is found to improve the accuracy of the model and reduce the calculation time; however, the prior art still has defects. In the prior art, due to the high cost of the impact response experiment, it is difficult to obtain a sufficient amount of training data, it is difficult to ensure the accuracy of the results, and reliance on raw data may limit the generalization ability of the training model; and the prior art only analyzes the existing responses, lacking a comprehensive analysis of the responses.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for predicting the explosion resistance of ultra-high performance concrete slabs based on machine learning to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for predicting explosion resistance response of ultra-high performance concrete slabs based on machine learning, the specific steps include:

[0009] Step 1: performing N explosion tests on the ultra-high performance concrete slab to obtain explosion test data, wherein the explosion test data includes explosion data, data before explosion, data after explosion, and an image of the ultra-high performance concrete slab after explosion;

[0010] Step 2: Analyze the image of the ultra-high performance concrete slab after blasting to obtain post-blasting image feature data, and combine the post-blasting data and the post-blasting image feature data to generate blasting response data;

[0011] Step 3: The blasting data, pre-blasting data and corresponding blasting response data of the N explosion experiments in step 1 form an explosion experiment set; construct a blasting data adversarial network, use the explosion experiment set to train the blasting data adversarial network, and use the trained blasting data adversarial network to generate a virtual explosion experiment data set;

[0012] Step 4: Using the blasting data and pre-blasting data in the virtual explosion experiment data set as input and the explosion response data as labels, the blasting response prediction model is trained, the pre-blasting data and blasting data of the current blasting work are obtained, and they are input into the blasting response model to obtain the predicted blasting response data;

[0013] Step 5: Comprehensively analyze the pre-blasting data and explosion response data corresponding to the current blasting work to obtain the property residual coefficient and the blasting damage coefficient; generate a comprehensive blasting response coefficient based on the property residual coefficient and the blasting damage coefficient.

[0014] Further, the blasting data includes the pressure and duration of the applied blasting shock wave, the distance between the blasting shock wave emission source and the plane of the ultra-high performance concrete slab, the angle between the line connecting the blasting shock wave emission source and the center position of the ultra-high performance concrete slab and the plane of the ultra-high performance concrete slab, and the pre-blasting data includes the thickness, compressive strength, tensile strength, elastic modulus, mass and density of the ultra-high performance concrete slab; the post-blasting data includes the compressive strength, elastic modulus and mass of the ultra-high performance concrete slab after blasting, and the stress distribution on the back of the ultra-high performance concrete slab;

[0015] The blast response data include images, compressive strength, elastic modulus and mass of the UHPC slab after blasting, and stress distribution on the back side of the UHPC slab, which is the side opposite to the blasting side of the UHPC slab.

[0016] Further, the explosion response data includes the crack proportion, relative ratio of maximum crack depth, deep crack proportion, stress distribution variance of the back of the slab, compressive strength, elastic modulus and mass of the ultra-high performance concrete slab after blasting;

[0017] The specific logic for calculating the crack ratio, the relative ratio of the maximum crack depth, and the deep crack ratio is as follows: the frequency of the pixel points of each level in the 256 gray levels in the image of the ultra-high performance concrete slab is counted to form a grayscale histogram, and the inter-class variance of each threshold in the histogram is calculated. The best separation threshold is selected through the inter-class variance, and the image of the front of the ultra-high performance concrete slab is threshold segmented using the best segmentation threshold to identify the crack part and the concrete part.

[0018] The inter-class variance of the corresponding threshold is calculated according to the following formula:

[0019] σ(t) 2 =w B (t)*w ( (t)*(μ B (t)-μ F (t)) 2

[0020] Among them, σ(t) 2 represents the inter-class variance when the threshold is t, t∈[0, 255], and t∈N + ,w B (t), w F (t) are the number of pixels in the crack part and the concrete part when the threshold is t, μ B (t), μ F (t) are the mean grayscale values ​​of the pixels in the background and foreground when the threshold is t;

[0021] Calculate the threshold value when the inter-class variance reaches the maximum, and define this value as the optimal separation threshold;

[0022] The number of pixels occupied by the crack part is divided by the total number of pixels to obtain the crack ratio; the average gray value of the concrete part is calculated, the pixel with the smallest gray value in the crack part is compared with the average gray value of the concrete part to obtain the maximum crack depth, a crack gray threshold is preset, and the pixel with a gray value of the crack part less than the crack gray threshold is defined as a deep pixel, the number of deep pixels is counted, and the number of deep pixels is divided by the number of pixels in the crack part to obtain the deep crack ratio; the crack gray threshold is 50% of the average gray value of the concrete part;

[0023] The specific formulas for calculating the crack ratio, the relative ratio of the maximum crack depth, the deep crack ratio, and the stress distribution variance of the back of the plate are:

[0024]

[0025] Among them, Ph is the average gray value of the concrete part, Ph i is the gray value of the i-th pixel of the concrete part, N H is the total number of pixels in the concrete part, i is the index of the pixel in the concrete part, LX is the crack ratio, N P is the total number of pixels in the crack part, N A is the total number of pixels, LPH is the maximum crack depth relative ratio, PL min is the gray value of the pixel with the smallest gray value in the crack part; PLD is the proportion of deep cracks, N D is the total number of deep pixels.

[0026] Furthermore, the blasting data adversarial network includes a generator and a discriminator, and the generator includes an input layer, an output layer and a hidden layer; the input layer of the generator is used to receive the explosion experiment set, and add random noise to the blasting data, pre-blasting data and corresponding blasting response data as initial input; the hidden layer of the generator is composed of multiple layers, each layer contains a number of nodes, and each node is connected to the previous layer through a weight; the ReLU activation function is used to perform a nonlinear transformation on the input data, so that the model can generate complex distributions; an independent neuron is set in the output layer of the generator, which is responsible for converting the transformed data of the hidden layer into virtual data generated by the generator;

[0027] The discriminator includes an input layer, an output layer and a hidden layer; the discriminator input layer receives the explosion test set and the virtual data generated by the generator; the hidden layer of the discriminator is composed of multiple layers, each layer contains a number of nodes, and each node is connected to the previous layer through a weight; the ReLU activation function is used to perform a nonlinear transformation on the input data so that the model can generate a complex distribution; an independent neuron is set in the discriminator output layer, and the independent neuron of the discriminator output layer uses a Sigmoid activation function, and the output range is [0,1] to reflect the output probability of the discriminator for real data and generated virtual data;

[0028] The formula of the loss function is:

[0029]

[0030] Among them, L is the loss function of the blasting data adversarial network, D(x) is the output probability of the discriminator for the real data; E is the mathematical expectation operation, x~p date For x, we need to get the real data p date Get a sample from the , D(G(z)) is the output probability of the discriminator for the generated virtual data; z~p z is z from a random noise distribution p z Get a sample from .

[0031] Furthermore, the specific logic for obtaining the residual coefficient of properties is as follows: the compressive strength, elastic modulus and mass of the ultra-high performance concrete slab are comprehensively analyzed to obtain the residual strength ratio and residual mass ratio;

[0032] The specific formula for calculating the residual coefficient of the property is:

[0033]

[0034] Among them, R s is the residual coefficient of the property, f o is the compressive strength of the ultra-high performance concrete slab before blasting, f p is the compressive strength of the ultra-high performance concrete slab after blasting, K p is the elastic modulus of the ultra-high performance concrete slab before blasting, K p Elastic modulus of ultra-high performance concrete slab after blasting; m o is the mass of the ultra-high performance concrete slab before blasting, m p The quality of ultra-high performance concrete slab after blasting.

[0035] Furthermore, the blasting damage coefficient is generated according to the crack proportion, the relative ratio of the maximum crack depth, the deep crack proportion and the stress distribution variance of the plate back. The specific formula for generating the blasting damage coefficient is:

[0036]

[0037] Among them, BH is the explosion damage coefficient, LX is the crack ratio; LPH is the relative ratio of the maximum crack depth, LPD is the deep crack ratio, σ B is the stress distribution variance of the back of the plate;

[0038] The comprehensive blasting response coefficient is generated according to the residual coefficient of the property and the blasting damage coefficient. The specific formula for generating the comprehensive blasting response coefficient is:

[0039] BXP=BH*R s

[0040] Among them, BXP is the comprehensive blasting response coefficient, R s is the property residual coefficient.

[0041] The present invention further provides a system for predicting the explosion resistance of ultra-high performance concrete slabs based on machine learning, and the system is used to implement the method for predicting the explosion resistance of ultra-high performance concrete slabs based on machine learning, specifically comprising:

[0042] A data acquisition module, used to perform N explosion tests on the ultra-high performance concrete slab to acquire explosion test data, wherein the explosion test data includes explosion data, data before explosion, data after explosion, and an image of the ultra-high performance concrete slab after explosion;

[0043] A response analysis module is used to analyze the image of the ultra-high performance concrete slab after blasting to obtain post-blasting image feature data, and combine the post-blasting data and the post-blasting image feature data to generate blasting response data;

[0044] The data virtualization module is used to form an explosion experiment set with the explosion data, pre-explosion data and corresponding explosion response data of the N explosion experiments in step 1; construct a blasting data adversarial network, train the blasting data adversarial network using the explosion experiment set, and generate a virtual explosion experiment data set using the trained blasting data adversarial network;

[0045] The model building module is used to take the blasting data and pre-blasting data in the virtual blasting experiment data set as input and the blasting response data as labels, train the blasting response prediction model, obtain the pre-blasting data and blasting data of the current blasting work, and input them into the blasting response model to obtain the predicted blasting response data;

[0046] The comprehensive analysis module is used to comprehensively analyze the pre-blasting data and explosion response data corresponding to the current blasting work to obtain the property residual coefficient and the blasting damage coefficient; and generate the comprehensive blasting response coefficient based on the property residual coefficient and the blasting damage coefficient.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] According to the above scheme, the present invention generates virtual experimental data through a blasting data adversarial network, which significantly enhances the diversity and coverage of data samples, thereby improving the adaptability and prediction accuracy of the model to complex scenarios.

[0049] The present invention also generates a property residual coefficient and a blasting damage coefficient by integrating a variety of physical indicators through the above scheme, and further analyzes to obtain a comprehensive blasting response coefficient, which can comprehensively reflect the explosion resistance of the ultra-high performance concrete slab and has stronger applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the overall method flow of the present invention.

[0051] Figure 2 It is a schematic diagram of the overall system structure of the present invention. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0053] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0054] Example:

[0055] See also Figure 1 , the present invention provides a technical solution:

[0056] A method for predicting explosion resistance response of ultra-high performance concrete slabs based on machine learning, the specific steps include:

[0057] Step 1: performing N explosion tests on the ultra-high performance concrete slab to obtain explosion test data, wherein the explosion test data includes explosion data, data before explosion, data after explosion, and an image of the ultra-high performance concrete slab after explosion;

[0058] Further, the blasting data includes the pressure and duration of the applied blasting shock wave, the distance between the blasting shock wave emission source and the plane of the ultra-high performance concrete slab, the angle between the line connecting the blasting shock wave emission source and the center position of the ultra-high performance concrete slab and the plane of the ultra-high performance concrete slab, and the pre-blasting data includes the thickness, compressive strength, tensile strength, elastic modulus, mass and density of the ultra-high performance concrete slab; the post-blasting data includes the compressive strength, elastic modulus and mass of the ultra-high performance concrete slab after blasting, and the stress distribution on the back of the ultra-high performance concrete slab;

[0059] The pre-blasting data is obtained by selecting one of the ultra-high performance concrete panels undergoing the explosion experiment as a sample and sending it to the laboratory for measurement; if the ultra-high performance concrete panel is not broken into multiple pieces after the blasting, the ultra-high performance concrete panel after the blasting is directly sent to the laboratory to obtain the post-blasting data; if the ultra-high performance concrete panel is broken into multiple pieces, the largest piece is selected and sent to the laboratory to obtain the post-blasting data.

[0060] Step 2: Analyze the image of the ultra-high performance concrete slab after blasting to obtain post-blasting image feature data, and combine the post-blasting data and the post-blasting image feature data to generate blasting response data;

[0061] Further, the explosion response data includes the crack proportion, relative ratio of maximum crack depth, deep crack proportion, stress distribution variance of the back of the slab, compressive strength, elastic modulus and mass of the ultra-high performance concrete slab after blasting;

[0062] The specific logic for calculating the crack ratio, the relative ratio of the maximum crack depth, and the deep crack ratio is as follows: the frequency of the pixel points of each level in the 256 gray levels in the image of the ultra-high performance concrete slab is counted to form a grayscale histogram, and the inter-class variance of each threshold in the histogram is calculated. The best separation threshold is selected through the inter-class variance, and the image of the front of the ultra-high performance concrete slab is threshold segmented using the best segmentation threshold to identify the crack part and the concrete part.

[0063] The inter-class variance of the corresponding threshold is calculated according to the following formula:

[0064] σ(t) 2 =w B (t)*w F (t)*(μ B (t)-μ F (t)) 2

[0065] Among them, σ(t) 2 represents the inter-class variance when the threshold is t, t∈[0, 255], and t∈N + ,w B(t), w F (t) are the number of pixels in the crack part and the concrete part when the threshold is t, μ B (t), μ F (t) are the mean grayscale values ​​of the pixels in the background and foreground when the threshold is t;

[0066] Calculate the threshold value when the inter-class variance reaches the maximum, and define this value as the optimal separation threshold;

[0067] The number of pixels occupied by the crack part is divided by the total number of pixels to obtain the crack ratio; the average gray value of the concrete part is calculated, the pixel with the smallest gray value in the crack part is compared with the average gray value of the concrete part to obtain the maximum crack depth, the crack gray value threshold is preset, the pixel with a gray value of the crack part less than the crack gray value threshold is defined as a deep pixel, the number of deep pixels is counted, and the number of deep pixels is divided by the number of pixels in the crack part to obtain the deep crack ratio. ; The crack gray value threshold is 50% of the average gray value of the concrete part;

[0068] The specific formulas for calculating the crack ratio, the relative ratio of the maximum crack depth, the deep crack ratio, and the stress distribution variance of the back of the plate are:

[0069]

[0070] Among them, Ph is the average gray value of the concrete part, Ph i is the gray value of the i-th pixel of the concrete part, N H is the total number of pixels in the concrete part, i is the index of the pixel in the concrete part, LX is the crack ratio, N P is the total number of pixels in the crack part, N A is the total number of pixels, LPH is the maximum crack depth relative ratio, PL min is the gray value of the pixel with the smallest gray value in the crack part; LPD is the proportion of deep cracks, N D is the total number of deep pixels.

[0071] In grayscale images, generally speaking, the larger the grayscale value, the shallower the crack. This is because the crack area usually reflects more light, resulting in a higher grayscale value. In contrast, the area surrounding the crack appears darker. Therefore, in the calculation of the maximum crack depth relative comparison, the minimum grayscale value of the crack area is used to represent the maximum depth; the pixel point whose grayscale value of the crack part is less than the crack grayscale threshold is defined as a deep pixel point.

[0072] Step 3: The blasting data, pre-blasting data and corresponding blasting response data of the N explosion experiments in step 1 form an explosion experiment set; construct a blasting data adversarial network, use the explosion experiment set to train the blasting data adversarial network, and use the trained blasting data adversarial network to generate a virtual explosion experiment data set;

[0073] The blasting data adversarial network includes a generator and a discriminator, and the generator includes an input layer, an output layer and a hidden layer; the input layer of the generator is used to receive the explosion experiment set, and add random noise to the blasting data, pre-blasting data and corresponding blasting response data as the initial input; the hidden layer of the generator is composed of multiple layers, each layer contains a number of nodes, and each node is connected to the previous layer through a weight; the ReLU activation function is used to perform a nonlinear transformation on the input data, so that the model can generate complex distributions; an independent neuron is set in the output layer of the generator, which is responsible for converting the transformed data of the hidden layer into virtual data generated by the generator;

[0074] The discriminator includes an input layer, an output layer and a hidden layer; the discriminator input layer receives the explosion test set and the virtual data generated by the generator; the hidden layer of the discriminator is composed of multiple layers, each layer contains a number of nodes, and each node is connected to the previous layer through a weight; the ReLU activation function is used to perform a nonlinear transformation on the input data so that the model can generate a complex distribution; an independent neuron is set in the discriminator output layer, and the independent neuron of the discriminator output layer uses a Sigmoid activation function, and the output range is [0,1] to reflect the output probability of the discriminator for real data and generated virtual data;

[0075] The formula of the loss function is:

[0076]

[0077] Among them, L is the loss function of the blasting data adversarial network, D(x) is the output probability of the discriminator for the real data; E is the mathematical expectation operation, x~p date For x, we need to get the real data p date Get a sample from the , D9G(z)) is the output probability of the discriminator for the generated virtual data; z~p z is z from a random noise distribution p z Get a sample from .

[0078] Due to the high cost of explosion experiments, it is impossible to obtain a large amount of explosion experiment data to train the subsequent blasting response prediction model. Without a large amount of explosion experiment data for training, the accuracy of the results obtained by the blasting response prediction model will be insufficient; therefore, it is necessary to generate a large amount of virtual explosion experiment data through the blasting data adversarial network to complete the training of the blasting response prediction model.

[0079] Step 4: Using the blasting data and pre-blasting data in the virtual explosion experiment data set as input and the explosion response data as labels, the blasting response prediction model is trained, the pre-blasting data and blasting data of the current blasting work are obtained, and they are input into the blasting response model to obtain the predicted blasting response data;

[0080] The blast response prediction model adopts a feedforward neural network, with the explosion response data as the label, and the blast data and pre-blast data in the virtual explosion experiment data set are used to train and optimize the blast response prediction model using existing technologies; specifically including: an input layer, a hidden layer, an output layer and an activation function, the input layer is responsible for receiving the environmental data at historical moments; the hidden layer is used for data processing of the environmental data at historical moments; it is composed of multiple layers, each layer contains 4 nodes, and the nodes of each hidden layer are connected to the previous layer through weights, which are used to perform feature abstraction and nonlinear transformation on the input environmental data at historical moments; by using the ReLu activation function, nonlinear relationships are introduced so that the model can fit complex feature relationships; an independent neuron is set in the output layer, which is responsible for converting the local and high-level feature representations extracted by the hidden layer for outputting the response value of the air purifier; a root mean square error loss function is used; the input data is calculated once through the network to obtain the output result, the loss function is calculated according to the predicted value and the true value, the gradient of the loss function for each weight and bias is calculated by the chain rule, and the weights and biases of the network are updated using the gradient descent algorithm to minimize the loss function.

[0081] Step 5: Comprehensively analyze the pre-blasting data and explosion response data corresponding to the current blasting work to obtain the property residual coefficient and the blasting damage coefficient; generate a comprehensive blasting response coefficient based on the property residual coefficient and the blasting damage coefficient.

[0082] The blast response data include images, compressive strength, elastic modulus and mass of the UHPC slab after blasting, and stress distribution on the back side of the UHPC slab, which is the side opposite to the blasting side of the UHPC slab.

[0083] The specific logic for obtaining the residual coefficient of properties is as follows: the compressive strength, elastic modulus and mass of the ultra-high performance concrete slab are comprehensively analyzed to obtain the residual strength ratio and residual mass ratio;

[0084] The specific formula for calculating the residual coefficient of the property is:

[0085]

[0086] Among them, R s is the residual coefficient of the property, f o is the compressive strength of the ultra-high performance concrete slab before blasting, f pis the compressive strength of the ultra-high performance concrete slab after blasting, K p is the elastic modulus of the ultra-high performance concrete slab before blasting, K p Elastic modulus of ultra-high performance concrete slab after blasting; m o is the mass of the ultra-high performance concrete slab before blasting, m p The quality of ultra-high performance concrete slab after blasting.

[0087] The property residual coefficient indicates the degree of retention of the comprehensive performance of the ultra-high performance concrete slab after blasting relative to that before blasting. The larger its value, the closer the comprehensive performance of the concrete slab after blasting is to the initial state, and the better the explosion resistance. The property residual coefficient directly compares the retention ratio of various properties of the material before and after blasting, reflects the change in bearing capacity, and also takes into account the weakening of the material's deformation capacity, providing a more comprehensive damage response index. The compressive strength of the ultra-high performance concrete slab reflects the bearing capacity under axial force, and the elastic modulus of the ultra-high performance concrete slab reflects the deformation resistance of the ultra-high performance concrete slab under stress. It reflects the degree of retention of the compressive strength of the ultra-high performance concrete slab after blasting relative to that before blasting. The larger the value, the closer the compressive strength of the concrete slab after blasting is to the initial state, and the better the explosion resistance. It reflects the degree of retention of the elastic modulus of the ultra-high performance concrete slab after blasting relative to that before blasting. The larger the value, the closer the elastic modulus of the concrete slab after blasting is to the initial state, and the better the explosion resistance. It reflects the degree of retention of the mass of the ultra-high performance concrete slab after blasting relative to that before blasting. The larger the value, the closer the mass of the concrete slab after blasting is to the initial state, and the better the explosion resistance.

[0088] The blasting damage coefficient is generated according to the crack ratio, the relative ratio of the maximum crack depth, the deep crack ratio and the stress distribution variance of the plate back. The specific formula for generating the blasting damage coefficient is:

[0089]

[0090] Among them, BH is the explosion damage coefficient, LX is the crack ratio; LPH is the relative ratio of the maximum crack depth, LPD is the deep crack ratio, σ B is the stress distribution variance of the back of the plate;

[0091] The blasting damage coefficient is a coefficient that reflects the degree of damage to the plate under blasting after comprehensively considering factors such as crack ratio, crack depth and stress distribution. It is an indicator to measure the degree of damage to the plate. The smaller the BH is, the more serious the damage to the plate under blasting. The crack ratio refers to the proportion of the crack area in the entire plate area, which measures the breadth or extension of the crack; the larger its value, the wider the crack distribution and the more serious the damage to the plate. The more cracks there are, the more vulnerable the plate is to blasting or impact damage. The deep crack ratio indicates the proportion of deeper cracks in the plate; the larger the LPD is, the higher the proportion of deep cracks in the plate, which means that these deep cracks may be more likely to cause damage to the plate, because deep cracks are usually easier to expand or cause fractures than shallow cracks. Since both the crack ratio and the deep crack ratio reflect the degree of damage to the concrete slab through the proportion of cracks, The two are coupled to jointly describe the degree of damage to the concrete slab; the gray value of the pixel with the smallest gray value in the crack part represents the smallest pixel in the gray value of the crack area, which usually indicates the deepest part of the crack or the most vulnerable area. The larger the LPH, the shallower the deepest position, the relatively less serious crack or the smaller the surface damage. However, if the LPH is small, it means that the crack is deeper and poses a greater threat to the stability of the structure. Before the blasting, due to the relatively regular structure of the concrete slab, the stress distribution on the back of the slab is relatively uniform, and the variance of the stress distribution on the back of the slab is relatively small. However, after the blasting, due to the destruction of the concrete slab, its structure is no longer regular, so the stress distribution on the back of the slab is also uneven. Generally speaking, the greater the damage to the concrete slab, the higher the degree of structural irregularity, and the greater the variance of the stress distribution on the back of the slab.

[0092] The comprehensive blasting response coefficient is generated according to the residual coefficient of the property and the blasting damage coefficient. The specific formula for generating the comprehensive blasting response coefficient is:

[0093] BXP=BH*R s

[0094] Among them, BXP is the comprehensive blasting response coefficient, R sis the residual coefficient of properties. The comprehensive blasting response coefficient reflects the comprehensive response of the ultra-high performance concrete slab. The larger its value, the better the performance of the ultra-high performance concrete slab when responding to blasting. The generation of this coefficient can provide an important basis for the evaluation of the comprehensive response performance of the ultra-high performance concrete slab; the blasting damage coefficient is a coefficient that reflects the degree of damage to the slab under the action of blasting. It is an indicator to measure the degree of damage to the slab. The larger the BH is, the smaller the degree of damage to the slab under the action of blasting, and the better the properties of the ultra-high performance concrete slab when responding to blasting. The residual coefficient of properties indicates the degree of retention of the comprehensive performance of the ultra-high performance concrete slab after blasting relative to that before blasting. The larger its value, the closer the comprehensive performance of the concrete slab after blasting is to the initial state; the better the performance of the ultra-high performance concrete slab when responding to blasting.

[0095] See also Figure 2 The present invention further provides a system for predicting the explosion resistance of ultra-high performance concrete slabs based on machine learning, and the system is used to implement the method for predicting the explosion resistance of ultra-high performance concrete slabs based on machine learning, specifically comprising:

[0096] A data acquisition module, used to perform N explosion tests on the ultra-high performance concrete slab to acquire explosion test data, wherein the explosion test data includes explosion data, data before explosion, data after explosion, and an image of the ultra-high performance concrete slab after explosion;

[0097] A response analysis module is used to analyze the image of the ultra-high performance concrete slab after blasting to obtain post-blasting image feature data, and combine the post-blasting data and the post-blasting image feature data to generate blasting response data;

[0098] The data virtualization module is used to form an explosion experiment set with the explosion data, pre-explosion data and corresponding explosion response data of the N explosion experiments in step 1; construct a blasting data adversarial network, train the blasting data adversarial network using the explosion experiment set, and generate a virtual explosion experiment data set using the trained blasting data adversarial network;

[0099] The model building module is used to take the blasting data and pre-blasting data in the virtual blasting experiment data set as input and the blasting response data as labels, train the blasting response prediction model, obtain the pre-blasting data and blasting data of the current blasting work, and input them into the blasting response model to obtain the predicted blasting response data;

[0100] The comprehensive analysis module is used to comprehensively analyze the pre-blasting data and explosion response data corresponding to the current blasting work to obtain the property residual coefficient and the blasting damage coefficient; and generate the comprehensive blasting response coefficient based on the property residual coefficient and the blasting damage coefficient.

[0101] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0102] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0103] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0104] The above is only a specific implementation method of the present application, but the protection scope of the present application is not limited thereto. Any technical personnel familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A method for predicting the explosion resistance of ultra-high performance concrete slabs based on machine learning, characterized in that: The specific steps include: Step 1: performing N explosion tests on the ultra-high performance concrete slab to obtain explosion test data, wherein the explosion test data includes explosion data, data before explosion, data after explosion, and an image of the ultra-high performance concrete slab after explosion; Step 2: Analyze the image of the ultra-high performance concrete slab after blasting to obtain post-blasting image feature data, and combine the post-blasting data and the post-blasting image feature data to generate blasting response data; Step 3: The blasting data, pre-blasting data and corresponding blasting response data of the N explosion experiments in step 1 form an explosion experiment set; construct a blasting data adversarial network, use the explosion experiment set to train the blasting data adversarial network, and use the trained blasting data adversarial network to generate a virtual explosion experiment data set; Step 4: Using the blasting data and pre-blasting data in the virtual explosion experiment data set as input and the explosion response data as labels, the blasting response prediction model is trained, the pre-blasting data and blasting data of the current blasting work are obtained, and they are input into the blasting response model to obtain the predicted blasting response data; Step 5: Comprehensively analyze the pre-blasting data and explosion response data corresponding to the current blasting work to obtain the property residual coefficient and the blasting damage coefficient; generate a comprehensive blasting response coefficient based on the property residual coefficient and the blasting damage coefficient.

2. The method for predicting explosion-resistant response of ultra-high performance concrete slabs based on machine learning according to claim 1, characterized in that: The blasting data include the pressure and duration of the applied blasting shock wave, the distance between the blasting shock wave emission source and the plane of the ultra-high performance concrete slab, the angle between the line connecting the blasting shock wave emission source and the center position of the ultra-high performance concrete slab and the plane of the ultra-high performance concrete slab, the pre-blasting data include the thickness, compressive strength, tensile strength, elastic modulus, mass and density of the ultra-high performance concrete slab; the post-blasting data include the compressive strength, elastic modulus and mass of the ultra-high performance concrete slab after blasting, and the stress distribution on the back of the ultra-high performance concrete slab.

3. The method for predicting explosion-resistant response of ultra-high performance concrete slabs based on machine learning according to claim 1, characterized in that: The explosion response data include the crack ratio, maximum crack depth relative ratio, deep crack ratio, stress distribution variance of the back of the slab, compressive strength, elastic modulus and mass of the ultra-high performance concrete slab after blasting; The specific logic for calculating the crack ratio, the relative ratio of the maximum crack depth, and the deep crack ratio is as follows: the frequency of the pixel points of each level in the 256 gray levels in the image of the ultra-high performance concrete slab is counted to form a grayscale histogram, and the inter-class variance of each threshold in the histogram is calculated. The best separation threshold is selected through the inter-class variance, and the image of the front of the ultra-high performance concrete slab is threshold segmented using the best segmentation threshold to identify the crack part and the concrete part. The inter-class variance of the corresponding threshold is calculated according to the following formula: σ(t) 2 =w B (t)*w F (t)*(μ B (t)-m F (t)) 2 Among them, σ(t) 2 represents the inter-class variance when the threshold is t, t∈[0, 255], and t∈N + ,w B (t), w F (t) are the number of pixels in the crack part and the concrete part when the threshold is t, μ B (t), μ F (t) are the mean grayscale values ​​of the pixels in the background and foreground when the threshold is t; Calculate the threshold value when the inter-class variance reaches the maximum, and define this value as the optimal separation threshold; The number of pixels occupied by the crack part is divided by the total number of pixels to obtain the crack ratio; the average gray value of the concrete part is calculated, the pixel with the smallest gray value in the crack part is compared with the average gray value of the concrete part to obtain the maximum crack depth, a crack gray value threshold is preset, and the pixel with a gray value of the crack part less than the crack gray value threshold is defined as a deep pixel, the number of deep pixels is counted, and the number of deep pixels is divided by the number of pixels in the crack part to obtain the deep crack ratio, and the crack gray value threshold is 50% of the average gray value of the concrete part; The specific formulas for calculating the crack ratio, the relative ratio of the maximum crack depth, the deep crack ratio, and the stress distribution variance of the back of the plate are: Among them, Ph is the average gray value of the concrete part, Ph i is the gray value of the i-th pixel of the concrete part, N H is the total number of pixels in the concrete part, i is the index of the pixel in the concrete part, LX is the crack ratio, N P is the total number of pixels in the crack part, N A is the total number of pixels, LPH is the maximum crack depth relative ratio, PL min is the gray value of the pixel with the smallest gray value in the crack part; LPD is the proportion of deep cracks, N D is the total number of deep pixels.

4. The method for predicting explosion-resistant response of ultra-high performance concrete slabs based on machine learning according to claim 1, characterized in that: The blasting data adversarial network includes a generator and a discriminator, and the generator includes an input layer, an output layer and a hidden layer; the input layer of the generator is used to receive the explosion experiment set, and add random noise to the blasting data, pre-blasting data and corresponding blasting response data as the initial input; the hidden layer of the generator is composed of multiple layers, each layer contains a number of nodes, and each node is connected to the previous layer through a weight; the ReLU activation function is used to perform a nonlinear transformation on the input data, so that the model can generate complex distributions; an independent neuron is set in the output layer of the generator, which is responsible for converting the transformed data of the hidden layer into virtual data generated by the generator; The discriminator includes an input layer, an output layer and a hidden layer; the discriminator input layer receives the explosion test set and the virtual data generated by the generator; the hidden layer of the discriminator is composed of multiple layers, each layer contains a number of nodes, and each node is connected to the previous layer through a weight; the ReLU activation function is used to perform a nonlinear transformation on the input data so that the model can generate a complex distribution; an independent neuron is set in the discriminator output layer, and the independent neuron of the discriminator output layer uses a Sigmoid activation function, and the output range is [0,1] to reflect the output probability of the discriminator for real data and generated virtual data; The formula of the loss function is: Among them, L is the loss function of the blasting data adversarial network, D(x) is the output probability of the discriminator for the real data; E is the mathematical expectation operation, x~p date For x, we need to get the real data p date Get a sample from the , D(G(z)) is the output probability of the discriminator for the generated virtual data; z~p z is z from a random noise distribution p z Get a sample from .

5. The method for predicting explosion-resistant response of ultra-high performance concrete slabs based on machine learning according to claim 1, characterized in that: The specific logic for obtaining the residual coefficient of properties is as follows: the compressive strength, elastic modulus and mass of the ultra-high performance concrete slab are comprehensively analyzed to obtain the residual strength ratio and residual mass ratio; The specific formula for calculating the residual coefficient of the property is: Among them, R s is the residual coefficient of the property, f o is the compressive strength of the ultra-high performance concrete slab before blasting, f p is the compressive strength of the ultra-high performance concrete slab after blasting, K p is the elastic modulus of the ultra-high performance concrete slab before blasting, K p Elastic modulus of ultra-high performance concrete slab after blasting; m o is the mass of the ultra-high performance concrete slab before blasting, m p The quality of ultra-high performance concrete slab after blasting.

6. The method for predicting explosion-resistant response of ultra-high performance concrete slabs based on machine learning according to claim 2, characterized in that: The blasting damage coefficient is generated according to the crack ratio, the relative ratio of the maximum crack depth, the deep crack ratio and the stress distribution variance of the plate back. The specific formula for generating the blasting damage coefficient is: Among them, BH is the blasting damage coefficient, LX is the crack ratio; LPH is the relative ratio of the maximum crack depth, LPD is the deep crack ratio, σ B is the stress distribution variance of the back of the plate; The comprehensive blasting response coefficient is generated according to the residual coefficient of the property and the blasting damage coefficient. The specific formula for generating the comprehensive blasting response coefficient is: BXP=BH*R s Among them, BXP is the comprehensive blasting response coefficient, R s is the property residual coefficient.

7. A machine learning-based ultra-high performance concrete slab explosion-resistant response prediction system, characterized in that: The system is used to implement the method for predicting the explosion-resistant response of ultra-high performance concrete slabs based on machine learning as described in any one of claims 1 to 6, and specifically includes: A data acquisition module, used to perform N explosion tests on the ultra-high performance concrete slab to acquire explosion test data, wherein the explosion test data includes explosion data, data before explosion, data after explosion, and an image of the ultra-high performance concrete slab after explosion; A response analysis module is used to analyze the image of the ultra-high performance concrete slab after blasting to obtain post-blasting image feature data, and combine the post-blasting data and the post-blasting image feature data to generate blasting response data; The data virtualization module is used to form an explosion experiment set with the explosion data, pre-explosion data and corresponding explosion response data of the N explosion experiments in step 1; construct a blasting data adversarial network, train the blasting data adversarial network using the explosion experiment set, and generate a virtual explosion experiment data set using the trained blasting data adversarial network; The model building module is used to take the blasting data and pre-blasting data in the virtual blasting experiment data set as input and the blasting response data as labels, train the blasting response prediction model, obtain the pre-blasting data and blasting data of the current blasting work, and input them into the blasting response model to obtain the predicted blasting response data; The comprehensive analysis module is used to comprehensively analyze the pre-blasting data and explosion response data corresponding to the current blasting work to obtain the property residual coefficient and the blasting damage coefficient; and generate the comprehensive blasting response coefficient based on the property residual coefficient and the blasting damage coefficient.

Images