ECC ratio optimization method and system based on continuous conditional generative adversarial network, and terminal

Abstract

The invention relates to the technical field of data processing, and discloses an ECC proportion optimization method, system and terminal based on a continuous conditional generative adversarial network, and the method comprises the steps: inputting a target performance condition vector and a random noise vector into a generator, and outputting a generated proportion vector; a real matching vector is obtained, the generated matching vector, the real matching vector and the target performance condition vector are output to a discriminator, and the authenticity degree is output; calculating discriminator loss and generator loss according to the authenticity degree, and optimizing the conditional generative adversarial network according to the discriminator loss and the generator loss to obtain a target network; obtaining a current target performance condition of the ECC, inputting the current target performance condition into the target network, and screening the output to obtain a qualified ratio; and calculating the total carbon emission corresponding to the qualified ratios, and selecting the qualified ratio with the lowest total carbon emission as the target ECC ratio. According to the invention, the efficiency and the stability of the proportion design are improved, and meanwhile, the carbon emission is obviously reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to an ECC proportioning optimization method, system, terminal and computer readable storage medium based on a continuous condition generative adversarial network. BACKGROUND

[0002] Engineered Cementitious Composites (ECC) is a new type of cement-based material with high ductility, high toughness and excellent crack control ability. Under tensile action, ECC can produce multiple cracking and strain hardening phenomena, and its ultimate tensile strain is usually hundreds of times that of ordinary concrete, so it has wide application prospects in the fields of bridges, tunnels, ocean engineering and infrastructure repair and reinforcement. The performance of ECC is highly dependent on its proportioning design, and reasonable proportioning not only relates to the mechanical properties such as ultimate tensile stress and ultimate tensile strain, but also determines the environmental impact and carbon emission level of the material. However, due to the large number of design parameters and the complexity of their mutual coupling, how to realize intelligent proportioning optimization that takes into account both mechanical properties and low-carbon goals has become a key problem in current research and application.

[0003] Existing ECC proportioning design methods mainly include experimental optimization methods, mechanical model-based methods, numerical simulation methods, data-driven methods and generative model methods. The experimental method relies on a large number of experiments, is low in efficiency and has a limited optimization range; the mechanical model needs strong assumptions and is difficult to fully reflect the complexity of the system; the numerical simulation method relies on finite element or phase field models, has high computational cost and is difficult to combine with low-carbon constraints; the data-driven method can use machine learning to predict performance and combine optimization algorithms to back-propagate the proportion, but it is a two-step process of prediction and optimization, lacks end-to-end performance driving capability, and often only focuses on a single performance indicator; the generative model method can directly generate the proportion, but has poor training stability, lacks reasonable constraints, and has not considered carbon emissions as an optimization target. Therefore, the existing methods have obvious deficiencies in multi-objective optimization, generation stability and low-carbon design.

[0004] Therefore, the prior art still needs to be improved and developed. SUMMARY

[0005] The main purpose of the present application is to provide an ECC proportioning optimization method, system, terminal and computer readable storage medium based on a continuous condition generative adversarial network, which aims to solve the problem that the existing ECC proportioning design methods have obvious deficiencies in multi-objective optimization, generation stability and low-carbon design, and cannot realize the consideration of both mechanical properties and low-carbon goals, resulting in low efficiency and stability of ECC proportioning.

[0006] In order to achieve the above object, the application provides an ECC ratio optimization method based on a continuous conditional generative adversarial network, which comprises the following steps: A conditional generative adversarial network is constructed, which comprises a generator and a discriminator. A target performance condition vector is set, a random noise vector is obtained, the target performance condition vector and the random noise vector are input into the generator, and a generated ratio vector is output. A ratio-performance database is constructed, a real ratio vector is obtained according to the ratio-performance database, the generated ratio vector, the real ratio vector and the target performance condition vector are output to the discriminator, and a first degree of authenticity and a second degree of authenticity are output. The discriminator loss and the generator loss are calculated according to the first degree of authenticity and the second degree of authenticity, the conditional generative adversarial network is optimized according to the discriminator loss and the generator loss, and a trained network is obtained. The generated ratio quality of the trained network is evaluated, and if the generated ratio quality evaluation is passed, a target network is obtained. The current target performance condition of the ECC is obtained, the current target performance condition is input into the target network, a plurality of candidate ratios are generated, the limit performance of each candidate ratio is calculated respectively, and a plurality of qualified ratios are selected from all the candidate ratios according to all the limit performances. The total carbon emission corresponding to each qualified ratio is calculated, and the qualified ratio with the lowest total carbon emission is selected as the target ECC ratio and output.

[0007] Optionally, the ECC ratio optimization method based on the continuous conditional generative adversarial network, wherein the target performance condition vector and the random noise vector are input into the generator to output a generated ratio vector, specifically comprising: The random noise vector is input into the feature extraction layer of the generator for feature extraction to obtain a first current hidden layer feature. The target performance condition vector is input into the condition mapping network of the generator for processing to generate a first feature linear modulation parameter and a second feature linear modulation parameter. ; Wherein, And The first feature linear modulation parameter and the second feature linear modulation parameter are represented by The condition mapping network is represented by The target performance condition vector is represented by , represents the target performance condition vector, composed of limit strain and limit stress ; inputting the first current hidden layer feature, the first feature linear modulation parameter and the second feature linear modulation parameter into a linear feature adjustment layer of the generator, linearly modulating the current hidden layer feature through the first feature linear modulation parameter and the second feature linear modulation parameter to obtain a first modulation feature: ; wherein, represents the first current hidden layer feature, represents the first modulation feature output by the linear feature adjustment layer, represents the first feature linear modulation parameter, represents the second feature linear modulation parameter; outputting a generated blending vector after processing the modulation feature through an activation function of the generator.

[0008] Optionally, the ECC blending optimization method based on the continuous condition generative adversarial network, wherein the blending-performance database is constructed, and a real blending vector is obtained from the blending-performance database, and the method specifically comprises: constructing a blending-performance database, performing data cleaning, outlier removal, missing value filling and normalization processing on real ECC blending and performance data in the blending-performance database to obtain a target database; obtaining a real ECC blending from the target database as a real blending vector used for training the conditional generative adversarial network.

[0009] Optionally, the ECC blending optimization method based on the continuous condition generative adversarial network, wherein the generated blending vector, the real blending vector and the target performance condition vector are output to the discriminator to output a first degree of authenticity and a second degree of authenticity, and the method specifically comprises: outputting the generated blending vector and the real blending vector to a feature extraction layer of the discriminator for feature extraction to obtain a second current hidden layer feature; inputting the target performance condition vector into a condition mapping network of the discriminator for processing to generate a first feature linear modulation parameter and a second feature linear modulation parameter; inputting the second current hidden layer feature, the first feature linear modulation parameter and the second feature linear modulation parameter into a linear feature adjustment layer of the generator, linearly modulating the current hidden layer feature through the first feature linear modulation parameter and the second feature linear modulation parameter to obtain a second modulation feature; The second modulation feature is subjected to dimensionality reduction and quantization processing through the preset layer structure and activation function of the discriminator, and the first degree of authenticity and the second degree of authenticity are output.

[0010] Optionally, in the ECC allocation optimization method based on continuous conditional generative adversarial networks, the discriminator loss is Wasserstein-GP loss and the generator loss is Wasserstein-BD loss. The step of calculating the discriminator loss and generator loss based on the first and second truth levels, and optimizing the conditional generative adversarial network based on the discriminator loss and generator loss to obtain the trained network specifically includes: The Wasserstein-GP loss is calculated based on the first and second degrees of truth: ; in, This represents the Wasserstein-GP loss. This represents a weighted average of all possible values ​​of a random variable. Indicates the degree of authenticity of the first statement. Indicates the second degree of truth or falsehood. This represents the generated ratio vector. This represents the true proportion vector. This represents the target performance condition vector. Indicates the gradient penalty weights. Represents the gradient penalty term. , This represents the linear interpolation between the generated proportion vector and the actual proportion vector. Indicates to Find the gradient. Denotes the Euclidean norm; The Wasserstein-BD loss is calculated based on the first and second degrees of truth: ; in, This represents the Wasserstein-BD loss. BD represents the boundary loss weights. , This represents a modified linear function. and These represent the upper and lower bounds of the range for each ratio parameter, respectively, and N represents the sample size. Indicates the first The generated matching vector for each sample.

[0011] Optionally, the ECC matching optimization method based on continuous conditional generative adversarial networks, wherein evaluating the quality of the generated matching of the trained network, and obtaining the target network if the quality evaluation of the generated matching passes, specifically includes: The quality of the generated matchmaking by the trained network is evaluated using the target database to obtain the single-point bias and average dispersion of the generated matchmaking: ; ; in, This indicates the single-point deviation. This represents the average dispersion. This represents the average strain of the current generated mix proportion. This represents the average stress of the current generated mix proportion. This refers to the strain labels in the target database. This refers to the stress labels in the target database. Indicates the number of samples. Indicates the ordinal number of the sample. This represents the i-th strain in the current generated mix. This represents the i-th stress in the current generated mix proportion; If both the single-point deviation and the average dispersion are less than their respective thresholds, the quality assessment of the generated ratio is passed, and the target network is obtained.

[0012] Optionally, the ECC mix design optimization method based on continuous conditional generative adversarial networks, wherein the qualified mix design includes: cement content, fly ash content, silica fume content, blast furnace slag content, ratio of sand content to cementitious material content, ratio of water content to cementitious material content, and fiber volume fraction. The calculation of the total carbon emissions corresponding to the qualified ratio specifically includes: Obtain the amount of each material in the qualified formula, and calculate the total carbon emissions corresponding to the qualified formula based on the amount of each material in the qualified formula: ; in, This represents the total carbon emissions. This indicates the amount of each material used. Carbon emission factor per unit material for each material This indicates the type number of the current material.

[0013] Furthermore, to achieve the above objectives, the present invention also provides an ECC allocation optimization system based on a continuous conditional generative adversarial network, wherein the ECC allocation optimization system based on a continuous conditional generative adversarial network includes: A network model building module is used to construct a conditional generative adversarial network, wherein the conditional generative adversarial network includes a generator and a discriminator; A ratio calculation module is used to set a target performance condition vector, obtain a random noise vector, input the target performance condition vector and the random noise vector into the generator, and output a ratio calculation vector. The authenticity calculation module is used to construct a ratio-performance database, obtain the true ratio vector according to the ratio-performance database, and output the generated ratio vector, the true ratio vector and the target performance condition vector to the discriminator, and output the first authenticity degree and the second authenticity degree. The network training optimization module is used to calculate the discriminator loss and generator loss based on the first degree of truth and the second degree of truth, and to optimize the conditional generative adversarial network based on the discriminator loss and the generator loss to obtain the trained network. The network quality evaluation module is used to evaluate the quality of the generated configurations of the trained network. If the quality evaluation of the generated configurations passes, the target network is obtained. The qualified mix generation module is used to obtain the current target performance conditions of ECC, input the current target performance conditions into the target network, generate multiple candidate mixes, calculate the limit performance of each candidate mix, and select multiple qualified mixes from all the candidate mixes based on all the limit performances. The carbon emission assessment module is used to calculate the total carbon emissions corresponding to each of the qualified ratios, select the qualified ratio with the lowest total carbon emissions from all the qualified ratios as the target ECC ratio, and output it.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and an ECC ratio optimization program based on a continuous conditional generative adversarial network stored in the memory and executable on the processor, wherein when the ECC ratio optimization program based on a continuous conditional generative adversarial network is executed by the processor, the steps of the ECC ratio optimization method based on a continuous conditional generative adversarial network as described above are implemented.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores an ECC ratio optimization program based on a continuous conditional generative adversarial network, and when the ECC ratio optimization program based on a continuous conditional generative adversarial network is executed by a processor, it implements the steps of the ECC ratio optimization method based on a continuous conditional generative adversarial network as described above.

[0016] In this invention, a target performance condition vector and a random noise vector are input into a generator, which outputs a generated ratio vector. A true ratio vector is obtained, and the generated ratio vector, the true ratio vector, and the target performance condition vector are output to a discriminator, which outputs the degree of truth or falsehood. The discriminator loss and generator loss are calculated based on the degree of truth or falsehood. The conditional generative adversarial network (GAN) is optimized based on these losses to obtain the target network. The current target performance conditions of the ECC are obtained and input into the target network. The outputs are filtered to obtain qualified ratios. The total carbon emissions corresponding to the qualified ratios are calculated, and the qualified ratio with the lowest total carbon emissions is selected as the target ECC ratio and output. This invention improves the efficiency and stability of ratio design, significantly reduces carbon emissions while ensuring the tensile performance of ECC, and achieves multi-objective optimization and green design, demonstrating good engineering application prospects and promotional value. Attached Figure Description

[0017] Figure 1 This is a flowchart of a preferred embodiment of the ECC allocation optimization method based on continuous conditional generative adversarial networks of the present invention; Figure 2 This is a diagram of the architecture of the conditional generative adversarial network in the ECC ratio optimization method based on continuous conditional generative adversarial network of the present invention. Figure 3 This is a flowchart of another preferred embodiment of the ECC allocation optimization method based on continuous conditional generative adversarial networks of the present invention; Figure 4 This is a structural diagram of a preferred embodiment of the ECC allocation optimization system based on continuous conditional generative adversarial networks of the present invention; Figure 5 This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0018] This application provides a method, system, and terminal for ECC allocation optimization based on continuous conditional generative adversarial networks. To make the objectives, technical solutions, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0019] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0020] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0021] The preferred embodiment of the ECC allocation optimization method based on continuous conditional generative adversarial networks described in this invention, such as... Figure 1 As shown, the ECC allocation optimization method based on continuous conditional generative adversarial networks includes the following steps: Step S10: Construct a conditional generative adversarial network, which includes a generator and a discriminator.

[0022] Specifically, such as Figure 2 As shown, the conditional generative adversarial network includes a generator and a discriminator. The generator generates candidate configurations based on random noise and target performance conditions. The core structure of the generator is a linear feature adjustment layer (FiLM) and a condition map network (ConditionMap), along with other structures such as linear layers, leaky rectified linear units, and condition map layers. The generator linearly modulates the hidden layer features of the generator through the feature linear modulation (FiLM) parameters generated by the Condition Map network (ConditionMap), thereby achieving conditional control of the generated configurations on the target performance.

[0023] Furthermore, the core structure of the discriminator is similar to that of the generator, also consisting of a Linear Feature Adjustment Layer (FiLM) and a Condition Map network, which will not be elaborated further here. It is understandable that the discriminator adds random deactivation processing after the leaked rectified linear unit. Its purpose is to prevent overfitting. By randomly shutting down some neurons during training, it enhances the discriminator's generalization ability to the input data and avoids its over-reliance on specific features to determine the authenticity of data.

[0024] Step S20: Set the target performance condition vector, obtain the random noise vector, input the target performance condition vector and the random noise vector into the generator, and output the generated ratio vector.

[0025] In this embodiment, a target performance condition vector is set, which represents the desired performance data after ECC matching. A random noise vector is obtained as input to the generator; this random noise vector is a list (or array) composed of completely random or pseudo-random numbers. The value of each element in the vector is independently and randomly generated, following a Gaussian normal distribution.

[0026] Further, the target performance condition vector and the random noise vector are input into the generator, and the generated ratio vector is output, specifically including: The random noise vector is input into the feature extraction layer of the generator to extract features and obtain the first current hidden layer features. The target performance condition vector is input into the conditional mapping network of the generator for processing, generating the first feature linear modulation parameter and the second feature linear modulation parameter: ; in, and This represents the first characteristic linear modulation parameter and the second characteristic linear modulation parameter. This represents the conditional mapping network. Represents the target performance condition vector. , The target performance condition vector is represented by the limit strain. and ultimate stress composition; The first current hidden layer feature, the first feature linear modulation parameter, and the second feature linear modulation parameter are input into the linear feature adjustment layer of the generator. The current hidden layer feature is linearly modulated using the first feature linear modulation parameter and the second feature linear modulation parameter to obtain the first modulation feature. ; in, This represents the first current hidden layer feature. This represents the first modulation feature output after passing through the linear feature adjustment layer. Indicates the first characteristic linear modulation parameter, This represents the second characteristic linear modulation parameter; After the modulation features are processed by the activation function of the generator, the generated ratio vector is output.

[0027] Understandably, the generator produces candidate allocations based on random noise and target performance conditions. The target performance condition vector is input to the generator. and random noise vector, target performance condition vector FiLM parameters are generated using a Condition Map network. and Linear modulation (FiLM) is applied to the hidden layer features of the generator to conditionally control the generation ratio on the target performance. Finally, the generator obtains the generation ratio vector through activation function mapping. .

[0028] Step S30: Construct a ratio-performance database, obtain the true ratio vector based on the ratio-performance database, and output the generated ratio vector, the true ratio vector, and the target performance condition vector to the discriminator, and output the first degree of authenticity and the second degree of authenticity.

[0029] The construction of the proportion-performance database, and the acquisition of the true proportion vector based on the proportion-performance database, specifically includes: Construct a ratio-performance database, and perform data cleaning, outlier removal, missing value imputation and normalization on the real ECC ratio and performance data in the ratio-performance database to obtain the target database; The real ECC ratios are obtained from the target database as the real ratio vectors used to train the conditional generative adversarial network.

[0030] In this embodiment, proportioning data and corresponding performance data are collected. A proportioning-performance database is constructed based on the comparison data and the performance data. The data is then preprocessed, including: processing the collected ECC proportions (i.e., Figure 2 The ratio vector in the data and performance data (i.e., the ratio vector in the data) Figure 2 Mechanical properties, among which, and These represent the first two materials in the ECC ratio. and The ultimate tensile stress and ultimate tensile strain corresponding to the ECC ratios of all materials are represented respectively. 1, 2...n represent sample numbers, and n is the total number of samples. Data cleaning is performed to remove outliers and duplicate data, and missing values ​​are appropriately filled. At the same time, the ratio parameters are normalized to ensure the integrity, consistency and usability of the data for subsequent model training, and to provide high-quality input for the model.

[0031] Further, the step of outputting the generated ratio vector, the true ratio vector, and the target performance condition vector to the discriminator, and outputting a first degree of truth and a second degree of truth, specifically includes: The generated ratio vector and the real ratio vector are output to the feature extraction layer of the discriminator for feature extraction to obtain the second current hidden layer feature. The target performance condition vector is input into the condition mapping network of the discriminator for processing to generate the first feature linear modulation parameter and the second feature linear modulation parameter. The second current hidden layer feature, the first feature linear modulation parameter, and the second feature linear modulation parameter are input into the linear feature adjustment layer of the generator. The current hidden layer feature is linearly modulated by the first feature linear modulation parameter and the second feature linear modulation parameter to obtain the second modulation feature. The second modulation feature is subjected to dimensionality reduction and quantization processing through the preset layer structure and activation function of the discriminator, and the first degree of authenticity and the second degree of authenticity are output.

[0032] Understandably, in this embodiment, the generated matching vector and the true matching vector are output to the discriminator for processing. The discriminator first extracts features from the generated matching vector and the true matching vector through a feature extraction layer to obtain the second current hidden layer feature. Then, the target performance condition vector is input into the condition mapping network of the discriminator to generate feature linear modulation parameters, and the current hidden layer feature is linearly modulated using the feature linear modulation parameters to obtain the second modulation feature. Finally, the second modulation feature is subjected to dimensionality reduction and quantization processing, and the judgment result is output, namely the first degree of truth and the second degree of truth.

[0033] Step S40: Calculate the discriminator loss and generator loss based on the first and second truth levels, and optimize the conditional generative adversarial network based on the discriminator loss and the generator loss to obtain the trained network.

[0034] In this embodiment, the discriminator loss is Wasserstein-GP loss, and the generator loss is Wasserstein-BD loss.

[0035] The step of calculating the discriminator loss and generator loss based on the first and second truth levels, and optimizing the conditional generative adversarial network based on the discriminator loss and generator loss to obtain the trained network specifically includes: The Wasserstein-GP loss is calculated based on the first and second degrees of truth: ; in, This represents the Wasserstein-GP loss. This represents a weighted average of all possible values ​​of a random variable. Indicates the degree of authenticity of the first statement. Indicates the second degree of truth or falsehood. This represents the generated ratio vector. This represents the true proportion vector. This represents the target performance condition vector. Indicates the gradient penalty weights. Represents the gradient penalty term. , This represents the linear interpolation between the generated proportion vector and the actual proportion vector. Indicates to Find the gradient. Denotes the Euclidean norm; The Wasserstein-BD loss is calculated based on the first and second degrees of truth: ; in, This represents the Wasserstein-BD loss. BD represents the boundary loss weights. , This represents a modified linear function. and These represent the upper and lower bounds of the range for each ratio parameter, respectively, and N represents the sample size. Indicates the first The generated matching vector for each sample.

[0036] Understandably, the Wasserstein-GP loss is a Wasserstein loss with gradient penalty. It uses the Wasserstein distance instead of probability judgment, directly measuring the "transfer cost" between the "generated distribution" and the "true distribution." Even if the discriminator doesn't classify perfectly, it can still provide a stable gradient signal. The addition of gradient penalty (GP) constrains the discriminator's gradient norm to not exceed 1, preventing excessively large discriminator weights from causing gradient explosion, while also suppressing mode collapse.

[0037] Wasserstein-BD loss is a weighted fusion of Wasserstein loss and boundary loss. It enhances the model's ability to distinguish class boundaries by strengthening the penalty for "hard-to-distinguish samples" (i.e. samples located near class boundaries and easily misclassified).

[0038] Furthermore, the discriminator loss and the generator loss guide the parameter optimization direction of the conditional generative adversarial network, thereby obtaining a trained network that meets the training requirements.

[0039] Step S50: Evaluate the quality of the generated ratios of the trained network. If the quality evaluation of the generated ratios passes, the target network is obtained.

[0040] Specifically, the quality of the generated matchmaking by the trained network is evaluated using the target database to obtain the single-point bias and average dispersion of the generated matchmaking: ; ; in, This indicates the single-point deviation. This represents the average dispersion. This represents the average strain of the current generated mix proportion. This represents the average stress of the current generated mix proportion. This refers to the strain labels in the target database. This refers to the stress labels in the target database. Indicates the number of samples. Indicates the ordinal number of the sample. This represents the i-th strain in the current generated mix. This represents the i-th stress in the current generated mix proportion; If both the single-point deviation and the average dispersion are less than their respective thresholds, the quality assessment of the generated ratio is passed, and the target network is obtained.

[0041] In this embodiment, the single-point deviation and average dispersion of the generated ratios are used to evaluate the quality of the generated ratios of the trained model. If the expected results are not achieved, the model's hyperparameters are readjusted during the training phase.

[0042] Step S60: Obtain the current target performance conditions of ECC, input the current target performance conditions into the target network, generate multiple candidate ratios, calculate the limit performance of each candidate ratio, and select multiple qualified ratios from all the candidate ratios based on all the limit performances.

[0043] Specifically, the current target performance conditions of the ECC are obtained, and these conditions are input into the target network. A generator is used to generate a large number of candidate mix proportions under the given performance conditions. The ultimate tensile stress and ultimate strain of each mix proportion are calculated using a prediction model (which can be implemented by empirical formulas or machine learning models). Multiple qualified mix proportions that meet the design performance requirements are then selected from these.

[0044] Step S70: Calculate the total carbon emissions corresponding to each qualified ratio, select the qualified ratio with the lowest total carbon emissions from all qualified ratios as the target ECC ratio and output it.

[0045] It is understood that the qualified proportions include: cement content, fly ash content, silica fume content, blast furnace slag content, the ratio of sand content to cementitious material content, the ratio of water content to cementitious material content, and fiber volume fraction.

[0046] The calculation of the total carbon emissions corresponding to the qualified ratio specifically includes: Obtain the amount of each material in the qualified formula, and calculate the total carbon emissions corresponding to the qualified formula based on the amount of each material in the qualified formula: ; in, This represents the total carbon emissions. This indicates the amount of each material used. Carbon emission factor per unit material for each material This indicates the type number of the current material.

[0047] In this embodiment, the qualified mix proportions with satisfactory performance are input into the carbon emission assessment module. The total carbon emission is calculated based on the amounts of each raw material (i.e., cement content, fly ash content, silica fume content, blast furnace slag content, the ratio of sand to cementitious material content, the ratio of water to cementitious material content, and fiber volume fraction) and the unit carbon emission factor of each raw material. Among the candidate mix proportions, the mix proportion with the lowest carbon emission is selected using the total carbon emission calculation formula, achieving multi-objective optimization that balances mechanical performance and low-carbon requirements. Finally, the ECC mix proportion that meets the target mechanical performance and has the lowest carbon emission is output, providing a reliable multi-objective optimization design scheme for practical engineering applications.

[0048] Furthermore, such as Figure 3 As shown, in another preferred embodiment of the present invention, the implementation steps of the ECC allocation optimization method based on continuous conditional generative adversarial networks are as follows: Step 1: Prepare the dataset: Clean the collected ECC ratio and performance data, remove outliers and duplicate data, and fill in missing values ​​appropriately. At the same time, normalize each ratio parameter to ensure the integrity, consistency and availability of the data for subsequent model training, and obtain the dataset for training.

[0049] Step 2: Training the continuous conditional generative adversarial neural network: (1) Determine model hyperparameters and initialize source model: determine hyperparameters, hidden layer size, activation function, loss weight, generator optimizer, generator learning rate; discriminator hyperparameters: hidden layer size, activation function, gradient penalty weight, discriminator optimizer, discriminator learning rate; conditional mapping network hyperparameters: hidden layer size.

[0050] (2) Model building: Initialize the generator and discriminator.

[0051] (3) Training the source model.

[0052] 1. Sample real data.

[0053] 2. The generator generates fake sample data.

[0054] 3. Update the discriminator and minimize it. .

[0055] 4. Update the generator and minimize it. .

[0056] 5. Repeat steps 1-4 until the model converges, then stop training.

[0057] (4) Generate a ratio assessment. If the expected results are not achieved, return to (2) to adjust the hyperparameters.

[0058] (5) Repeat operations (1)-(4) to determine the model hyperparameters, obtain the optimal model, train and save the model.

[0059] Step 3: Generate mix proportions and predict performance: After training, use the generator to generate a large number of candidate mix proportions under given performance conditions, and use the prediction model (which can be implemented by empirical formulas or machine learning models) to calculate the ultimate tensile stress and ultimate strain of each mix proportion, and select the set of candidate mix proportions that meet the design performance requirements.

[0060] Step 4: Carbon Emission Assessment of Mix Design: Input qualified candidate mix designs into the carbon emission assessment module, and calculate the total carbon emissions based on the amount of each raw material and its unit carbon emission factor. Select the mix design with the lowest carbon emissions from the candidate mix designs to achieve multi-objective optimization, taking into account both mechanical performance and low-carbon requirements.

[0061] Step 5: Output multi-objective mix ratio: Output the ECC mix ratio that meets the target mechanical performance and has the lowest carbon emissions, providing a reliable multi-objective optimization design scheme for practical engineering applications.

[0062] As can be seen, this invention proposes a mix design method for engineering cement-based composite materials (ECC) based on a deep generative model. This method takes target performance indicators (including ultimate stress and ultimate strain) as input and introduces a conditional generative adversarial network (GAN) incorporating FiLM feature modulation. Through adversarial training between the generator and discriminator, it outputs ECC mix proportion parameters (including cement, fly ash, silica fume, slag, sand, and fiber) that meet the target performance, achieving performance-driven reverse mix proportion generation. Simultaneously, this invention establishes a scatter distance method for evaluating the generative model. By calculating the distance between the mean point of the generated samples and the target performance points, as well as the overall dispersion of the generated samples, the accuracy and diversity of the generated results are quantitatively evaluated, thereby ensuring that the generated mix proportions not only meet the performance targets but also possess reasonable distribution characteristics and generalization ability.

[0063] Furthermore, such as Figure 4As shown, based on the above-mentioned ECC allocation optimization method based on continuous conditional generative adversarial networks, the present invention also provides an ECC allocation optimization system based on continuous conditional generative adversarial networks, wherein the ECC allocation optimization system based on continuous conditional generative adversarial networks includes: The network model construction module 51 is used to construct a conditional generative adversarial network, which includes a generator and a discriminator. The ratio calculation module 52 is used to set the target performance condition vector, obtain the random noise vector, input the target performance condition vector and the random noise vector into the generator, and output the ratio calculation vector. The authenticity calculation module 53 is used to construct a ratio-performance database, obtain a true ratio vector based on the ratio-performance database, and output the generated ratio vector, the true ratio vector and the target performance condition vector to the discriminator, and output a first authenticity degree and a second authenticity degree. The network training optimization module 54 is used to calculate the discriminator loss and generator loss based on the first degree of truth and the second degree of truth, and optimize the conditional generative adversarial network based on the discriminator loss and the generator loss to obtain the trained network. The network quality evaluation module 55 is used to evaluate the quality of the generated ratio of the trained network. If the quality evaluation of the generated ratio passes, the target network is obtained. The qualified ratio generation module 56 is used to obtain the current target performance conditions of ECC, input the current target performance conditions into the target network, generate multiple candidate ratios, calculate the limit performance of each candidate ratio, and select multiple qualified ratios from all the candidate ratios based on all the limit performances. The carbon emission assessment module 57 is used to calculate the total carbon emissions corresponding to each of the qualified ratios, select the qualified ratio with the lowest total carbon emissions from all the qualified ratios as the target ECC ratio and output it.

[0064] Furthermore, such as Figure 5 As shown, based on the above-mentioned ECC allocation optimization method and system based on continuous conditional generative adversarial networks, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 5 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0065] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores an ECC matching optimization program 40 based on a continuous conditional generative adversarial network (CGA). This ECC matching optimization program 40 can be executed by the processor 10 to implement the ECC matching optimization method based on a CGA in this application.

[0066] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the ECC ratio optimization method based on continuous conditional generative adversarial networks.

[0067] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components of the terminal communicate with each other via a system bus.

[0068] In one embodiment, when the processor 10 executes the ECC matching optimization program 40 based on the continuous conditional generative adversarial network in the memory 20, the following steps are performed: A conditional generative adversarial network is constructed, which includes a generator and a discriminator. Set a target performance condition vector, obtain a random noise vector, input the target performance condition vector and the random noise vector into the generator, and output a generated ratio vector. Construct a ratio-performance database, obtain a true ratio vector based on the ratio-performance database, and output the generated ratio vector, the true ratio vector, and the target performance condition vector to the discriminator, and output a first degree of authenticity and a second degree of authenticity. The discriminator loss and generator loss are calculated based on the first and second truth levels. The conditional generative adversarial network is then optimized based on the discriminator loss and the generator loss to obtain the trained network. The quality of the generated matchings of the trained network is evaluated. If the quality evaluation of the generated matchings passes, the target network is obtained. Obtain the current target performance conditions of ECC, input the current target performance conditions into the target network, generate multiple candidate configurations, calculate the limit performance of each candidate configuration, and select multiple qualified configurations from all the candidate configurations based on all the limit performances. Calculate the total carbon emissions corresponding to each of the qualified ratios, select the qualified ratio with the lowest total carbon emissions from all the qualified ratios as the target ECC ratio and output it.

[0069] Specifically, the step of inputting the target performance condition vector and the random noise vector into the generator and outputting a generated ratio vector includes: The random noise vector is input into the feature extraction layer of the generator to extract features and obtain the first current hidden layer features. The target performance condition vector is input into the conditional mapping network of the generator for processing, generating the first feature linear modulation parameter and the second feature linear modulation parameter: ; in, and This represents the first characteristic linear modulation parameter and the second characteristic linear modulation parameter. This represents the conditional mapping network. Represents the target performance condition vector. , The target performance condition vector is represented by the limit strain. and ultimate stress composition; The first current hidden layer feature, the first feature linear modulation parameter, and the second feature linear modulation parameter are input into the linear feature adjustment layer of the generator. The current hidden layer feature is linearly modulated using the first feature linear modulation parameter and the second feature linear modulation parameter to obtain the first modulation feature. ; in, This represents the first current hidden layer feature. This represents the first modulation feature output after passing through the linear feature adjustment layer. Indicates the first characteristic linear modulation parameter, This represents the second characteristic linear modulation parameter; After the modulation features are processed by the activation function of the generator, the generated ratio vector is output.

[0070] The construction of the proportion-performance database, and the acquisition of the true proportion vector based on the proportion-performance database, specifically includes: Construct a ratio-performance database, and perform data cleaning, outlier removal, missing value imputation and normalization on the real ECC ratio and performance data in the ratio-performance database to obtain the target database; The real ECC ratios are obtained from the target database as the real ratio vectors used to train the conditional generative adversarial network.

[0071] Specifically, the step of outputting the generated ratio vector, the true ratio vector, and the target performance condition vector to the discriminator, and outputting a first degree of truth and a second degree of truth, includes: The generated ratio vector and the real ratio vector are output to the feature extraction layer of the discriminator for feature extraction to obtain the second current hidden layer feature. The target performance condition vector is input into the condition mapping network of the discriminator for processing to generate the first feature linear modulation parameter and the second feature linear modulation parameter. The second current hidden layer feature, the first feature linear modulation parameter, and the second feature linear modulation parameter are input into the linear feature adjustment layer of the generator. The current hidden layer feature is linearly modulated by the first feature linear modulation parameter and the second feature linear modulation parameter to obtain the second modulation feature. The second modulation feature is subjected to dimensionality reduction and quantization processing through the preset layer structure and activation function of the discriminator, and the first degree of authenticity and the second degree of authenticity are output.

[0072] Wherein, the discriminator loss is Wasserstein-GP loss, and the generator loss is Wasserstein-BD loss; The step of calculating the discriminator loss and generator loss based on the first and second truth levels, and optimizing the conditional generative adversarial network based on the discriminator loss and generator loss to obtain the trained network specifically includes: The Wasserstein-GP loss is calculated based on the first and second degrees of truth: ; in, This represents the Wasserstein-GP loss. This represents a weighted average of all possible values ​​of a random variable. Indicates the degree of authenticity of the first statement. Indicates the second degree of truth or falsehood. This represents the generated ratio vector. This represents the true proportion vector. This represents the target performance condition vector. Indicates the gradient penalty weights. Represents the gradient penalty term. , This represents the linear interpolation between the generated proportion vector and the actual proportion vector. Indicates to Find the gradient. Denotes the Euclidean norm; The Wasserstein-BD loss is calculated based on the first and second degrees of truth: ; in, This represents the Wasserstein-BD loss. BD represents the boundary loss weights. , This represents a modified linear function. and These represent the upper and lower bounds of the range for each ratio parameter, respectively, and N represents the sample size. Indicates the first The generated matching vector for each sample.

[0073] The step of evaluating the quality of the generated matchings in the trained network, and obtaining the target network if the quality evaluation is passed, specifically includes: The quality of the generated matchmaking by the trained network is evaluated using the target database to obtain the single-point bias and average dispersion of the generated matchmaking: ; ; in, This indicates the single-point deviation. This represents the average dispersion. This represents the average strain of the current generated mix proportion. This represents the average stress of the current generated mix proportion. This refers to the strain labels in the target database. This refers to the stress labels in the target database. Indicates the number of samples. Indicates the ordinal number of the sample. This represents the i-th strain in the current generated mix. This represents the i-th stress in the current generated mix proportion; If both the single-point deviation and the average dispersion are less than their respective thresholds, the quality assessment of the generated ratio is passed, and the target network is obtained.

[0074] The qualified mix proportions include: cement content, fly ash content, silica fume content, blast furnace slag content, sand content to cementitious material content ratio, water content to cementitious material content ratio, and fiber volume fraction. The calculation of the total carbon emissions corresponding to the qualified ratio specifically includes: Obtain the amount of each material in the qualified formula, and calculate the total carbon emissions corresponding to the qualified formula based on the amount of each material in the qualified formula: ; in, This represents the total carbon emissions. This indicates the amount of each material used. Carbon emission factor per unit material for each material This indicates the type number of the current material.

[0075] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores an ECC ratio optimization program based on a continuous conditional generative adversarial network, and the ECC ratio optimization program based on a continuous conditional generative adversarial network, when executed by a processor, implements the steps of the ECC ratio optimization method based on a continuous conditional generative adversarial network as described above.

[0076] In summary, this invention provides a method, system, and terminal for ECC ratio optimization based on a continuous conditional generative adversarial network (GAN). The method includes: inputting a target performance condition vector and a random noise vector into a generator, and outputting a generated ratio vector; obtaining a true ratio vector, and outputting the generated ratio vector, the true ratio vector, and the target performance condition vector to a discriminator, outputting the degree of truth or falsehood; calculating the discriminator loss and the generator loss based on the degree of truth or falsehood, and optimizing the conditional GAN ​​based on the discriminator loss and the generator loss to obtain a target network; obtaining the current target performance conditions of the ECC, inputting the current target performance conditions into the target network, and obtaining qualified ratios after filtering the outputs; calculating the total carbon emissions corresponding to the qualified ratios, selecting the qualified ratio with the lowest total carbon emissions as the target ECC ratio, and outputting it. This invention improves the efficiency and stability of ratio design, significantly reduces carbon emissions while ensuring the tensile performance of ECC, and achieves multi-objective optimization and green design, with good engineering application prospects and promotional value.

[0077] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0078] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0079] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for ECC allocation optimization based on continuous conditional generative adversarial networks, characterized in that, The ECC allocation optimization method based on continuous conditional generative adversarial networks includes: A conditional generative adversarial network is constructed, which includes a generator and a discriminator. Set a target performance condition vector, obtain a random noise vector, input the target performance condition vector and the random noise vector into the generator, and output a generated ratio vector. Construct a ratio-performance database, obtain a true ratio vector based on the ratio-performance database, and output the generated ratio vector, the true ratio vector, and the target performance condition vector to the discriminator, and output a first degree of authenticity and a second degree of authenticity. The discriminator loss and generator loss are calculated based on the first and second truth levels. The conditional generative adversarial network is then optimized based on the discriminator loss and the generator loss to obtain the trained network. The quality of the generated matchings of the trained network is evaluated. If the quality evaluation of the generated matchings passes, the target network is obtained. Obtain the current target performance conditions of ECC, input the current target performance conditions into the target network, generate multiple candidate configurations, calculate the limit performance of each candidate configuration, and select multiple qualified configurations from all the candidate configurations based on all the limit performances. Calculate the total carbon emissions corresponding to each of the qualified ratios, select the qualified ratio with the lowest total carbon emissions from all the qualified ratios as the target ECC ratio and output it.

2. The ECC allocation optimization method based on continuous conditional generative adversarial networks according to claim 1, characterized in that, The step of inputting the target performance condition vector and the random noise vector into the generator and outputting a generated ratio vector specifically includes: The random noise vector is input into the feature extraction layer of the generator to extract features and obtain the first current hidden layer features. The target performance condition vector is input into the conditional mapping network of the generator for processing, generating the first feature linear modulation parameter and the second feature linear modulation parameter: ； in, and This represents the first characteristic linear modulation parameter and the second characteristic linear modulation parameter. This represents the conditional mapping network. Represents the target performance condition vector. , The target performance condition vector is represented by the limit strain. and ultimate stress composition; The first current hidden layer feature, the first feature linear modulation parameter, and the second feature linear modulation parameter are input into the linear feature adjustment layer of the generator. The current hidden layer feature is linearly modulated using the first feature linear modulation parameter and the second feature linear modulation parameter to obtain the first modulation feature. ； in, This represents the first current hidden layer feature. This represents the first modulation feature output after passing through the linear feature adjustment layer; After the modulation features are processed by the activation function of the generator, the generated ratio vector is output.

3. The ECC allocation optimization method based on continuous conditional generative adversarial networks according to claim 1, characterized in that, The construction of the proportion-performance database, and the acquisition of the true proportion vector based on the proportion-performance database, specifically includes: Construct a ratio-performance database, and perform data cleaning, outlier removal, missing value imputation and normalization on the real ECC ratio and performance data in the ratio-performance database to obtain the target database; The real ECC ratios are obtained from the target database as the real ratio vectors used to train the conditional generative adversarial network.

4. The ECC allocation optimization method based on continuous conditional generative adversarial networks according to claim 3, characterized in that, The step of outputting the generated ratio vector, the true ratio vector, and the target performance condition vector to the discriminator, and outputting a first degree of truth and a second degree of truth, specifically includes: The generated ratio vector and the real ratio vector are output to the feature extraction layer of the discriminator for feature extraction to obtain the second current hidden layer feature. The target performance condition vector is input into the condition mapping network of the discriminator for processing to generate the first feature linear modulation parameter and the second feature linear modulation parameter. The second current hidden layer feature, the first feature linear modulation parameter, and the second feature linear modulation parameter are input into the linear feature adjustment layer of the generator. The current hidden layer feature is linearly modulated by the first feature linear modulation parameter and the second feature linear modulation parameter to obtain the second modulation feature. The second modulation feature is subjected to dimensionality reduction and quantization processing through the preset layer structure and activation function of the discriminator, and the first degree of authenticity and the second degree of authenticity are output.

5. The ECC allocation optimization method based on continuous conditional generative adversarial networks according to claim 1, characterized in that, The discriminator loss is Wasserstein-GP loss, and the generator loss is Wasserstein-BD loss; The step of calculating the discriminator loss and generator loss based on the first and second truth levels, and optimizing the conditional generative adversarial network based on the discriminator loss and generator loss to obtain the trained network specifically includes: The Wasserstein-GP loss is calculated based on the first and second degrees of truth: ； in, This represents the Wasserstein-GP loss. This represents a weighted average of all possible values ​​of a random variable. Indicates the degree of authenticity of the first statement. Indicates the second degree of truth or falsehood. This represents the generated ratio vector. This represents the true proportion vector. This represents the target performance condition vector. Indicates the gradient penalty weights. Represents the gradient penalty term. , This represents the linear interpolation between the generated proportion vector and the actual proportion vector. Indicates to Find the gradient. Denotes the Euclidean norm; The Wasserstein-BD loss is calculated based on the first and second degrees of truth: ； in, This represents the Wasserstein-BD loss. BD represents the boundary loss weights. , This represents a modified linear function. and These represent the upper and lower bounds of the range for each ratio parameter, respectively, and N represents the sample size. Indicates the first The generated matching vector for each sample.

6. The ECC allocation optimization method based on continuous conditional generative adversarial networks according to claim 3, characterized in that, The step of evaluating the quality of the generated matchings in the trained network, and obtaining the target network if the quality evaluation is passed, specifically includes: The quality of the generated matchmaking by the trained network is evaluated using the target database to obtain the single-point bias and average dispersion of the generated matchmaking: ； ； in, This indicates the single-point deviation. This represents the average dispersion. This represents the average strain of the current generated mix proportion. This represents the average stress of the current generated mix proportion. This refers to the strain labels in the target database. This refers to the stress labels in the target database. Indicates the number of samples. Indicates the ordinal number of the sample. This represents the i-th strain in the current generated mix. This represents the i-th stress in the current generated mix proportion; If both the single-point deviation and the average dispersion are less than their respective thresholds, the quality assessment of the generated ratio is passed, and the target network is obtained.

7. The ECC allocation optimization method based on continuous conditional generative adversarial networks according to claim 1, characterized in that, The qualified mix proportions include: cement content, fly ash content, silica fume content, blast furnace slag content, the ratio of sand content to cementitious material content, the ratio of water content to cementitious material content, and fiber volume fraction; The calculation of the total carbon emissions corresponding to the qualified ratio specifically includes: Obtain the amount of each material in the qualified formula, and calculate the total carbon emissions corresponding to the qualified formula based on the amount of each material in the qualified formula: ； in, This represents the total carbon emissions. This indicates the amount of each material used. Carbon emission factor per unit material for each material This indicates the type number of the current material.

8. An ECC allocation optimization system based on continuous conditional generative adversarial networks, characterized in that, The ECC allocation optimization system based on continuous conditional generative adversarial networks includes: A network model building module is used to construct a conditional generative adversarial network, wherein the conditional generative adversarial network includes a generator and a discriminator; A ratio calculation module is used to set a target performance condition vector, obtain a random noise vector, input the target performance condition vector and the random noise vector into the generator, and output a ratio calculation vector. The authenticity calculation module is used to construct a ratio-performance database, obtain the true ratio vector according to the ratio-performance database, and output the generated ratio vector, the true ratio vector and the target performance condition vector to the discriminator, and output the first authenticity degree and the second authenticity degree. The network training optimization module is used to calculate the discriminator loss and generator loss based on the first degree of truth and the second degree of truth, and to optimize the conditional generative adversarial network based on the discriminator loss and the generator loss to obtain the trained network. The network quality evaluation module is used to evaluate the quality of the generated configurations of the trained network. If the quality evaluation of the generated configurations passes, the target network is obtained. The qualified mix generation module is used to obtain the current target performance conditions of ECC, input the current target performance conditions into the target network, generate multiple candidate mixes, calculate the limit performance of each candidate mix, and select multiple qualified mixes from all the candidate mixes based on all the limit performances. The carbon emission assessment module is used to calculate the total carbon emissions corresponding to each of the qualified ratios, select the qualified ratio with the lowest total carbon emissions from all the qualified ratios as the target ECC ratio, and output it.

9. A terminal, characterized in that, The terminal includes: a memory, a processor, and an ECC ratio optimization program based on a continuous conditional generative adversarial network stored in the memory and executable on the processor. When the ECC ratio optimization program based on the continuous conditional generative adversarial network is executed by the processor, it implements the steps of the ECC ratio optimization method based on a continuous conditional generative adversarial network as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an ECC ratio optimization program based on a continuous conditional generative adversarial network (CGA). When the ECC ratio optimization program based on a CGA is executed by a processor, it implements the steps of the ECC ratio optimization method based on a CGA as described in any one of claims 1-7.

Images