An intelligent prediction method and system for a dangerous chemical loading and unloading leakage diffusion area
By constructing a hazardous materials loading and unloading leakage and diffusion prediction model using FLUNT software and a multi-channel convolutional neural network, the problems of slow speed and accuracy in hazardous materials leakage and diffusion area delineation were solved, achieving efficient and accurate hazardous area delineation and supporting rapid rescue.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA JILIANG UNIV
- Filing Date
- 2024-12-02
- Publication Date
- 2026-05-08
AI Technical Summary
Existing technologies are slow in delineating the diffusion area of hazardous chemical leaks, and the prediction results depend on the accuracy of the model. They also have difficulties in handling nonlinear and complex data, resulting in low rescue efficiency.
Using FLUNT software combined with a multi-channel convolutional neural network, experimental simulations were conducted to simulate the factors affecting the diffusion range, a database of hazardous material leakage and diffusion was established, and a hazardous material loading and unloading leakage and diffusion prediction model was constructed through training and testing of the multi-channel convolutional neural network. Performance was evaluated using evaluation indicators.
It enabled the modeling and objective prediction of the spread of hazardous materials, improved the timeliness of rescue operations, and reduced casualties and property losses.
Smart Images
Figure CN119692539B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of diffusion area delineation technology, specifically to an intelligent prediction method and system for the diffusion area of hazardous chemical loading and unloading leaks. Background Technology
[0002] Hazardous chemicals are flammable and explosive, and proper delineation of the spill area after a hazardous chemical leak can facilitate the orderly conduct of rescue operations.
[0003] The traditional method used to delineate the diffusion zone of hazardous material leaks is numerical simulation. This method uses diffusion models to calculate the concentration of hazardous materials in different areas, thus defining the diffusion zone. However, numerical simulation suffers from drawbacks, including slow computation, reliance on model accuracy for predictions, and difficulty handling nonlinear and complex data. Consequently, it often results in slow and inaccurate zone delineation, leading to delays in rescue operations or low rescue efficiency. Summary of the Invention
[0004] The purpose of this invention is to provide an intelligent prediction method for the leakage and diffusion area of hazardous chemical loading and unloading, so as to solve the technical problems of slow calculation, prediction results depending on the accuracy of the model, and difficulties in handling nonlinear and complex data in the existing technology.
[0005] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution:
[0006] A method for intelligent prediction of leakage and diffusion areas during the loading and unloading of hazardous chemicals, characterized by the following steps:
[0007] Based on FLUNT software and the internal and external environment of the hazardous materials loading and unloading area, the factors that affect the spread of hazardous materials are identified and marked as factors affecting the spread.
[0008] Using FLUNT software, based on the values of factors affecting the diffusion range, an experimental simulation of the hazardous material leakage and diffusion process was conducted to obtain the hazardous material diffusion distance.
[0009] Establish a database of hazardous material leaks and spreads, and collect and store hazardous material spread distance data composed of the values of factors affecting the spread range and the spread distance of hazardous materials;
[0010] The hazardous materials leakage and diffusion data in the hazardous materials leakage and diffusion database are processed and divided into a training set for training multi-channel convolutional neural networks and a test set for testing multi-channel convolutional neural networks.
[0011] Based on the training set and test set, a multi-channel convolutional neural network is trained and tested to obtain a dangerous goods loading and unloading leakage diffusion prediction model for predicting the diffusion range of dangerous goods.
[0012] The performance of the dangerous goods loading and unloading leakage and diffusion prediction model in predicting the diffusion range is evaluated using preset evaluation indicators.
[0013] As a preferred embodiment of the present invention, the factors affecting the diffusion range include the size of the leakage orifice, natural wind speed, ground roughness, leakage height, and atmospheric stability.
[0014] As a preferred embodiment of the present invention, the multi-channel convolutional neural network is composed of multi-channel convolutional layers, pooling layers, fusion layers and fully connected layers.
[0015] As a preferred embodiment of the present invention, the method for processing hazardous material diffusion data in a hazardous material leakage and diffusion database includes:
[0016] When processing hazardous material spill and diffusion data in the hazardous material spill and diffusion database, missing values are filled using the K-nearest neighbor imputation method. The formula for the K-nearest neighbor imputation method is as follows:
[0017]
[0018] In the formula, New Value represents the value that needs to be filled in for missing values, x i This represents the K nearest neighbor observations of a missing value, where K represents the number of the most similar observations in the space.
[0019] After missing values are filled in, the hazardous materials dissemination data are normalized using the following formula:
[0020]
[0021] In the formula, X represents the hazardous materials dissemination data after missing values have been filled in. min It is the minimum value of the characteristic column belonging to X in the hazardous materials diffusion data, X max X is the maximum value of the characteristic column belonging to X in the hazardous materials diffusion data. norm It is the data value after X is normalized.
[0022] As a preferred embodiment of the present invention, each channel in the multi-channel convolutional layer performs an independent convolution operation on each diffusion range influencing factor of the normalized hazardous materials diffusion data using one-dimensional convolution. The calculation formula of the multi-channel convolutional layer is as follows:
[0023]
[0024] In the formula, Ik (t) represents the value of the diffusion range influencing factor sequence of the k-th channel input at time step t, S c,k (m) represents the weight of the convolution kernel corresponding to the c-th output channel at position m in the k-th channel, (I*S)(c,t) represents the output value of the c-th output channel after the convolution operation at time step t, C in This indicates the number of output channels, and c represents the index of the output channel.
[0025] In a preferred embodiment of the present invention, the pooling layer employs average pooling operation, and the calculation formula for average pooling is as follows:
[0026]
[0027] In the formula, P(u) represents the value of the pooled feature at time step u, |window| represents the size of the pooling window, and I(u+i) represents the average value of all values within the pooling window. All values within the pooling window are composed of (I*S)(c,t) obtained after the convolution operation.
[0028] As a preferred embodiment of the present invention, the fusion layer adopts a splicing and fusion method, and the calculation formula of the fusion layer is:
[0029]
[0030] In the formula, F represents the output value after concatenation. i C represents the combined value of each channel P(u) output after the pooling operation. in This indicates the number of output channels of a multi-channel convolutional layer.
[0031] As a preferred embodiment of the present invention, the calculation formula of the fully connected layer is:
[0032] A = f(Z) = f(W*H + b);
[0033] In the formula, A represents the output of the fully connected layer after passing through the activation function, f represents the activation function, Z represents the output of the fully connected layer before passing through the activation function, W represents the matrix weights of the fully connected layer, H represents the one-dimensional vector of the fusion layer output after the input of the fully connected layer is expanded, and b is the bias vector.
[0034] The activation function f is: In the formula, x is the input of ReLU.
[0035] As a preferred embodiment of the present invention, the evaluation index includes the adjusted goodness of fit R. 2 best Mean Absolute Error (MAE) and AIC criterion, among which,
[0036] Adjusted fit R 2 best The calculation formula is:
[0037] The formula for calculating the Mean Absolute Error (MAE) is as follows: The formula for calculating the AIC criterion is: AIC = -2In(L) + 2a;
[0038] In the formula, R 2 The value represents the goodness of fit before adjustment, v represents the number of data samples in the test set, w represents the number of data parameters in the test set, and q represents the goodness of fit before adjustment. i This represents the actual diffusion distance of the i-th data sample in the test set; denoted as the predicted diffusion distance of the i-th data sample in the test set, L represents the maximum likelihood function value under given data, and a represents the number of parameters in the hazardous materials loading and unloading leakage diffusion prediction model;
[0039] Adjusted R 2 A negative value is acceptable, indicating that the predictive power of the model used is less than that of the baseline model using the mean variable. (Similar to R...) 2 Compared to the adjusted R 2 The number of model variables was changed, thus avoiding the problem of inflated numbers caused by an increase in the number of dependent variables.
[0040] The loss function MSE calculation formula for the dangerous goods loading and unloading leakage and diffusion prediction model is as follows:
[0041]
[0042] In the formula, MSE is the neural loss function value, n represents the number of data samples in the training set, and y i This represents the actual diffusion distance of the i-th data sample in the training set. This represents the predicted diffusion distance of the i-th data sample in the training set.
[0043] As a preferred embodiment of the present invention, the present invention provides an intelligent prediction system for hazardous materials loading and unloading leakage and diffusion areas, applied to an intelligent prediction method for hazardous chemical loading and unloading leakage and diffusion areas. The system includes:
[0044] The influencing factor determination unit is used to determine the factors that affect the spread of dangerous goods based on FLUNT software and the internal and external environment of the dangerous goods loading and unloading area, and mark them as factors affecting the spread.
[0045] The diffusion experiment simulation unit is used to conduct experimental simulations of the diffusion process of hazardous materials leakage using FLUNT software based on the values of factors affecting the diffusion range, and to obtain the diffusion distance of hazardous materials.
[0046] The data collection and storage unit is used to establish a hazardous materials leakage and diffusion database, and to collect and store hazardous materials diffusion distance data composed of the values of factors affecting the diffusion range and the hazardous materials diffusion distance.
[0047] The data processing unit is used to process the hazardous materials leakage and diffusion data in the hazardous materials leakage and diffusion database, and divide it into a training set for training multi-channel convolutional neural networks and a test set for testing multi-channel convolutional neural networks.
[0048] The prediction model building unit is used to train and test the multi-channel convolutional neural network based on the training set and the test set to obtain a prediction model for the leakage and diffusion of hazardous materials during loading and unloading, which is used to predict the diffusion range of hazardous materials.
[0049] The prediction model evaluation unit is used to evaluate the performance of the dangerous goods loading and unloading leakage and diffusion prediction model in predicting the diffusion range using preset evaluation indicators.
[0050] Compared with the prior art, the present invention has the following advantages:
[0051] This invention uses FLUNT software to simulate hazardous material leaks, establishes a dataset and a test set for training the model, employs a multi-channel convolutional neural network model for intelligent prediction of hazardous material leaks, and uses a network search method for hyperparameter tuning to achieve the goal of rationally delineating hazardous areas, thereby effectively carrying out emergency response and rescue activities, reducing casualties, and minimizing property damage. Attached Figure Description
[0052] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.
[0053] Figure 1 A flowchart of the intelligent prediction method for leakage and diffusion areas during hazardous chemical loading and unloading provided in this embodiment of the invention;
[0054] Figure 2 A block diagram of an intelligent prediction system for the leakage and diffusion area of hazardous chemicals during loading and unloading, provided in an embodiment of the present invention. Detailed Implementation
[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0056] like Figure 1 As shown, this invention provides an intelligent prediction method for the leakage and diffusion area during the loading and unloading of hazardous chemicals, characterized by the following steps:
[0057] Based on FLUNT software and the internal and external environment of the hazardous materials loading and unloading area, the factors that affect the spread of hazardous materials are identified and marked as factors affecting the spread.
[0058] Using FLUNT software, based on the values of factors affecting the diffusion range, an experimental simulation of the hazardous material leakage and diffusion process was conducted to obtain the hazardous material diffusion distance.
[0059] Establish a database of hazardous material leaks and spreads, and collect and store hazardous material spread distance data composed of the values of factors affecting the spread range and the spread distance of hazardous materials;
[0060] The hazardous materials leakage and diffusion data in the hazardous materials leakage and diffusion database are processed and divided into a training set for training multi-channel convolutional neural networks and a test set for testing multi-channel convolutional neural networks.
[0061] Based on the training set and test set, a multi-channel convolutional neural network is trained and tested to obtain a dangerous goods loading and unloading leakage diffusion prediction model for predicting the diffusion range of dangerous goods.
[0062] The performance of the dangerous goods loading and unloading leakage and diffusion prediction model in predicting the diffusion range is evaluated using preset evaluation indicators.
[0063] In this invention, the values of the factors influencing the diffusion range determine the diffusion distance of hazardous materials. When learning a multi-channel convolutional neural network, it is necessary to use the values of the factors influencing the diffusion range and the diffusion distance of hazardous materials. By using a multi-channel convolutional neural network for learning, the goal of predicting the diffusion distance of hazardous materials based on the values of the factors influencing the diffusion range can be achieved. This realizes the modeling and objective prediction of the diffusion distance of hazardous materials, efficiently obtains the diffusion range of hazardous materials, and meets the timeliness requirements of hazardous materials rescue.
[0064] The factors affecting the diffusion range include the size of the leak hole, natural wind speed, ground roughness, leak height, and atmospheric stability. These factors can be added or removed as needed in actual use.
[0065] The multichannel convolutional neural network consists of multichannel convolutional layers, pooling layers, fusion layers, and fully connected layers.
[0066] Methods for processing hazardous materials diffusion data in a hazardous materials spill and diffusion database include:
[0067] When processing hazardous material spill and diffusion data in the hazardous material spill and diffusion database, missing values are filled using the K-nearest neighbor imputation method. The formula for the K-nearest neighbor imputation method is as follows:
[0068]
[0069] In the formula, New Value represents the value that needs to be filled in for missing values, x i This represents the K nearest neighbor observations of a missing value, where K represents the number of the most similar observations in the space.
[0070] After missing values are filled in, the hazardous materials dissemination data are normalized using the following formula:
[0071]
[0072] In the formula, X represents the hazardous materials dissemination data after missing values have been filled in. min It is the minimum value of the characteristic column belonging to X in the hazardous materials diffusion data, X max X is the maximum value of the characteristic column belonging to X in the hazardous materials diffusion data. norm It is the data value after X is normalized.
[0073] In the multi-channel convolutional layer, each channel uses one-dimensional convolution to independently perform convolution operations on each diffusion range influencing factor of the normalized hazardous materials diffusion data. The calculation formula for the multi-channel convolutional layer is as follows:
[0074]
[0075] In the formula, I k (t) represents the value of the diffusion range influencing factor sequence of the k-th channel input at time step t, S c,k (m) represents the weight of the convolution kernel corresponding to the c-th output channel at position m in the k-th channel, (I*S)(c,t) represents the output value of the c-th output channel after the convolution operation at time step t, C in This indicates the number of output channels, and c represents the index of the output channel.
[0076] The pooling layer employs average pooling, and the formula for average pooling is as follows:
[0077]
[0078] In the formula, P(u) represents the value of the pooled feature at time step u, |window| represents the size of the pooling window, and I(u+i) represents the average value of all values within the pooling window. All values within the pooling window are composed of (I*S)(c,t) obtained after the convolution operation.
[0079] The fusion layer uses a splicing and fusion method, and the calculation formula for the fusion layer is as follows:
[0080]
[0081] In the formula, F represents the output value after concatenation. i C represents the combined value of each channel P(u) output after the pooling operation. in This indicates the number of output channels of a multi-channel convolutional layer.
[0082] The calculation formula for the fully connected layer is:
[0083] A = f(Z) = f(W*H + b);
[0084] In the formula, A represents the output of the fully connected layer after passing through the activation function, f represents the activation function, Z represents the output of the fully connected layer before passing through the activation function, W represents the matrix weights of the fully connected layer, H represents the one-dimensional vector of the fusion layer output after the input of the fully connected layer is expanded, and b is the bias vector.
[0085] The activation function f is: In the formula, x is the input of ReLU.
[0086] The evaluation metrics include the adjusted goodness of fit R. 2 best Mean Absolute Error (MAE) and AIC criterion, among which,
[0087] Adjusted fit R 2 best The calculation formula is:
[0088] The formula for calculating the Mean Absolute Error (MAE) is as follows:
[0089] The formula for calculating the AIC criterion is: AIC = -2In(L) + 2a;
[0090] In the formula, R 2 The value represents the goodness of fit before adjustment, v represents the number of data samples in the test set, w represents the number of data parameters in the test set, and q represents the goodness of fit before adjustment. i This represents the actual diffusion distance of the i-th data sample in the test set; denoted as the predicted diffusion distance of the i-th data sample in the test set, L represents the maximum likelihood function value under given data, and a represents the number of parameters in the hazardous materials loading and unloading leakage diffusion prediction model;
[0091] The adjusted fit R in this invention 2 best It can be negative, indicating that the predictive power of the model used is not as good as the baseline model using the mean variable, and is related to the unadjusted R-squared. 2 Compared to the adjusted goodness of fit R 2 best By changing the number of model variables, the problem of artificially inflated model performance caused by an increase in the number of dependent variables is avoided. This allows for more accurate feedback on model performance, providing more precise guidance for parameter tuning and ultimately leading to the optimal model. Therefore, the adjusted fit Rfit is higher. 2 best This ensures that the prediction model for leakage and diffusion during the loading and unloading of hazardous materials is at its optimal performance.
[0092] In this invention, MAE is an indicator used to measure the accuracy of a prediction model. It is the average of the absolute values of the differences between the predicted and actual values. The smaller the MAE value, the more accurate the model's prediction. MAE can reflect the model's performance in the presence of noisy data because it is less sensitive to outliers and is not disturbed by noisy data. Furthermore, MAE directly reflects the model's prediction accuracy and is not masked by the positive or negative value of the prediction error.
[0093] The AIC criterion in this invention is a standard used to measure the goodness of fit of a statistical model. It is based on the concept of information entropy and aims to balance the complexity of the model and the goodness of the model fit to the data.
[0094] Adjusted fit R 2 best By combining the mean absolute error (MAE) and the analogy indices (AIC) criterion, the accuracy and complexity of the model can be evaluated using both MAE and AIC criteria. This approach avoids the problem of inflated performance indicators, restores the most objective evaluation indicator values, and achieves the most objective evaluation of the model. This serves as the benchmark for optimization, ultimately ensuring that the hazardous materials loading and unloading leakage and diffusion prediction model receives the best guidance and achieves the best performance.
[0095] The loss function MSE calculation formula for the dangerous goods loading and unloading leakage and diffusion prediction model is as follows:
[0096]
[0097] In the formula, MSE is the neural loss function value, n represents the number of data samples in the training set, and y i This represents the actual diffusion distance of the i-th data sample in the training set. This represents the predicted diffusion distance of the i-th data sample in the training set.
[0098] like Figure 2 As shown, this invention provides an intelligent prediction system for hazardous materials loading and unloading leakage and diffusion areas, applied to an intelligent prediction method for hazardous chemical loading and unloading leakage and diffusion areas. The system includes:
[0099] The influencing factor determination unit is used to determine the factors that affect the spread of dangerous goods based on FLUNT software and the internal and external environment of the dangerous goods loading and unloading area, and mark them as factors affecting the spread.
[0100] The diffusion experiment simulation unit is used to conduct experimental simulations of the diffusion process of hazardous materials leakage using FLUNT software based on the values of factors affecting the diffusion range, and to obtain the diffusion distance of hazardous materials.
[0101] The data collection and storage unit is used to establish a hazardous materials leakage and diffusion database, and to collect and store hazardous materials diffusion distance data composed of the values of factors affecting the diffusion range and the hazardous materials diffusion distance.
[0102] The data processing unit is used to process the hazardous materials leakage and diffusion data in the hazardous materials leakage and diffusion database, and divide it into a training set for training multi-channel convolutional neural networks and a test set for testing multi-channel convolutional neural networks.
[0103] The prediction model building unit is used to train and test the multi-channel convolutional neural network based on the training set and the test set to obtain a prediction model for the leakage and diffusion of hazardous materials during loading and unloading, which is used to predict the diffusion range of hazardous materials.
[0104] The prediction model evaluation unit is used to evaluate the performance of the dangerous goods loading and unloading leakage and diffusion prediction model in predicting the diffusion range using preset evaluation indicators.
[0105] This invention uses FLUNT software to simulate hazardous material leaks, establishes a dataset and a test set for training the model, employs a multi-channel convolutional neural network model for intelligent prediction of hazardous material leaks, and uses a network search method for hyperparameter tuning to achieve the goal of rationally delineating hazardous areas, thereby effectively carrying out emergency response and rescue activities, reducing casualties, and minimizing property damage.
[0106] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.
Claims
1. A method for intelligent prediction of leakage and diffusion areas during the loading and unloading of hazardous chemicals, characterized in that, Includes the following steps: Based on FLUENT software and the internal and external environment of the hazardous materials loading and unloading area, the factors that affect the spread of hazardous materials are identified and marked as factors affecting the spread. Using FLUENT software, based on the values of factors affecting the diffusion range, an experimental simulation of the hazardous material leakage and diffusion process was conducted to obtain the hazardous material diffusion distance. Establish a hazardous materials leakage and diffusion database to collect and store hazardous materials diffusion distance data, which consists of the values of factors affecting the diffusion range and the hazardous materials diffusion distance. The hazardous materials spill and diffusion distance data in the hazardous materials spill and diffusion database are processed and divided into a training set for training multi-channel convolutional neural networks and a test set for testing multi-channel convolutional neural networks. Based on the training set and test set, a multi-channel convolutional neural network is trained and tested to obtain a dangerous goods loading and unloading leakage diffusion prediction model for predicting the diffusion range of dangerous goods. The performance of the dangerous goods loading and unloading leakage and diffusion prediction model in predicting the diffusion range is evaluated using preset evaluation indicators. In a multi-channel convolutional layer, each channel uses one-dimensional convolution to independently perform convolution operations on each diffusion range influencing factor of the normalized hazardous material diffusion distance data. The calculation formula for the multi-channel convolutional layer is as follows: ； In the formula, Indicates the first The sequence of factors influencing the diffusion range of each channel input at time step The value of , Indicates the first The convolution kernel corresponding to the nth output channel is in the nth... Location on each channel The weight, Indicates the first Each output channel at time step The output value obtained after the convolution operation, Indicates the number of output channels. Indicates the index of the output channel; The pooling layer uses average pooling, and the formula for average pooling is: ； In the formula, This indicates the pooled features at time step The value on, Indicates the size of the pooling window. This represents the average of all values within the pooling window, where all values are obtained after the convolution operation. constitute; The blending layer uses a splicing and blending method, and the calculation formula for the blending layer is: ； In the formula, This represents the concatenated output value. This represents the output channels after the pooling operation. The integrated value, This indicates the number of output channels of a multi-channel convolutional layer; The calculation formula for a fully connected layer is: ； In the formula, This represents the output of the fully connected layer after the activation function. This represents the activation function. This represents the output of the fully connected layer before the activation function. This represents the matrix weights of the fully connected layer. This represents a one-dimensional vector representing the expanded output of the fusion layer after input to the fully connected layer. It is the bias vector; Activation function for: In the formula, for Input.
2. The intelligent prediction method for the leakage and diffusion area during hazardous chemical loading and unloading according to claim 1, characterized in that: The factors affecting the diffusion range include the size of the leak orifice, natural wind speed, ground roughness, leak height, and atmospheric stability.
3. The intelligent prediction method for the leakage and diffusion area during hazardous chemical loading and unloading according to claim 1, characterized in that: The multichannel convolutional neural network consists of multichannel convolutional layers, pooling layers, fusion layers, and fully connected layers.
4. The intelligent prediction method for the leakage and diffusion area during hazardous chemical loading and unloading according to claim 1, characterized in that: Methods for processing hazardous material diffusion distance data in a hazardous material spill and diffusion database include: When processing hazardous material spill and diffusion distance data in the hazardous material spill and diffusion database, missing values are filled using the K-nearest neighbor imputation method. The formula for the K-nearest neighbor imputation method is as follows: ； In the formula, This indicates the value to fill in for missing values. Indicates missing values The nearest neighbor observation, This represents the number of the most similar observations in space; After the missing values are filled in, the hazardous material diffusion distance data is normalized. The normalization formula is as follows: ； In the formula, This represents the hazardous materials dispersal distance data after missing values have been filled in. It is the minimum value of X in the characteristic column of the hazardous materials diffusion distance data. It is the maximum value of X in the characteristic column belonging to the same category in the hazardous materials diffusion distance data. It is the data value after X is normalized.
5. The intelligent prediction method for the leakage and diffusion area during hazardous chemical loading and unloading according to claim 1, characterized in that: The evaluation metrics include the adjusted goodness of fit. Mean Absolute Error (MAE) and AIC criterion, among which, Adjusted fit The calculation formula is: ; The formula for calculating the Mean Absolute Error (MAE) is as follows: ; The calculation formula for the AIC criterion is as follows: ; In the formula, This indicates the goodness of fit before adjustment. This indicates the number of data samples in the test set. This indicates the number of data parameters in the test set. This represents the actual diffusion distance of the i-th data sample in the test set; This represents the predicted diffusion distance of the i-th data sample in the test set. This represents the maximum likelihood function value given the data. Indicates the number of parameters in the hazardous materials loading and unloading leakage and diffusion prediction model; Adjusted A negative value indicates that the predictive power of the model used is inferior to that of the baseline model using the mean variable. Compared to the adjusted The number of model variables was changed, thus avoiding the problem of inflated numbers caused by an increase in the number of dependent variables; The loss function MSE calculation formula for the dangerous goods loading and unloading leakage and diffusion prediction model is as follows: ； In the formula, The neural loss function value. This indicates the number of data samples in the training set. This represents the actual diffusion distance of the i-th data sample in the training set. This represents the predicted diffusion distance of the i-th data sample in the training set.
6. An intelligent prediction system for the leakage and diffusion area during the loading and unloading of hazardous materials, characterized in that, The intelligent prediction method for the leakage and diffusion area of hazardous chemical loading and unloading, as described in any one of claims 1-5, comprises: The influencing factor determination unit is used to determine the factors that affect the spread of dangerous goods based on FLUENT software and the internal and external environment of the dangerous goods loading and unloading area, and mark them as factors affecting the spread range. The diffusion experiment simulation unit is used to conduct experimental simulations of the diffusion process of hazardous materials leakage using FLUENT software based on the values of factors affecting the diffusion range, and to obtain the diffusion distance of hazardous materials. The data collection and storage unit is used to establish a hazardous materials leakage and diffusion database, and to collect and store hazardous materials diffusion distance data composed of the values of factors affecting the diffusion range and the hazardous materials diffusion distance. The data processing unit is used to process the hazardous material diffusion distance data in the hazardous material leakage and diffusion database, and divide it into a training set for training a multi-channel convolutional neural network and a test set for testing a multi-channel convolutional neural network. The prediction model building unit is used to train and test the multi-channel convolutional neural network based on the training set and the test set to obtain a prediction model for the leakage and diffusion of hazardous materials during loading and unloading, which is used to predict the diffusion range of hazardous materials. The prediction model evaluation unit is used to evaluate the performance of the dangerous goods loading and unloading leakage and diffusion prediction model in predicting the diffusion range using preset evaluation indicators.