CN-120597760-B - Method for dynamically determining explosion safety distance based on deep learning model

CN120597760BCN 120597760 BCN120597760 BCN 120597760BCN-120597760-B

Abstract

The invention relates to the technical field of risk assessment and discloses a method for dynamically determining explosion safety distance based on a deep learning model, which comprises the following steps of S1, obtaining a plurality of groups of data through an engineering simulation platform and CFD software, constructing the plurality of groups of data into a training data set, S2, training the deep learning model through the training data set, obtaining the relation between input features and output features through the deep learning model, S3, obtaining real-time data of pressure, temperature, gas concentration and flow velocity in an environment at a corresponding time point through a data acquisition module, S4, carrying out numerical simulation on an explosion process of the created corresponding explosion scene model through the engineering simulation platform and the CFD software, S5, obtaining a predicted value through the deep learning model, S6, calculating the explosion safety distance through an explosion safety distance calculation formula, S7, repeating the steps of S3-S6, and dynamically determining the explosion safety distance in real time. The method can be applied to different explosion scenes, and can accurately acquire the safety distance in the explosion scenes in real time.

Inventors

NIU YIHUI
GONG YUKE
DU BINGSHU
WANG WENHE
MI HONGFU
LI ZIRAN
JIANG LINGUI
ZHANG YUZHUO

Assignees

重庆科技大学

Dates

Publication Date: 20260508
Application Date: 20250528

Claims (5)

1. The method for dynamically determining the explosion safety distance based on the deep learning model is characterized by comprising the following steps of: S1, creating different explosion scene models through an engineering simulation platform, carrying out numerical simulation on explosion processes under different scenes through CFD software to obtain multiple groups of data, wherein each group of data comprises space coordinates, time steps, local pressure values, shock wave propagation speeds, overpressure threshold values of specific positions, total energy released by explosion and dangerous range radius, and constructing the multiple groups of data into a training data set; S2, training a deep learning model through a training data set, wherein the deep learning model obtains the relations of input characteristic space coordinates, time steps, local pressure values, shock wave propagation speeds, overpressure threshold values of specific positions of output characteristics, total energy released by explosion and dangerous range radius; s2 comprises the following steps: s21, inputting a training data set into a deep learning model, wherein a first layer hidden layer, a second layer hidden layer and a full-connection layer in the deep learning model automatically obtain corresponding weights and biases through the training data set, simultaneously inputting input features in the training data set into the deep learning model in batches in a single group mode, extracting features through the first layer hidden layer, the second layer hidden layer and the full-connection layer to obtain three feature tensors, wherein each feature tensor corresponds to time sequence or space features of a group of input data, and each feature tensor is a row matrix structure distributed according to output features and feature dimensions; s22, processing single features in the corresponding feature tensor through a softmax layer in the deep learning model, calculating to obtain the probability of the single features in the feature tensor relative to the sum of the single features in the feature tensor, and selecting the single feature with the highest probability as a predicted value; s23, calculating the deviation degree between the predicted value and the output characteristics in the training data set through a mean square error loss function, and optimizing the weights and the biases corresponding to the first hidden layer, the second hidden layer and the full-connection layer in the deep learning model; S24, repeating the steps S21-S24, and automatically stopping training when the maximum iteration times are reached; S3, acquiring real-time data of pressure, temperature, gas concentration and flow rate in the environment at a corresponding time point through a data acquisition module; S4, processing the acquired real-time data of pressure, temperature, gas concentration and flow velocity in the environment and geometric parameters of the explosive objects through an engineering simulation platform to create a corresponding explosion scene model, and carrying out numerical simulation on the explosion process of the created corresponding explosion scene model through CFD software to obtain space coordinates, time steps, local pressure values and shock wave propagation speeds; s5, inputting the obtained space coordinates, time step, local pressure value and shock wave propagation speed into a trained deep learning model, and obtaining a predicted overpressure threshold value of a specific position, total energy released by explosion and a dangerous range radius through the deep learning model; S6, calculating the obtained overpressure threshold, total energy released by explosion and the radius of a dangerous range through an explosion safety distance calculation formula to obtain the explosion safety distance of the corresponding time point; and S7, repeating the steps S3-S6, and transmitting the acquired information to the engineering simulation platform by the data acquisition module according to a preset frequency, and dynamically determining the explosion safety distance in real time.
2. The method for dynamically determining explosion safety distance based on deep learning model according to claim 1, wherein S1 comprises the following steps: S11, defining a solid boundary, an opening boundary and an environment boundary in an engineering simulation platform, setting initial pressure, temperature, gas concentration and flow rate for an explosion region, defining geometric parameters of the explosion region, and creating a corresponding explosion scene model; S12, carrying out grid division on the explosion scene model through Fluent Meshing tools in CFD software; S13, setting a turbulence model in a Fluent tool in CFD software, simulating an electric spark ignition process in a component transmission mode, and setting a time step and a calculation time length; S14, starting a Fluent simulation program, and carrying out numerical simulation on the explosion process in the scene; s15, obtaining simulation results including a pressure distribution cloud chart and a shock wave propagation path chart through a Fluent tool in CFD software; S16, deriving a pressure distribution cloud chart and a shock wave propagation path chart, extracting the values of the pressure distribution cloud chart and the shock wave propagation path chart through an image analysis and program reading function, acquiring space coordinates according to grid nodes of an explosion scene model, acquiring local pressure values according to the pressure distribution cloud chart and the space coordinates, acquiring a shock wave propagation speed according to the shock wave propagation path chart, and acquiring an overpressure threshold value, total energy released by explosion and a dangerous range radius from a CFD software result report; S17, repeating S11-S16, creating different explosion scene models, and obtaining a plurality of groups of data, wherein the space coordinates, the local pressure value, the shock wave propagation speed, the time step, the overpressure threshold value, the total energy released by explosion and the dangerous range radius in each group of data are used as labels, and the plurality of groups of data together form a training data set.
3. The method for dynamically determining explosion safety distance based on deep learning model according to claim 1, wherein S21 further comprises the steps of: s211, a deep learning model comprises a first layer of hidden layers, a second layer of hidden layers, a full-connection layer and a softmax layer, wherein the first layer of hidden layers comprises 16 LSTM units, and the second layer of hidden layers comprises 32 LSTM units; And S212, inputting a training data set into a deep learning model, wherein a first layer of hidden layer, a second layer of hidden layer and a full-connection layer in the deep learning model automatically obtain corresponding weights and biases through the training data set, inputting features in the training data set into the deep learning model in batches in a single group mode, extracting features through the first layer of hidden layer, the second layer of hidden layer and the full-connection layer, and obtaining three feature tensors, wherein each feature tensor corresponds to time sequence or space features of a group of samples, and each feature tensor is a row-column matrix structure distributed according to the samples and feature dimensions.
4. The method for dynamically determining explosion safety distance based on deep learning model according to claim 1, wherein S5 further comprises the steps of: S51, inputting the obtained space coordinates, time step, local pressure value, shock wave propagation velocity as input features into the trained deep learning model, as shown in formula (1), (1), In the formula (1), The input characteristics are represented as such, 、、 The respective coordinates of the space are provided in each case, Is the step of the time that is required, Is the value of the local pressure and, Is the shock wave propagation velocity; s52, extracting feature tensors of input features by the first layer hidden layer, the second layer hidden layer and the full connection layer in the deep learning model, wherein the feature tensors are shown in formulas (2), (3), (4), (5), (6), (7), (8) and (9), (2), (3), (4), (5), (6), (7), (8), (9), In the formulas (2), (3), (4), (5), (6), (7), (8) and (9), Is an input feature that is used to determine the input, 、 The output of the first layer hidden layer and the output of the second layer hidden layer are respectively, Is the output of the full-connection layer, Is a hidden state as the first layer, is set to a zero vector, Is a forgetful door which is a door which is left, Is an input door, and is provided with a plurality of input doors, Is a candidate state for use in the method, Is in the state of a cell and, Is an output door which is provided with a plurality of output doors, 、、、、 Respectively, a weight matrix is provided, 、、、、 Respectively offset; S53, processing single characteristics in the extracted corresponding characteristic tensor by a softmax layer in the deep learning model, calculating the probability of each single characteristic in the characteristic tensor relative to the sum of the single characteristics in the characteristic tensor, selecting the single characteristic with the maximum probability as a predicted value, as shown in a formula (10), (10), In the formula (10) of the present invention, Is the full connection layer A plurality of features; S54, repeating S53, calculating the probability of each single feature in the other two feature tensors relative to the sum of the single features in the feature tensors, and taking the single feature with the highest probability as the corresponding predicted value.
5. The method for dynamically determining explosion safety distance based on deep learning model according to claim 1, wherein S6 comprises the following steps: s61, calculating the obtained overpressure threshold, total energy released by explosion and dangerous range radius through an explosion safety distance calculation formula to obtain the explosion safety distance of the corresponding time point, wherein the explosion safety distance is shown in a formula (11), (11), In the formula (11), the color of the sample is, Is the safe distance to be used for the operation, Is the radius of the dangerous range, Is the total energy released by the explosion and, Is the over-pressure threshold value, Is a constant that is related to the environmental conditions, Is an index related to the type of explosion.

Description

Method for dynamically determining explosion safety distance based on deep learning model Technical Field The invention relates to the technical field of risk assessment, in particular to a method for dynamically determining an explosion safety distance based on a deep learning model. Background In industrial production, explosion events are extremely harmful to equipment, personnel and the environment. The safety distance determination in the existing explosion event depends on an empirical formula or static simulation or a single data source, but the safety distance is obtained only by the empirical formula or static simulation, the complex working condition in the actual situation and the environmental condition which changes in real time can be ignored, the finally obtained safety distance is easy to predict with low precision, and if the safety distance is only obtained by means of a single data source, such as a chemical diffusion path, the safety distance determination method is only suitable for a chemical industry park, and the application range is small. Disclosure of Invention In order to overcome the defects of the prior art, the invention aims to provide a method for dynamically determining the explosion safety distance based on a deep learning model, which can be applied to different explosion scenes and can accurately acquire the safety distance in the explosion scenes in real time, thereby assisting workers to make accurate safety coping decisions. The technical scheme adopted by the invention is that the method for dynamically determining the explosion safety distance based on the deep learning model comprises the following steps: S1, creating different explosion scene models through an engineering simulation platform, carrying out numerical simulation on explosion processes under different scenes through CFD software to obtain multiple groups of data, wherein each group of data comprises space coordinates, time steps, local pressure values, shock wave propagation speeds, overpressure threshold values of specific positions, total energy released by explosion and dangerous range radius, and constructing the multiple groups of data into a training data set; S2, training a deep learning model through a training data set, wherein the deep learning model obtains the relations of input characteristic space coordinates, time steps, local pressure values, shock wave propagation speeds, overpressure threshold values of specific positions of output characteristics, total energy released by explosion and dangerous range radius; S3, acquiring real-time data of pressure, temperature, gas concentration and flow rate in the environment at a corresponding time point through a data acquisition module; S4, processing the acquired real-time data of pressure, temperature, gas concentration and flow velocity in the environment and geometric parameters of the explosive objects through an engineering simulation platform to create a corresponding explosion scene model, and carrying out numerical simulation on the explosion process of the created corresponding explosion scene model through CFD software to obtain space coordinates, time steps, local pressure values and shock wave propagation speeds; s5, inputting the obtained space coordinates, time step, local pressure value and shock wave propagation speed into a trained deep learning model, and obtaining a predicted overpressure threshold value of a specific position, total energy released by explosion and a dangerous range radius through the deep learning model; S6, calculating the obtained overpressure threshold, total energy released by explosion and the radius of a dangerous range through an explosion safety distance calculation formula to obtain the explosion safety distance of the corresponding time point; and S7, repeating the steps S3-S6, and transmitting the acquired information to the engineering simulation platform by the data acquisition module according to a preset frequency, and dynamically determining the explosion safety distance in real time. A preferred embodiment of the present invention is characterized in that S1 comprises the steps of: S11, defining a solid boundary, an opening boundary and an environment boundary in an engineering simulation platform, setting initial pressure, temperature, gas concentration and flow rate for an explosion region, defining geometric parameters of the explosion region, and creating a corresponding explosion scene model; S12, carrying out grid division on the explosion scene model through Fluent Meshing tools in CFD software; S13, setting a turbulence model in a Fluent tool in CFD software, simulating an electric spark ignition process in a component transmission mode, and setting a time step and a calculation time length; S14, starting a Fluent simulation program, and carrying out numerical simulation on the explosion process in the scene; s15, obtaining simulation results including a pressure distribution clo