CN115481554A - Thermal diffusion digital twin model, temperature field real-time optimization control model and method in explosive fusion casting and curing process - Google Patents
Thermal diffusion digital twin model, temperature field real-time optimization control model and method in explosive fusion casting and curing process Download PDFInfo
- Publication number
- CN115481554A CN115481554A CN202211157180.5A CN202211157180A CN115481554A CN 115481554 A CN115481554 A CN 115481554A CN 202211157180 A CN202211157180 A CN 202211157180A CN 115481554 A CN115481554 A CN 115481554A
- Authority
- CN
- China
- Prior art keywords
- temperature field
- data
- explosive
- temperature
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
Abstract
The invention discloses a thermal diffusion digital twin model, a temperature field real-time optimization control model and a method in an explosive casting solidification process, wherein the thermal diffusion digital twin model is driven by constructed data, so that the real-time simulation of the temperature field in the explosive casting process can be realized, and the problem that the temperature field distribution in the whole solidification process cannot be quickly obtained in the prior art is solved; by constructing a temperature field optimization control model, real-time monitoring and future state prediction are carried out on a global temperature field based on a thermal diffusion digital twin model, and process parameters are adjusted according to the real-time monitoring and future state prediction, so that the temperature field control in the casting process is realized, and the problems of more forming defects, poor quality stability and the like caused by the constraints that the internal temperature field in the casting process cannot be monitored, the process parameters cannot be optimized according to the actual temperature field distribution and the like are solved. The method adopts an explosive fusion casting solidification process temperature field optimization control method, collects wall surface temperature data in real time, and regulates and controls technological parameters in the explosive fusion casting process in real time so as to realize temperature field control in the fusion casting solidification process.
Description
Technical Field
The invention belongs to the technical field of data analysis, and particularly relates to a thermal diffusion digital twin model, a temperature field real-time optimization control model and a method in an explosive fusion casting solidification process.
Background
The temperature field is used as the most important factor influencing the quality of fusion casting molding of the explosive, and the defects of shrinkage cavity, shrinkage porosity, cracks and the like are easily formed by improper control, so that the temperature field distribution of the fusion casting process of the explosive is monitored in real time, dynamic adjustment of process parameters is carried out according to the temperature field distribution, the generation of the defects of the explosive is reduced, and the method has important significance for improving the quality of the fusion casting process.
At present, methods for temperature field research mainly include an experimental method, a finite element numerical simulation method, a machine learning method and the like.
The experimental method comprises the following steps: under normal conditions, experimental study under the same working condition is the most effective and most accurate means for obtaining the dynamic change rule of the temperature field, but because the physical experiment temperature sensor is only convenient for measuring the surface temperature and the internal temperature is difficult to measure, the dynamic change of the temperature field in the fusion casting process is difficult to monitor through an experimental method.
Value simulation method: with the development of computational fluid mechanics, numerical simulation methods such as a finite element method and the like are widely applied, and with the improvement of a relevant theoretical model, the calculation precision can basically meet the requirements of engineering practical application, and the temperature field distribution in the curing process can be obtained. However, in the casting process, because the CFD numerical simulation has the disadvantages of large model, large number of grids, large simulation workload, low calculation efficiency, complex operation and time consumption, it can only be used for prior simulation, and cannot perform real-time simulation and monitoring on the actual temperature field affected by various dynamic uncertain factors in the forming process. Therefore, real-time determination of quality and process adjustment in the process cannot be supported.
And (3) machine learning: because the machine learning technology utilizes a large amount of models trained by historical data to discover the rules of data implications, compared with the traditional numerical method, the method does not need to understand complex physical equations therein, reduces time and calculation cost, and can realize the function of rapid prediction. However, at present, temperature prediction based on machine learning is mainly a temperature change rule of a certain position and a fitting rule is mainly performed through nonlinearity, and meanwhile, machine learning is a black box model, so that interpretability is not available, and similarity (symmetry) of a physical process is not effectively utilized, so that massive data is needed, and engineering is difficult to achieve.
The temperature field control is to perform corresponding adjustment by analyzing the current temperature field data in real time, and apply the decision result to a production workshop, so as to finally keep the temperature field in the casting process within a reasonable range. The existing temperature field control method has the following defects: firstly, the method mainly depends on the experience of workers, has poor accuracy and low efficiency, and cannot monitor the state of a temperature field in real time, so that the method has no basis and poor reliability when quality control is carried out; and secondly, the alloy is obtained by analyzing historical data or finite element models, so that the casting quality can be improved to a certain degree, but the alloy lacks timeliness and fidelity and is difficult to meet the requirements.
Disclosure of Invention
In view of the above, the invention aims to provide a thermal diffusion digital twin model, a temperature field real-time optimization control model and a method in an explosive casting and curing process, wherein the thermal diffusion digital twin model is driven by constructed data, so that real-time simulation of a temperature field in the explosive casting process can be realized, and the problem that the temperature field distribution of the whole curing process cannot be quickly obtained in the prior art is solved; the process parameters are dynamically optimized through a temperature field optimization control method, the temperature field control in the casting process is realized, and the problems of multiple forming defects, poor quality stability and the like caused by constraints such as incapability of monitoring an internal temperature field in the casting process, incapability of optimizing the process parameters according to actual temperature field distribution and the like are solved.
In order to achieve the purpose, the invention provides the following technical scheme:
the invention firstly provides a method for constructing a thermal diffusion digital twin model in an explosive fusion casting and curing process, which comprises the following steps:
the method comprises the following steps: constructing a finite element model of the explosive fusion casting solidification process;
step two: carrying out numerical simulation on the explosive casting and curing process, randomly changing process parameters consisting of the temperature of a hot core rod, the insertion depth of the hot core rod and the applied pressure in the simulation process, and extracting node data at all time steps in the curing process, wherein the node data comprises node coordinates and node temperature; synthesizing the node data under a plurality of time steps into a group of temperature field distribution data of which the process parameters change along with time, and constructing a data set by using the temperature field distribution data;
step three: construction of data-driven thermal diffusion digital twin model
31 Randomly extracting node data from the temperature field distribution data at the same time step as local information, and coding the node data from an input space into a potential space through a coder to obtain an initial node state of function local interpolation;
32 Realizing space dispersion by T round message transmission in the GNN model, updating node data and making local information global;
33 Mapping from the potential space to an output space with a decoder to predict temperature field distribution data at a next time step;
34 Comparing whether the error between the predicted next time step temperature field distribution data and the simulated next time step temperature field distribution data is less than a set threshold value: if yes, obtaining a data-driven thermal diffusion digital twin model; if not, the GNN model parameters are updated, and the temperature field distribution data of the next time step is taken as input, and the step 31) is executed.
Further, in the first step, the method for constructing the finite element model of the explosive fusion casting solidification process comprises the following steps:
11 Selecting an explosive fusion casting solidification molding process mathematical model;
12 Establishing a finite element model of the explosive fusion casting and curing process according to the actual working condition, and carrying out pretreatment on numerical simulation of the explosive fusion casting and curing process to complete grid division;
13 Boundary conditions and initialization parameters are set, and numerical simulation is performed by randomly changing process parameters.
Further, in the second step, normalization processing is performed on the temperature field distribution data.
Further, in the step 32), the node temperature is updated by using a temperature transfer function:
wherein, f (r) i ) Representing spatial nodes r i The temperature of (a); v j Representing the ratio of the mass to the density of surrounding spatial nodes; w (| r) i -r j L, h) represents a spatial node r i And a space node r j H represents a temperature diffusion range; | r i -r j I represents a spatial node r i And a space node r j The distance between them.
Further, in the step 33), the method for predicting the temperature field distribution data of the next time step includes:
T n+1 ＝T n +△tdT
wherein, T n A temperature representing the current time step; t is n+1 Represents the temperature of the next time step; Δ t represents the time interval of one time step; dT represents a temperature change.
The invention also discloses a construction method of the temperature field optimization control model in the explosive fusion casting solidification process, which comprises the following steps:
s1: initializing all parameters w of the Q network randomly, initializing all states and actions of the Q network based on the parameters w, obtaining corresponding values Q, and simultaneously emptying an experience playback pool D;
s2: collecting wall surface temperature data, and utilizing the data-driven thermal diffusion digital twin model constructed by the method to simulate the temperature field in real time to obtain the current state s of the temperature field j And global temperature field simulation data, wherein the global temperature field simulation data is used as a feature vector phi(s) of the Q network j )；
S3: using a feature vector phi(s) in a Q network j ) As input, obtaining Q value output corresponding to all actions of the Q network; selecting a corresponding action a in a current Q-value output by using an E-greedy method j J is an iterationThe number of times;
s4: in a state s j Performing the current action a j And obtaining a new state s by utilizing data-driven thermal diffusion digital twin model real-time simulation j+1 And the corresponding feature vector phi(s) j+1 ) And a prize r j+1 Judging the state s j+1 Whether it is the final state: if so, terminating the iteration to obtain a temperature field optimization control model; if not, executing step S5);
s5: quadruple(s) to be obtained j ，a j ，r j+1 ，s j+1 ) Added to the empirical replay pool D, m samples(s) are sampled from the empirical replay pool D i ，a i ，r i+1 ，s i+1 ) Wherein i = j-m +1, j-m +2, j-1, j; j is more than or equal to m; calculating the current target Q value y j Using the target Q value y j Calculating a mean square error loss function, and updating Q network parameters by using the mean square error loss function;
update the state, order s j ＝s j+1 ，r j ＝r j+1 ，j＝j+1；
Step S3 is performed.
Further, in the step S5, the current target Q value y j Comprises the following steps:
wherein γ represents an attenuation value;
represents a pair action a j+1 State of (c) j+1 The maximum estimate of (c).
Further, in step S5, the mean square error loss function is:
wherein m represents the number of samples; j represents the current iteration number; w represents all parameters of the Q network.
A temperature field optimization control method in an explosive casting and curing process is characterized in that wall surface temperature data are collected in real time, and a temperature field optimization control model constructed by the method is adopted to regulate and control process parameters in the explosive casting process in real time so as to realize temperature field control in the explosive casting and curing process.
The invention has the beneficial effects that:
the invention relates to a construction method of a thermal diffusion digital twin model in an explosive casting and solidifying process, which comprises the following steps of firstly, constructing a finite element model in the explosive casting and solidifying process, and providing data for constructing a data-driven thermal diffusion digital twin model through numerical simulation of finite elements; specifically, in the simulation process, three process parameters of the temperature of the hot core rod, the insertion depth of the hot core rod and the magnitude of applied pressure are randomly changed, node data under all time steps are extracted, sufficient temperature field distribution data are collected, and a data set is created; then, a data-driven thermal diffusion digital twin model is constructed, and because the global temperature data in the fusion casting process is difficult to measure, the invention randomly samples and extracts node data as local information, the local information is spread to any place in the space through a GNN model, and the prediction from the local temperature data to a global temperature field is realized through the learning of a thermal diffusion mechanism, so that the local information is global; finally, optimizing the data-driven thermal diffusion digital twin model until the precision of an output target is approached, obtaining the optimized data-driven thermal diffusion digital twin model considering the temperature of the hot core rod, the insertion depth of the hot core rod and the magnitude of applied pressure, and realizing the learning of the temperature field distribution rule of the explosive solidification process under different moments and different process technological parameters; in actual use, the wall surface temperature data acquired in real time by the thermocouple can be used for driving the thermal diffusion digital twin model to obtain the global temperature field distribution based on the data, so that the rapid real-time simulation of the temperature field is realized, and the problem that the temperature field distribution of the whole explosive casting and curing process cannot be rapidly obtained in the prior art is solved.
The method for constructing the temperature field optimization control model in the explosive casting and curing process is characterized in that a real-time simulation result of a data-driven thermal diffusion digital twin model in the explosive casting and curing process is utilized to quickly identify the temperature field distribution characteristics of the explosive casting and curing process, and process parameters are optimized in real time according to the temperature field distribution and the casting defect occurrence judgment criterion, so that the real-time regulation and control of the casting temperature field distribution are realized, and the explosive casting quality is improved; the data driving thermal diffusion digital twin model can quickly obtain the global temperature field distribution based on the thermal diffusion digital twin model only by monitoring the wall surface temperature in real time through a thermocouple, so that the real-time simulation of the temperature field in the explosive casting process is realized, and the problem that the temperature field distribution in the whole curing process cannot be quickly obtained in the prior art is solved; on the basis that a data-driven thermal diffusion digital twin model can perform real-time rapid simulation on an explosive solidification process, a deep reinforcement learning method is adopted to perform dynamic optimization on process parameters based on temperature field distribution in a casting process, wall surface temperature data are acquired in real time through a thermocouple, a global temperature field is monitored in real time and predicted in future state based on the thermal diffusion digital twin model, and the process parameters are adjusted according to the real-time monitoring and prediction, so that the control of the temperature field in the casting process is realized, and the problems of more molding defects, poor quality stability and the like caused by constraints that the internal temperature field in the casting process cannot be monitored, the process parameters cannot be optimized according to actual temperature field distribution and the like are solved.
In addition, because the traditional temperature field control method is poor in accuracy and efficiency and cannot monitor the state of the temperature field in real time, the method is free of data and poor in reliability during quality control, and cannot meet production requirements.
Drawings
In order to make the object, technical scheme and beneficial effect of the invention more clear, the invention provides the following drawings for explanation:
FIG. 1 is a schematic diagram of an embodiment of a thermal diffusion digital twin model construction method in an explosive fusion casting solidification process;
FIG. 2 is a schematic illustration of a constructed finite element model;
FIG. 3 is a mesh partition of a finite element model;
FIG. 4 is a diagram of the results of numerical simulation of temperature field distribution by a finite element model;
FIG. 5 is a schematic illustration of combining node data into temperature field distribution data;
FIG. 6 is a diagram of the correspondence between process parameters and temperature field distribution data;
FIG. 7 is a schematic diagram of an embodiment of a temperature field optimization control model construction method in an explosive fusion casting solidification process according to the present invention;
FIG. 8 is a schematic modeling diagram of a temperature field optimization control model;
FIG. 9 is a schematic diagram of an embodiment of the temperature field optimization control method in the process of casting and solidifying the explosive.
Detailed Description
The present invention is further described with reference to the following drawings and specific examples so that those skilled in the art can better understand the present invention and can practice the present invention, but the examples are not intended to limit the present invention.
As shown in FIG. 1, the method for constructing the thermal diffusion digital twin model in the explosive fusion casting solidification process of the embodiment comprises the following steps.
The method comprises the following steps: a finite element model of an explosive fusion casting solidification process is constructed, and the method comprises the following steps:
11 Selecting a mathematical model of the explosive fusion casting solidification molding process, namely a heat transfer model of a cooling process;
12 Establishing a finite element model of the explosive casting and curing process according to the actual working conditions, and performing pretreatment on numerical simulation of the explosive casting and curing process to complete grid division as shown in FIG. 3;
13 Set boundary conditions and initialization parameters and randomly change process parameters for numerical simulation as shown in fig. 4.
Step two: carrying out numerical simulation on the explosive casting and curing process, randomly changing process parameters consisting of the temperature of a hot core rod, the insertion depth of the hot core rod and the applied pressure in the simulation process, and extracting node data at all time steps in the curing process, wherein the node data comprises node coordinates and node temperature; and synthesizing the node data under a plurality of time steps into a group of temperature field distribution data of which the process parameters change along with time, and constructing a data set by utilizing the temperature field distribution data. Specifically, the method for creating the data set comprises the following steps:
21): synthesizing the node data at each time step in the curing process into a set of temperature field distribution data which changes along with the real time sequence under the process condition, as shown in FIG. 5;
22 To establish the relationship between the curing time and the temperature field distribution under different process parameters, the process parameters and the curing time are in one-to-one correspondence with the temperature field distribution data, as shown in fig. 6;
23 In order to eliminate the difference between the data magnitude orders, the interrelation among all factors is well reflected, the error in the data processing process is reduced, and the temperature field distribution data is subjected to normalization processing.
Step three: construction of data-driven thermal diffusion digital twin model
31 Randomly extracting node data from the temperature field distribution data at the same time step as local information, and coding the node data from an input space into a potential space through a coder to obtain an initial node state of function local interpolation;
32 Realizing space dispersion by T round message transmission in the GNN model, updating node data and making local information global;
33 Mapping from the potential space to an output space with a decoder to predict temperature field distribution data at a next time step;
34 Comparing whether the error between the predicted next time step temperature field distribution data and the simulated next time step temperature field distribution data is less than a set threshold value: if so, obtaining a data-driven thermal diffusion digital twin model; if not, the GNN model parameters are updated, and the temperature field distribution data of the next time step is taken as input, and the step 31) is executed.
After the liquid explosive is filled in the bullet mould, the explosive solidification mainly considers unstable heat conduction, which is usually described by an unstable heat conduction partial differential equation, and the method is also a traditional numerical simulation method. Based on this, the data-driven thermal diffusion digital twinning model considers representing the fluid domain as dense particles (nodes), each of which contains temperature information and position information. Because the heat can diffuse to the surroundings when the temperature is different, the temperature of each node can influence the temperature of the surrounding nodes, the diffusion rule of the node temperature is related to the position, the temperature and the diffusion distance of the node, and the thermal diffusion property can be expressed by using a continuity function, so if the positions and the temperature values of a plurality of nodes are known, the temperature can be diffused and transferred by using the temperature diffusion function of the nodes. Finally, by interpolating weighted sums between node values, the output value can be read from any node in the system and the value of any new query point in space can be predicted. Therefore, in the embodiment, the data-driven thermal diffusion digital twin model is to establish a temperature propagation model to predict the temperature through the mechanism learning of heat transfer. It is clear that the data-driven thermal diffusion digital twin model of the present embodiment does not learn to map the inputs directly to the global solution, but rather derives the global solution by learning a simpler local model and then using message passing. It is possible to have far fewer nodes than training examples and sparse connectivity, making it more computationally efficient.
When the liquid explosive is filled in the projectile body die, the Fourier law is satisfied, and the unstable heat conduction partial differential equation is used for describing that:
the three-dimensional heat conduction formula is simplified and transformed as follows:
in the formula:
wherein ρ represents an average density of a liquid phase and a solid phase, ρ L Denotes the density of the liquid phase, p s Denotes the density of the solid phase, f L Denotes the liquid phase volume fraction, f s The solid phase volume fraction is shown, c is the isobaric specific heat capacity, T is the temperature of the liquid medicine, T is time, k is the heat conductivity coefficient, and L is the latent heat of crystallization; k is a radical of x Represents the thermal conductivity in the X direction; k is a radical of y Represents the thermal conductivity in the Y direction; k is a radical of z Indicating the Z-direction thermal conductivity.
Time-discretizing it, from the temperature T at the present moment n Deducing the temperature T at the next moment n+1 ：
Wherein, T n Temperature representing the current time step; t is n+1 Represents the temperature of the next time step; Δ t represents the time interval of one time step; dT represents a temperature change.
After time dispersion, the temperature of any particle needs to be spatially dispersed, because the temperature of any particle, in addition to changing with time, also diffuses heat to surrounding particles, which is normally a continuous function, through which the degree of influence on the thermal diffusion of surrounding particles is represented, mainly related to the distance between the temperature of the particle itself and other particles, because the particle diffusion has a certain range, and the temperature of any particle can also be obtained by the surrounding particles according to a certain weighted sum, so that the node temperature is updated by using a temperature transfer function:
wherein, f (r) i ) Representing spatial nodes r i The temperature of (a); v j Representing the ratio of the mass to the density of the surrounding spatial nodes; w (| r) i -r j L, h) represents a spatial node r i And a space node r j H represents the temperature diffusion range; | r i -r j | represents a spatial node r i And a space node r j The distance between them.
The method for constructing the thermal diffusion digital twinning model in the explosive casting and curing process comprises the following steps of firstly, providing data for establishing a data-driven thermal diffusion digital twinning model through constructing a finite element model in the explosive casting and curing process and through finite element numerical simulation; specifically, in the simulation process, three process parameters of the temperature of the hot core rod, the insertion depth of the hot core rod and the magnitude of applied pressure are randomly changed, node data under all time steps are extracted, sufficient temperature field distribution data are collected, and a data set is created; then, a data-driven thermal diffusion digital twin model is constructed, and because the global temperature data in the fusion casting process is difficult to measure, the invention randomly samples and extracts node data as local information, the local information is spread to any place in the space through a GNN model, and the prediction from the local temperature data to a global temperature field is realized through the learning of a thermal diffusion mechanism, so that the local information is global; finally, optimizing the data-driven thermal diffusion digital twin model until the precision of an output target is approached, obtaining the optimized data-driven thermal diffusion digital twin model considering the temperature of the hot core rod, the insertion depth of the hot core rod and the magnitude of applied pressure, and realizing the learning of the temperature field distribution rule of the explosive solidification process under different moments and different process technological parameters; in actual use, the wall surface temperature data acquired in real time by the thermocouple can be used for driving the thermal diffusion digital twin model to obtain the global temperature field distribution based on the data, so that the rapid real-time simulation of the temperature field is realized, and the problem that the temperature field distribution of the whole explosive casting and curing process cannot be rapidly obtained in the prior art is solved.
As shown in fig. 7, this embodiment further provides a method for constructing an optimized control model of a temperature field in an explosive fusion casting solidification process, including the following steps:
s1: all parameters w of the Q network are initialized randomly, all states and actions of the Q network are initialized based on the parameters w, corresponding values Q are obtained, and meanwhile, the experience playback pool D is emptied. Specifically, the parameter w comprises iteration round number T, state characteristic dimension n, action set A, step length alpha, attenuation factor gamma, exploration rate epsilon, Q network structure and sample number m of batch gradient descent;
s2: collecting wall surface temperature data, and utilizing the data-driven thermal diffusion digital twin model constructed by the method to simulate the temperature field in real time to obtain the current state s of the temperature field j And global temperature field simulation data, wherein the global temperature field simulation data is used as a characteristic vector phi(s) of the Q network j )；
S3: using a feature vector phi(s) in a Q network j ) As input, obtaining Q value output corresponding to all actions of the Q network; selecting a corresponding action a in a current Q-value output by using an E-greedy method j J is the number of iterations;
s4: in a state s j Performing a current action a j And obtaining a new state s by utilizing real-time simulation of a data-driven thermal diffusion digital twin model j+1 And the corresponding feature vector phi(s) j+1 ) And a prize r j+1 Judging the state s j+1 Whether it is the final state: if so, terminating the iteration to obtain a temperature field optimization control model; if not, executing step S5);
s5: quadruple(s) to be obtained j ，a j ，r j+1 ，s j+1 ) Added to the empirical replay pool D, m samples(s) are sampled from the empirical replay pool D i ，a i ，r i+1 ，s i+1 ) Wherein i = j-m + 1, j-m + 2, j-1, j; j is more than or equal to m; calculating the current target Q value y j Using the target Q value y j Calculating a mean square error loss function, and updating Q network parameters by using the mean square error loss function;
update the state, order s j ＝s j+1 ，r j ＝r j+1 ，j＝j+1；
Step S3 is performed.
Wherein the current target Q value y j Comprises the following steps:
wherein γ represents an attenuation value;
represents a pair of actions a j+1 State of (c) j+1 The maximum estimate of (c).
The mean square error loss function is:
wherein m represents the number of samples; j represents the current iteration number; w represents all the parameters of the Q network.
After the thermal diffusion digital twin model is established, the global temperature field distribution can be obtained based on the thermal diffusion digital twin model only through wall surface temperature data acquired by the thermocouple in real time, and real-time simulation of the temperature field is realized. Because the generation of defects in the casting process is closely related to the distribution of the temperature field, in order to reduce the generation of the defects, the temperature field at each moment needs to be evaluated, and the defect generation judgment criterion based on the casting temperature field is used as the temperature field optimization standard to optimize process parameters. In order to realize the control of the temperature field in the casting process, a deep reinforcement learning method is adopted to optimize the technological parameters in the casting process in real time, and a casting process temperature field control model is established based on the deep reinforcement learning and mainly comprises an environment and an intelligent body, wherein the environment comprises a thermal diffusion digital twin model and is used for providing the temperature field change at different moments under different technological parameters, establishing history and providing training data for the temperature field control model. In the training process, the environment receives the action and the current temperature state, the current reward and the observed value are obtained according to the previous environment state and are input into the intelligent agent, and the intelligent agent outputs the next long action and starts the next interaction. And continuously updating network parameters by utilizing a deep reinforcement learning algorithm according to a target value in the process of interaction between the intelligent agent and the environment (the intelligent agent interacts with the environment every time, the environment returns to reward and tells the intelligent agent whether the action just performed is good or bad), and realizing the dynamic control of the temperature field in the casting process. When the device is actually used, firstly, the working state is initialized and initial process parameters are given, and then the global temperature field distribution of the current moment is obtained on the basis of the thermal diffusion digital twin model according to the wall surface temperature acquired in real time; and finally, comparing the temperature field data with the target value to judge whether the quality requirement is met, adjusting the action (process parameter) at the next moment based on the comparison, and repeatedly executing the process until the whole casting process is finished.
As shown in fig. 8, the agent is an executing individual, and the executing individual can be operated to make different selections (i.e., actions), such as adjusting process parameters and changing temperature field distribution; an environment having a state (temperature field distribution at different times and under different process parameters) for feeding back the state to the agent; the action is to adjust the process parameters, so that the environment, namely the state, is changed; rewards are used to assess whether an action provided by the agent is changing the environmental state; in the fusion casting process, the intelligent body continuously adjusts the action according to the comparison between the temperature field and the target temperature field, so that the state is changed, and finally the temperature field is controlled in real time.
According to the method for constructing the temperature field optimization control model in the explosive casting and curing process, the real-time simulation result of the data-driven thermal diffusion digital twin model in the explosive casting and curing process is utilized to quickly identify the temperature field distribution characteristics in the explosive casting and curing process, and the process parameters are optimized in real time according to the temperature field distribution and the casting defect occurrence judgment criterion, so that the real-time regulation and control of the casting temperature field distribution are realized, and the explosive casting quality is improved; the data driving thermal diffusion digital twin model can quickly obtain the global temperature field distribution based on the thermal diffusion digital twin model only by monitoring the wall surface temperature in real time through a thermocouple, so that the real-time simulation of the temperature field in the explosive casting process is realized, and the problem that the temperature field distribution in the whole curing process cannot be quickly obtained in the prior art is solved; on the basis that a data-driven thermal diffusion digital twin model can perform real-time rapid simulation on an explosive solidification process, a deep reinforcement learning method is adopted to perform dynamic optimization on process parameters based on temperature field distribution in a casting process, wall surface temperature data are acquired in real time through a thermocouple, a global temperature field is monitored in real time and predicted in future state based on the thermal diffusion digital twin model, and the process parameters are adjusted according to the real-time monitoring and prediction, so that the control of the temperature field in the casting process is realized, and the problems of more molding defects, poor quality stability and the like caused by constraints that the internal temperature field in the casting process cannot be monitored, the process parameters cannot be optimized according to actual temperature field distribution and the like are solved.
In addition, because the traditional temperature field control method is poor in accuracy and efficiency and cannot monitor the state of the temperature field in real time, the method is free of data and poor in reliability during quality control, and cannot meet production requirements.
As shown in fig. 9, the present embodiment further provides a temperature field optimization control method in an explosive casting and curing process, which collects wall surface temperature data in real time, and performs real-time regulation and control on process parameters in the explosive casting process by using the temperature field optimization control model constructed by the method in the present embodiment, so as to implement temperature field control in the casting and curing process.
The above-mentioned embodiments are merely preferred embodiments for fully illustrating the present invention, and the scope of the present invention is not limited thereto. The equivalent substitution or change made by the technical personnel in the technical field on the basis of the invention is all within the protection scope of the invention. The protection scope of the invention is subject to the claims.
Claims (9)
1. A method for constructing a thermal diffusion digital twin model in an explosive fusion casting and solidification process is characterized by comprising the following steps of: the method comprises the following steps:
the method comprises the following steps: constructing a finite element model of an explosive fusion casting and curing process;
step two: carrying out numerical simulation on the explosive fusion casting and curing process, randomly changing process parameters consisting of the temperature of a hot core rod, the insertion depth of the hot core rod and the magnitude of applied pressure in the simulation process, and extracting node data at all time steps in the curing process, wherein the node data comprises node coordinates and node temperature; synthesizing the node data under a plurality of time steps into a group of temperature field distribution data of which the process parameters change along with time, and constructing a data set by using the temperature field distribution data;
step three: construction of data-driven thermal diffusion digital twin model
31 Node data is randomly extracted from temperature field distribution data at the same time step to serve as local information, and the node data is coded from an input space into a potential space through a coder to obtain an initial node state of function local interpolation;
32 Realizing space dispersion by T round message transmission in the GNN model, updating node data and making local information global;
33 Mapping from the potential space to an output space with a decoder to predict temperature field distribution data at a next time step;
34 Comparing whether the error between the predicted next time step temperature field distribution data and the simulated next time step temperature field distribution data is less than a set threshold value: if yes, obtaining a data-driven thermal diffusion digital twin model; if not, the GNN model parameters are updated, and the temperature field distribution data of the next time step is taken as input, and the step 31) is executed.
2. The method for constructing the thermal diffusion digital twin model in the explosive fusion casting solidification process according to claim 1, which is characterized by comprising the following steps of: in the first step, the method for constructing the finite element model in the explosive fusion casting solidification process comprises the following steps:
11 Selecting a mathematical model of an explosive fusion casting solidification molding process;
12 Establishing a finite element model of the explosive casting and curing process according to the actual working condition, and carrying out pretreatment on numerical simulation of the explosive casting and curing process to complete grid division;
13 Boundary conditions and initialization parameters are set, and numerical simulation is performed by randomly changing process parameters.
3. The method for constructing the thermal diffusion digital twin model in the explosive fusion casting solidification process according to claim 1, characterized by comprising the following steps: and in the second step, the temperature field distribution data is subjected to normalization processing.
4. The method for constructing the thermal diffusion digital twin model in the explosive fusion casting solidification process according to claim 1, which is characterized by comprising the following steps of: in the step 32), the node temperature is updated by using a temperature transfer function:
wherein, f (r) i ) Representing spatial nodes r i The temperature of (a); v j Representing the ratio of the mass to the density of the surrounding spatial nodes; w (| r) i -r j L, h) represents a spatial node r i And a space node r j H represents the temperature diffusion range; | r i -r j | represents a spatial node r i And a space node r j The distance between them.
5. The method for constructing the thermal diffusion digital twin model in the explosive fusion casting solidification process according to claim 1, which is characterized by comprising the following steps of: in the step 33), the method for predicting the temperature field distribution data of the next time step includes:
T n+1 ＝T n +△tdT
wherein, T n Temperature representing the current time step; t is a unit of n+1 Represents the temperature of the next time step; Δ t represents a time interval of one time step; dT represents a temperature change.
6. A method for constructing an optimized control model of a temperature field in an explosive casting and curing process is characterized by comprising the following steps: the method comprises the following steps:
s1: initializing all parameters w of the Q network randomly, initializing all states and actions of the Q network based on the parameters w, obtaining corresponding values Q, and simultaneously emptying an experience playback pool D;
s2: collecting wall surface temperature data, and simulating a temperature field in real time by using a data-driven thermal diffusion digital twin model constructed by the method according to any one of claims 1 to 5 to obtain the current state s of the temperature field j And global temperature field simulation data, wherein the global temperature field simulation data is used as a characteristic vector phi(s) of the Q network j )；
S3: using a feature vector phi(s) in a Q network j ) As input, obtaining Q value output corresponding to all actions of the Q network; selecting a corresponding action a in the current Q-value output by using an e-greedy method j J is the number of iterations;
s4: in a state s j Performing a current action a j And obtaining a new state s by utilizing real-time simulation of a data-driven thermal diffusion digital twin model j+1 And the corresponding feature vector phi(s) j+1 ) And a prize r j+1 Judging the state s j+1 Whether it is the final state: if so, terminating the iteration to obtain a temperature field optimization control model; if not, executing step S5);
s5: the quadruple(s) to be obtained j ，a j ，r j+1 ，s j+1 ) Added to the empirical replay pool D, m samples(s) are sampled from the empirical replay pool D i ，a i ，r i+1 ，s i+1 ) Wherein i = j-m +1, j-m +2, j-1, j; j is more than or equal to m; calculating the current target Q value y j Using the target Q value y j Computing a mean square error loss functionUpdating Q network parameters by using a mean square error loss function;
update the state, let s j ＝s j+1 ，r j ＝r j+1 ，j＝j+1；
Step S3 is performed.
7. The method for constructing the temperature field optimization control model in the explosive fusion casting solidification process according to claim 6, wherein the method comprises the following steps: in the step S5, the current target Q value y j Comprises the following steps:
wherein γ represents an attenuation value;
represents a pair action a j+1 State of (1) j+1 Is calculated.
8. The method for constructing the temperature field optimization control model in the explosive fusion casting solidification process according to claim 6 or 7, wherein the method comprises the following steps: in step S5, the mean square error loss function is:
wherein m represents the number of samples; j represents the current iteration number; w represents all the parameters of the Q network.
9. An optimization control method for a temperature field in an explosive casting and curing process is characterized by comprising the following steps: collecting wall surface temperature data in real time, and adopting the temperature field optimization control model constructed by the method of any one of claims 6-8 to regulate and control process parameters in the fusion casting process of the explosive in real time so as to realize temperature field control in the fusion casting solidification process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211157180.5A CN115481554A (en) | 2022-09-22 | 2022-09-22 | Thermal diffusion digital twin model, temperature field real-time optimization control model and method in explosive fusion casting and curing process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211157180.5A CN115481554A (en) | 2022-09-22 | 2022-09-22 | Thermal diffusion digital twin model, temperature field real-time optimization control model and method in explosive fusion casting and curing process |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115481554A true CN115481554A (en) | 2022-12-16 |
Family
ID=84394242
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211157180.5A Pending CN115481554A (en) | 2022-09-22 | 2022-09-22 | Thermal diffusion digital twin model, temperature field real-time optimization control model and method in explosive fusion casting and curing process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115481554A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117170334A (en) * | 2023-11-02 | 2023-12-05 | 钥准医药科技(启东)有限公司 | Intelligent control method and system for rapid drug fusion |
-
2022
- 2022-09-22 CN CN202211157180.5A patent/CN115481554A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117170334A (en) * | 2023-11-02 | 2023-12-05 | 钥准医药科技(启东)有限公司 | Intelligent control method and system for rapid drug fusion |
| CN117170334B (en) * | 2023-11-02 | 2024-03-08 | 钥准医药科技(启东)有限公司 | Intelligent control method and system for rapid drug fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110427654B (en) | Landslide prediction model construction method and system based on sensitive state | |
| CN111324993B (en) | Turbulent flow field updating method, device and related equipment | |
| CN114693005B (en) | Three-dimensional underground oil reservoir dynamic prediction method based on convolution Fourier neural network | |
| Zhang et al. | MeshingNet3D: Efficient generation of adapted tetrahedral meshes for computational mechanics | |
| CN115935834A (en) | History fitting method based on deep autoregressive network and continuous learning strategy | |
| CN115481554A (en) | Thermal diffusion digital twin model, temperature field real-time optimization control model and method in explosive fusion casting and curing process | |
| CN115186555B (en) | Digital twinning-based drying equipment live simulation method and related equipment | |
| CN114912364A (en) | Natural gas well flow prediction method, device, equipment and computer readable medium | |
| CN112131794A (en) | Hydraulic structure multi-effect optimization prediction and visualization method based on LSTM network | |
| CN109540089B (en) | Bridge deck elevation fitting method based on Bayes-Kriging model | |
| CN112307536B (en) | Dam seepage parameter inversion method | |
| CN110991605A (en) | Low-pressure casting mold temperature prediction method of multivariable time series deep belief network | |
| CN105022873B (en) | A kind of on-line reorganization method of isothermal die forging process mold temperature field | |
| CN113468466A (en) | Multi-working-condition one-dimensional wave equation solving method based on neural network | |
| CN112836789A (en) | Ground connection wall deformation dynamic prediction method based on composite neural network algorithm | |
| CN115062551B (en) | Wet physical process parameterization method based on time sequence neural network | |
| CN116227359A (en) | Flow field prediction method based on attention and convolutional neural network codec | |
| CN114864017B (en) | Prediction calculation method and system for overall grain size of aluminum alloy casting | |
| Skamrock | Truncation error estimates for refinement criteria in nested and adaptive models | |
| CN113743022B (en) | Storage and visualization method for high-precision climate change data | |
| CN113468795B (en) | General construction method for different parameter material partition casting/filling dam deformation mixed model | |
| CN111210877B (en) | Method and device for deducing physical parameters | |
| CN115270239A (en) | Bridge reliability prediction method based on dynamic characteristics and intelligent algorithm response surface method | |
| CN112800690B (en) | Underwater folding and unfolding mechanism parameter optimization method based on group intelligent optimization algorithm | |
| CN111460548B (en) | Normal-state roller compacted concrete gravity dam combined damming safety assessment method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |