Industrial air-supported membrane structure indoor transient temperature field rapid prediction method

Abstract

The invention relates to a rapid prediction method for an indoor transient temperature field of an industrial gas-supported membrane structure, belongs to the technical field of temperature field prediction, and solves the problems that a temperature field prediction method for a gas-supported membrane structure in the prior art is complex in modeling, large in calculation amount, low in efficiency and incapable of adapting to dynamic changes of a heat source. Comprising the following steps: acquiring a discrete change heat source and a continuous change heat source in a to-be-tested industrial gas-supported membrane structure; for each discrete change heat source, calculating first temperature rise distribution based on the temperature rise variation of all the discrete change heat sources; for each continuous change heat source, calculating second temperature rise distribution based on the first temperature field distribution of all the continuous change heat sources; and based on an initial temperature field, the first temperature rise distribution and the second temperature rise distribution, calculating to obtain a transient temperature field of the to-be-measured industrial air-supported membrane structure. The method for rapidly predicting the indoor transient temperature field of the industrial air-supported membrane structure is simple in modeling, rapid, efficient and capable of adapting to dynamic changes of the temperature field.

Description

Technical Field

[0001] This invention relates to the field of temperature field prediction technology, and in particular to a method for rapid prediction of transient indoor temperature fields in industrial air-supported membrane structures. Background Technology

[0002] Air-supported membrane structures are characterized by their light weight, rapid construction, large span, flexible design, energy efficiency, environmental friendliness, and reusability, making them widely used in industrial settings such as warehousing and logistics, environmental enclosures, and radiation shielding chambers. As a fully enclosed structure that relies on internal and external pressure differences to maintain its shape, the thermal environment characteristics of air-supported membrane structures differ significantly from traditional buildings: their internal temperature field is easily affected by fluctuations in the outdoor environment and disturbances from internal heat and cold sources, exhibiting a marked uneven temperature distribution. Especially in industrial settings, air-supported membrane structures typically contain multiple heat sources, such as self-heating biomass (e.g., coal) and large operating equipment (e.g., excavators, conveyors), leading to significant temperature stratification and generally higher temperatures in working areas. Furthermore, changes in the location of heat sources further exacerbate the dynamic fluctuations in the temperature field.

[0003] Existing technologies for temperature field prediction mainly include computational fluid dynamics (CFD), region modeling, fast fluid dynamics (FFD), and deep learning models. Among these, FFD is the primary method for predicting building temperature fields, but its complex modeling and analysis processes result in significant computational time consumption. Region models reduce computational load by merging computational nodes with similar boundary characteristics, thus achieving a balance between computational speed and prediction accuracy. Although this method significantly improves efficiency compared to traditional CFD, it is still based on the CFD framework and cannot meet the needs of rapid prediction of global temperature fields. Fast fluid dynamics (FFD) uses semi-Lagrangian techniques to solve the convection term of the momentum equation, avoiding iterative calculations and thus achieving high computational efficiency. However, current applications of FFD are mainly focused on predicting fluid velocity fields, with limited research on temperature field simulation. Deep learning models bypass the complex CFD calculation process by directly establishing a mapping relationship between boundary conditions and the temperature field, showing promising application prospects. However, this method relies on large-scale CFD datasets for training, and the model's transfer learning capability still needs improvement. Furthermore, existing temperature field prediction methods mostly focus on steady-state temperature distribution, and their ability to predict dynamic temperature fields remains insufficient.

[0004] Therefore, there is a need for a method that is simple to model, fast and efficient, and can adapt to dynamic changes in the temperature field for rapid prediction of the indoor transient temperature field of industrial air-supported membrane structures. Summary of the Invention

[0005] Based on the above analysis, the present invention aims to provide a rapid prediction method for the indoor transient temperature field of industrial air-supported membrane structures, in order to solve the problems of complex modeling, large computational load, low efficiency, and inability to adapt to dynamic changes in heat sources in existing air-supported membrane structure temperature field prediction methods.

[0006] This invention provides a method for rapid prediction of the transient temperature field inside an industrial air-supported membrane structure, comprising:

[0007] To obtain discrete and continuous heat sources in the industrial gas-supported membrane structure under test;

[0008] For each discrete heat source, the temperature rise change is calculated using the discrete change distribution prediction model corresponding to that discrete heat source, and the first temperature rise distribution is calculated based on the temperature rise changes of all discrete heat sources.

[0009] For each continuously varying heat source, the first temperature field distribution is calculated using the continuously varying distribution prediction model corresponding to that heat source, and the second temperature rise distribution is calculated based on the first temperature field distribution of all continuously varying heat sources.

[0010] The transient temperature field of the industrial gas-supported membrane structure under test is calculated based on the initial temperature field, the first temperature rise distribution, and the second temperature rise distribution.

[0011] Based on a further improvement of the above method, the discrete variation distribution prediction model includes:

[0012]

[0013] in, Let n be the temperature change at position p of the discrete heat source n. Let n be the response coefficient of the discrete heat source n at position p. Let C be the heat generated per unit time by a discrete, varying heat source n, where ρ is the air density and C is the heat output per unit time. p V is the specific heat capacity of air, and V is the volume of the room.

[0014] A further improvement to the above method, the calculation of the first temperature rise distribution based on the temperature rise changes of all discrete heat sources, includes:

[0015]

[0016] Δt=t-t0,

[0017] in, Let denot be the cumulative temperature rise of discrete heat source n at position p over a time interval Δt, where Δt is the duration of the discrete heat source's action, t0 is the initial time, t is the target time, and N is the number of discrete heat sources, n = 1, 2, 3, ..., N.

[0018] Based on a further improvement of the above method, the continuously changing distribution prediction model is implemented based on the U-net architecture, including:

[0019] The first input layer is used to receive environmental parameters and property parameters of the industrial gas-supported membrane structure under test.

[0020] The first fully connected layer is used to extract features from the environmental parameters and the attribute parameters to obtain a first feature vector, and then output it to the second fully connected layer.

[0021] The second fully connected layer is used to perform a non-linear transformation on the first feature vector to obtain a second feature vector, and input it to the bridging layer of the U-net architecture.

[0022] The second input layer is used to receive the temperature field distribution from the previous time step.

[0023] An encoder is used to perform a downsampling operation on the temperature field distribution to obtain a third feature vector;

[0024] A bridge layer is used to concatenate the second feature vector and the third feature vector to obtain a fourth feature vector;

[0025] A decoder is used to perform an upsampling operation on the fourth feature vector to obtain the predicted temperature field distribution;

[0026] The second output layer is used to output the predicted temperature field distribution.

[0027] Based on further improvements to the above method, the environmental parameters include outdoor temperature, air conditioning supply air temperature, and ventilation system wind speed.

[0028] A further improvement to the above method, the calculation of the second temperature rise distribution based on the first temperature field distribution of all continuously varying heat sources, includes:

[0029]

[0030] Among them, TC j (t) represents the temperature field distribution of the continuously changing heat source j at time t, TC j (t0) represents the temperature field distribution of the continuously changing heat source j at time t0, where j = 1, 2, 3, ..., m.

[0031] Based on a further improvement of the above method, the calculation of the transient temperature field of the industrial gas-supported membrane structure under test based on the initial temperature field, the first temperature field, and the second temperature field includes:

[0032]

[0033] Among them, T total (Δt) represents the transient temperature field of the industrial gas-supported membrane structure under test at time t, T inital This represents the transient temperature field distribution of the industrial gas-supported membrane structure under test at the initial time t0.

[0034] Based on further improvements to the above method, the training sample dataset for the continuously changing distribution prediction model is constructed as follows:

[0035] S1: Construct finite element models of industrial air-supported membrane structures of different sizes;

[0036] S2: Set the number and type of discrete and continuous heat sources;

[0037] S3: Set the range of location variation, floor area variation, and heat output variation for discrete heat sources; set the range of outdoor temperature variation, air conditioner temperature variation, and wind speed variation when a continuous heat source is running.

[0038] S4: Construct a simulation data table of working condition parameters based on the variation range of each parameter in steps S2-S3;

[0039] S5: For each finite element model in step S1, perform the following operation:

[0040] Perform finite element simulation on each data point in the simulation data table of operating conditions parameters, and record the temperature field data of that data point at each time step. Use the attribute parameters of the finite element model, the attribute parameters of each discrete heat source, the attribute parameters of each continuously changing heat source, and the temperature field data of that time step as a training sample.

[0041] Based on further improvements to the above method, the training process of the continuously changing distribution prediction model includes:

[0042] Pre-construct a continuous change distribution prediction model corresponding to different continuously changing heat source types;

[0043] The number of discrete and continuously varying heat sources is determined based on the training samples.

[0044] For each discrete heat source, the first temperature field distribution is calculated using the discrete change distribution prediction model corresponding to that discrete heat source.

[0045] For each continuously varying heat source, the second temperature field distribution is calculated using the continuously varying distribution prediction model corresponding to that heat source.

[0046] The predicted transient temperature field is calculated based on the initial temperature field, the temperature field distribution of each discrete heat source, and the temperature field distribution of each continuously changing heat source.

[0047] The models of each continuous heat source are corrected based on the predicted transient temperature field and the temperature field data corresponding to that data.

[0048] Based on further improvements to the above method, the construction of finite element models for industrial gas-supported membrane structures of different sizes includes:

[0049] A1: Based on business requirements, define the length range, width range, height range, membrane thickness range, radius of curvature range, and material type of the industrial air-supported membrane structure;

[0050] A2: Construct a parameter table for industrial air-supported membrane structures. The value of each parameter in this table must conform to the parameter range constraints in step A1.

[0051] A3: Construct a finite element model corresponding to each data point in the parameter table using ANSYS Fluent.

[0052] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0053] This invention provides a rapid prediction method for the transient temperature field inside an industrial air-supported membrane structure. The method divides the heat sources in the industrial air-supported membrane structure into discrete and continuously changing heat sources. For discrete heat sources, a physical mechanism is used to calculate the temperature field, while for continuously changing heat sources, a deep learning model is used. This method, which separates the calculation of heat sources by integrating physical mechanisms and deep learning, enables rapid calculation for varying heat sources inside industrial air-supported membrane structures. Compared to traditional computational fluid dynamics, region modeling, fast fluid dynamics, and deep learning models, this method does not require remodeling or modification of the deep learning model. It not only simplifies the complexity and computational burden of existing rapid prediction models for the transient temperature field inside industrial air-supported membrane structures but also improves the efficiency of temperature field calculation and is applicable to problems involving dynamically changing heat sources.

[0054] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0055] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0056] Figure 1 This is an example diagram of a method for rapid prediction of the indoor transient temperature field of an industrial air-supported membrane structure according to an embodiment of the present invention. Detailed Implementation

[0057] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0058] Industrial air-supported membrane structures are an innovative architectural form utilizing air pressure support. Their core advantages lie in their ultra-large spans, rapid construction, economic flexibility, and high space efficiency, making them particularly suitable for industrial applications with high internal space requirements, short construction periods, limited budgets, or the need for temporary / mobile facilities. The internal temperature field of an industrial air-supported membrane structure is mainly affected by two types of heat sources: external solar radiation and internal heat sources, including self-heating materials such as coal and operating equipment such as conveyors and excavators. Since industrial air-supported membrane structures are primarily used for material storage and transfer operations, the internal material stacking positions and equipment operating states are dynamically changing. Combined with the regulating effect of the ventilation system, this results in a significant spatiotemporal dynamic characteristic of the temperature field. This complex time-varying characteristic poses a significant challenge to temperature field prediction: if traditional CFD simulation combined with machine learning is used, on the one hand, a large number of numerical simulations are needed for various possible operating conditions to build a training database, resulting in extremely high computational costs; on the other hand, existing methods often ignore the influence of time-varying factors, making it difficult to achieve accurate real-time prediction of transient temperature fields. To address this, the present invention provides a rapid prediction method for transient temperature fields in industrial air-supported membrane structures. By dividing the heat sources in the industrial air-supported membrane into discrete and continuously changing heat sources, and integrating physical mechanisms with a machine learning model based on deep learning, a fusion and superposition prediction method is performed to achieve high-precision, low-latency transient temperature field prediction. This solves the problems of existing temperature field prediction methods for air-supported membrane structures, such as complex modeling, large computational load, low efficiency, and inability to adapt to dynamic changes in heat sources.

[0059] A specific embodiment of the present invention discloses a method for rapid prediction of the transient temperature field indoors of an industrial air-supported membrane structure, such as... Figure 1 As shown, it includes:

[0060] S1: Obtain discrete and continuous heat sources in the industrial air-supported membrane structure under test.

[0061] The temperature field of industrial air-supported membrane structures is affected by various heat sources, typically including internal heat sources, heat exchange with the external environment, and the operation of the ventilation system. This invention categorizes the heat sources affecting the temperature field into two types: discrete heat sources and continuously varying heat sources. Discrete heat sources include internal heat gain, such as heat generated by self-heating materials and mechanical equipment; these sources have a fixed heat loss. Continuously varying heat sources include factors such as air conditioning and ventilation systems and external environmental influences; the impact of these heat sources on the temperature field is continuously changing and difficult to quantify into a finite number of discrete changes. Therefore, this invention defines the heat sources such as materials and mechanical equipment piled up in industrial air-supported membrane structures as discrete heat sources, using heat, floor space, and location as influencing factors; and defines external temperature and ventilation systems as continuously varying heat sources, considering their time-varying characteristics of temperature and wind speed.

[0062] For example, the self-heating material can be coal, grain, metal powder, etc., and the mechanical equipment can be bucket wheel excavator, loader, forklift, backup power generation equipment, etc.

[0063] Specifically, when using the method proposed in this invention to predict transient temperature fields, technicians or engineers first need to identify the types and number of heat sources in the industrial gas-supported membrane structure to be tested (i.e., the target research object). Only after statistically obtaining the types and number of discrete and continuous heat sources can subsequent operations be performed.

[0064] S2: For each discrete heat source, the temperature rise change is calculated using the discrete change distribution prediction model corresponding to that discrete heat source, and the first temperature rise distribution is calculated based on the temperature rise changes of all discrete heat sources.

[0065] The influence of discrete heat sources on the temperature field of large-volume industrial air films mainly exhibits local characteristics. Therefore, it is analyzed by calculating the generated local temperature field. For internal discrete heat sources, the change process has discrete characteristics, specifically manifested in changes in the location, area of ​​effect, and intensity of heat generation. In the activated state, the heat output per unit time of such heat sources usually remains constant. The discrete change distribution prediction model is as follows:

[0066]

[0067] in, Let be the temperature change of the discrete heat source n at position p, that is, the temperature change of the heat source n per unit time. Let n be the response coefficient of the discrete heat source n at position p. Let C be the heat generated per unit time by a discrete, varying heat source n, where ρ is the air density and C is the heat output per unit time. p V is the specific heat capacity of air, and V is the volume of the room.

[0068] After obtaining the discrete heat sources of the industrial gas-supported membrane structure under test in step S1, the temperature rise change of each discrete heat source is calculated using a discrete variation distribution prediction model. As for the calculation method of the response coefficient, this invention does not impose any limitations; any calculable method in the prior art can be used as the method for obtaining it.

[0069] The calculation of the first temperature rise distribution based on the temperature rise changes of all discrete heat sources includes:

[0070]

[0071] Δt=t-t0,

[0072] in, Let Δt be the cumulative temperature rise of discrete heat source n at position p during a time interval of Δt, where Δt is the duration of the discrete heat source, t0 is the initial time, t is the target time, and N is the number of discrete heat sources, n = 1, 2, 3, ..., N. This formula actually has two meanings: (1) This formula calculates the temperature field change of each discrete heat source at a certain position within a time interval; (2) N is the total number of heat sources. For each heat source, the difference in this formula lies only in the response coefficient and the heat output of the heat source.

[0073] It is worth noting that the temperature rise distribution also needs to be calculated in step S3. The time interval between steps S2 and S3 must be consistent to ensure the accuracy of the transient temperature field calculation in S4.

[0074] S3: For each continuously changing heat source, the first temperature field distribution is calculated using the continuously changing distribution prediction model corresponding to that heat source, and the second temperature rise distribution is calculated based on the first temperature field distribution of all continuously changing heat sources.

[0075] The heat output of heat sources such as the air conditioning and ventilation system inside the industrial air-supported membrane and the outdoor ambient temperature usually changes dynamically over time, and its temperature influence exhibits continuous nonlinear characteristics. Therefore, it is difficult to accurately describe the temperature field changes of a constant heat source using a discrete model. Hence, this invention uses a deep network model to describe the temperature field changes of a continuously changing heat source.

[0076] Specifically, the continuously changing distribution prediction model is implemented based on the U-net architecture, including:

[0077] The first input layer is used to receive environmental parameters and property parameters of the industrial air-supported membrane structure under test. The environmental parameters include: outdoor temperature, air conditioning supply air temperature, and ventilation system wind speed. The property parameters of the industrial air-supported membrane structure under test include: membrane structure length, membrane structure width, membrane structure height, membrane structure thickness, membrane structure radius of curvature, and membrane structure material type.

[0078] The first fully connected layer is used to extract features from the environmental parameters and the attribute parameters to obtain a first feature vector, which is then output to the second fully connected layer.

[0079] The second fully connected layer is used to perform a nonlinear transformation on the first feature vector to obtain a second feature vector, and then input it into the bridging layer of the U-net architecture.

[0080] The second input layer is used to receive the temperature field distribution from the previous time step. The temperature field distribution from the previous time step is represented using a 3D tensor.

[0081] An encoder is used to perform a downsampling operation on the temperature field distribution to obtain a third feature vector.

[0082] A bridging layer is used to concatenate the second and third feature vectors to obtain a fourth feature vector. For example, this step can be to directly concatenate the second and third feature vectors.

[0083] A decoder is used to perform an upsampling operation on the fourth feature vector to obtain the predicted temperature field distribution.

[0084] The second output layer is used to output the predicted temperature field distribution. Specifically, the predicted temperature field distribution output in this step is also represented in 3D tensor form.

[0085] After obtaining the continuously varying heat source of the industrial gas-supported membrane structure under test in step S1, the temperature rise change of each continuously varying heat source is calculated using a continuously varying distribution prediction model corresponding to its type. This invention establishes a corresponding continuously varying distribution prediction model for each type of continuously varying heat source. This design fully considers the characteristics of each continuously varying heat source itself and better highlights the role of each heat source in the temperature field distribution.

[0086] The second temperature rise distribution is calculated based on the first temperature field distribution of all continuously varying heat sources, including:

[0087]

[0088] Among them, TC j (t) represents the temperature field distribution of the continuously changing heat source j at time t, TC j (t0) represents the temperature field distribution of the continuously changing heat source j at time t0, where j = 1, 2, 3, ..., m.

[0089] For the continuously changing distribution prediction model, the training sample dataset is constructed as follows:

[0090] M1: Construct finite element models of industrial gas-supported membrane structures of different sizes, including:

[0091] A1: Based on business requirements, define the length range, width range, height range, membrane thickness range, radius of curvature range, and material type of the industrial air-supported membrane structure.

[0092] The range of each parameter and the type of material can be determined by business needs. Understandably, the larger the range of parameters set, the more finite element models can be built, the more accurate the later model predictions will be, and the better the differences in temperature fields in industrial gas-supported membrane structures of different sizes can be reflected.

[0093] A2: Construct a parameter table for industrial air-supported membrane structures. The value of each parameter in this table must conform to the parameter range constraints in step A1.

[0094] After determining the range of each parameter in step A1, the step size for increasing or decreasing each range can be set according to actual needs. For example, if the length range is 50m to 100m, the step size can be set to 10m. That is, if other parameters remain unchanged, five different finite element models of industrial air-supported membrane structures can be set.

[0095] Understandably, for an industrial air-supported membrane structure, the temperature field distribution at heights of 25m and 28m is not significantly different. Therefore, when setting the step size for each parameter, it can be determined based on the experience of technicians or process engineers, using the smallest difference that can produce a significant change in the parameter as the step size.

[0096] A3: Construct a finite element model corresponding to each data point in the parameter table using ANSYS Fluent.

[0097] For example, ANSYS Fluent 2021 is used to construct the finite element model.

[0098] M2: Sets the number and type of discrete and continuous heat sources.

[0099] Step S1 provides the types of some discrete and continuously varying heat sources, and the number of each type of heat source can be set according to business needs.

[0100] M3: Set the range of location variation, floor area variation, and heat output variation for discrete heat sources; set the range of outdoor temperature variation, air conditioner temperature variation, and wind speed variation when a continuous heat source is running.

[0101] The range of each parameter and the type of material can be determined by business needs. Understandably, the larger the range of parameters set, the more finite element models can be built, and the more accurate the model predictions will be in the later stages.

[0102] M4: Construct a simulation data table of working condition parameters based on the variation range of each parameter in steps M2-M3.

[0103] Specifically, for the simulation data table of operating condition parameters, one of the data can be represented as: {{Distance from heat source 1, location, floor area, heat generation}; {Distance from heat source 3, location, floor area, heat generation}; ...; {Distance from heat source N, location, floor area, heat generation}; {Continuous heat source 1, outdoor temperature, air conditioning temperature, wind speed}; {Continuous heat source 2, outdoor temperature, air conditioning temperature, wind speed}; ...; {Continuous heat source 3, outdoor temperature, air conditioning temperature, wind speed}}.

[0104] M5: For each finite element model in step M1, perform the following operation:

[0105] Perform finite element simulation on each data point in the simulation data table of operating conditions parameters, and record the temperature field data of that data point at each time step. Use the attribute parameters of the finite element model, the attribute parameters of each discrete heat source, the attribute parameters of each continuously changing heat source, and the temperature field data of that time step as a training sample.

[0106] For example, training samples can be represented as: {{distance from heat source 1, location, floor area, heat generation}; {distance from heat source 3, location, floor area, heat generation}; ...; {distance from heat source N, location, floor area, heat generation}; {continuous heat source 1, outdoor temperature, air conditioning temperature, wind speed}; {continuous heat source 2, outdoor temperature, air conditioning temperature, wind speed}; ...; {continuous heat source 3, outdoor temperature, air conditioning temperature, wind speed}; membrane structure length; membrane structure width; membrane structure height; membrane structure thickness; membrane structure radius of curvature; membrane structure material type; temperature field data}.

[0107] For example, to balance computational efficiency and accuracy, the RNGk-ε turbulence model was selected to simulate indoor airflow, the Boussinesq approximation method was used to handle natural convection effects, the SIMPLE algorithm was used to solve the simulation, and the time step of the transient simulation was set to one minute.

[0108] Specifically, the training process for the continuously changing distribution prediction model includes:

[0109] C1: Pre-construct continuous variation distribution prediction models corresponding to different types of continuously varying heat sources. Specifically, for each type of continuously varying heat source, a deep network model based on the U-net architecture is pre-constructed to correspond to that type.

[0110] C2: Determine the number of discrete and continuously varying heat sources based on the training samples.

[0111] Specifically, in step M5, the number of discrete and continuous heat sources included in each training sample is obtained.

[0112] C3: For each discrete heat source, the first temperature field distribution is calculated using the discrete change distribution prediction model corresponding to that discrete heat source.

[0113] C4: For each continuously varying heat source, the second temperature field distribution is calculated using the continuously varying distribution prediction model corresponding to that heat source.

[0114] C5: The predicted transient temperature field is calculated based on the initial temperature field, the temperature field distribution of each discrete heat source, and the temperature field distribution of each continuously changing heat source.

[0115] Specifically, the predicted transient temperature field is obtained using the following formula:

[0116]

[0117] Among them, T total (Δt) represents the transient temperature field of the industrial gas-supported membrane structure under test at time t, T inital This represents the transient temperature field distribution of the industrial gas-supported membrane structure under test at the initial time t0.

[0118] C6: Based on the predicted transient temperature field and the temperature field data corresponding to this data, the models of each continuous heat source are corrected.

[0119] To verify the prediction accuracy of each model, the accuracy was verified based on the CFD calculation results, with mean squared error (RMSE) or mean absolute error (MAE) used as the standard to measure the generalization ability of the model.

[0120] S4: The transient temperature field of the industrial gas-supported membrane structure under test is calculated based on the initial temperature field, the first temperature rise distribution, and the second temperature rise distribution, including:

[0121]

[0122] Among them, T total (Δt) represents the transient temperature field of the industrial gas-supported membrane structure under test at time t, T inital This represents the transient temperature field distribution of the industrial gas-supported membrane structure under test at the initial time t0.

[0123] Compared with existing technologies, this embodiment provides a rapid prediction method for the indoor transient temperature field of an industrial air-supported membrane structure. The method divides the industrial air-supported membrane structure under study into discrete and continuously varying heat sources. For discrete heat sources, a physical mechanism is used to calculate the temperature field, while for continuously varying heat sources, a deep learning model is used. This method, which separates the calculation of heat sources by integrating physical mechanisms and deep learning, enables rapid calculation for changes in the indoor heat sources of industrial air-supported membrane structures. Compared with traditional computational fluid dynamics, region modeling, fast fluid dynamics, and deep learning modeling methods, it does not require remodeling or modification of the deep learning model. This not only simplifies the complexity and computational burden of existing rapid prediction models for the indoor transient temperature field of industrial air-supported membrane structures but also improves the efficiency of temperature field calculation and is applicable to problems involving dynamically changing heat sources.

[0124] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0125] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for rapid prediction of transient temperature field in an industrial air-supported membrane structure, characterized in that, include: To obtain discrete and continuous heat sources in the industrial gas-supported membrane structure under test; For each discrete heat source, the temperature rise change is calculated using the discrete change distribution prediction model corresponding to that discrete heat source, and the first temperature rise distribution is calculated based on the temperature rise changes of all discrete heat sources. For each continuously varying heat source, the first temperature field distribution is calculated using the continuously varying distribution prediction model corresponding to that heat source, and the second temperature rise distribution is calculated based on the first temperature field distribution of all continuously varying heat sources. The transient temperature field of the industrial gas-supported membrane structure under test is calculated based on the initial temperature field, the first temperature rise distribution, and the second temperature rise distribution.

2. The method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 1, characterized in that, The discrete variation distribution prediction model includes: in, Let n be the temperature change at position p of the discrete heat source n. Let n be the response coefficient of the discrete heat source n at position p. Let C be the heat generated per unit time by a discrete, varying heat source n, where ρ is the air density and C is the heat output per unit time. p V is the specific heat capacity of air, and V is the volume of the room.

3. The method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 2, characterized in that, The calculation of the first temperature rise distribution based on the temperature rise changes of all discrete heat sources includes: Δt=t-t0, in, Let denot be the cumulative temperature rise of discrete heat source n at position p over a time interval Δt, where Δt is the duration of the discrete heat source's action, t0 is the initial time, t is the target time, and N is the number of discrete heat sources, n = 1, 2, 3, ..., N.

4. The method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 3, characterized in that, The continuously changing distribution prediction model is implemented based on the U-net architecture and includes: The first input layer is used to receive environmental parameters and property parameters of the industrial gas-supported membrane structure under test. The first fully connected layer is used to extract features from the environmental parameters and the attribute parameters to obtain a first feature vector, and then output it to the second fully connected layer. The second fully connected layer is used to perform a non-linear transformation on the first feature vector to obtain a second feature vector, and input it to the bridging layer of the U-net architecture. The second input layer is used to receive the temperature field distribution from the previous time step. An encoder is used to perform a downsampling operation on the temperature field distribution to obtain a third feature vector; A bridge layer is used to concatenate the second feature vector and the third feature vector to obtain a fourth feature vector; A decoder is used to perform an upsampling operation on the fourth feature vector to obtain the predicted temperature field distribution; The second output layer is used to output the predicted temperature field distribution.

5. The method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 4, characterized in that, The environmental parameters include outdoor temperature, air conditioning supply temperature, and ventilation system wind speed.

6. The method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 5, characterized in that, The calculation of the second temperature rise distribution based on the first temperature field distribution of all continuously changing heat sources includes: Among them, TC j (t) represents the temperature field distribution of the continuously changing heat source j at time t, TC j (t0) represents the temperature field distribution of the continuously changing heat source j at time t0, where j = 1, 2, 3, ..., m.

7. The method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 6, characterized in that, The transient temperature field of the industrial gas-supported membrane structure under test, calculated based on the initial temperature field, the first temperature field, and the second temperature field, includes: Among them, T total (Δt) represents the transient temperature field of the industrial gas-supported membrane structure under test at time t, T inital This represents the transient temperature field distribution of the industrial gas-supported membrane structure under test at the initial time t0.

8. The method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 6, characterized in that, The training sample dataset for the continuously changing distribution prediction model is constructed in the following manner: S1: Construct finite element models of industrial air-supported membrane structures of different sizes; S2: Set the number and type of discrete and continuous heat sources; S3: Set the range of location variation, floor area variation, and heat output variation for discrete heat sources; set the range of outdoor temperature variation, air conditioner temperature variation, and wind speed variation when a continuous heat source is running. S4: Construct a simulation data table of working condition parameters based on the variation range of each parameter in steps S2-S3; S5: For each finite element model in step S1, perform the following operation: Perform finite element simulation on each data point in the simulation data table of operating conditions parameters, and record the temperature field data of that data point at each time step. Use the attribute parameters of the finite element model, the attribute parameters of each discrete heat source, the attribute parameters of each continuously changing heat source, and the temperature field data of that time step as a training sample.

9. A method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 8, characterized in that, The training process of the continuously changing distribution prediction model includes: Pre-construct a continuous change distribution prediction model corresponding to different continuously changing heat source types; The number of discrete and continuously varying heat sources is determined based on the training samples. For each discrete heat source, the first temperature field distribution is calculated using the discrete change distribution prediction model corresponding to that discrete heat source. For each continuously varying heat source, the second temperature field distribution is calculated using the continuously varying distribution prediction model corresponding to that heat source. The predicted transient temperature field is calculated based on the initial temperature field, the temperature field distribution of each discrete heat source, and the temperature field distribution of each continuously changing heat source. The models of each continuous heat source are corrected based on the predicted transient temperature field and the temperature field data corresponding to that data.

10. A method for rapid prediction of indoor transient temperature field of an industrial air-supported membrane structure according to claim 8, characterized in that, The finite element models for constructing industrial gas-supported membrane structures of different sizes include: A1: Based on business requirements, define the length range, width range, height range, membrane thickness range, radius of curvature range, and material type of the industrial air-supported membrane structure; A2: Construct a parameter table for industrial air-supported membrane structures. The value of each parameter in this table must conform to the parameter range constraints in step A1. A3: Construct a finite element model corresponding to each data point in the parameter table using ANSYS Fluent.

Images