Mode training method, and task execution method and device

Abstract

Disclosed are a model training method, and a task execution method and device. The model training method comprises: acquiring each piece of physical field data of a mixture state in a specified space at least at two moments, and inputting each piece of physical field data at a first moment into a prediction model to be trained, so that on the basis of the each piece of physical field data, the prediction model predicts each piece of physical field data of the mixture state in the specified space at a second moment as prediction data; on the basis of the prediction data and the plurality of pieces of physical field data at the first moment, determining a deviation between an inflow amount of a substance in a fluid state in each unit space in the specified space per unit time and an increment of the mass of the substance contained in each unit space per unit time as a first deviation; determining a deviation between the prediction data and each piece of physical field data at the second moment as a prediction deviation; and on the basis of at least the first deviation and the prediction deviation, training the prediction model.

Description

A model training method and task execution method and device Technical Field

[0001] The present disclosure relates to the fields of artificial intelligence and fluid mechanics, and in particular to a model training method and a task execution method and device. Background Art

[0002] Multiphase flow has widespread applications in industrial production and everyday life, such as in energy, chemical engineering, aerospace, and metallurgy, as well as in dust removal. Efficient and accurate prediction of the physical fields involved in multiphase flow (volume fraction, flow, pressure, and temperature) is crucial for optimizing the design of related processes.

[0003] Currently, numerical simulation is commonly used to obtain the physical field details of a mixed state under specific conditions (such as a physical state composed of a mixture of fluid and solid substances), but it often takes a long time.

[0004] Summary of the Invention

[0005] The present disclosure provides a model training method and a task execution method and device to partially solve the above-mentioned problems existing in the prior art. The present disclosure adopts the following technical solutions.

[0006] In a first aspect, the present disclosure provides a method for model training, comprising: obtaining a plurality of physical field data of a mixture state in a specified space at at least two moments, wherein the mixture state includes at least a substance in a fluid state, the specified space includes an inlet and an outlet, the mixture state enters the specified space from the inlet and leaves the specified space from the outlet, the at least two moments include a first moment and a second moment, the second moment being a moment after the first moment; inputting the plurality of physical field data at the first moment into a prediction model to be trained, so that the prediction model predicts the mixture state in the specified space based on the plurality of physical field data at the first moment. The multiple physical field data of the mixture state at the second moment are used as prediction data; based on the prediction data and the multiple physical field data of the mixture state at the first moment, the deviation between the inflow of the substance in the fluid state in the specified space into each unit space per unit time and the increase in the mass of the substance contained in each unit space per unit time is determined as a first deviation, and the deviation between the prediction data and the multiple physical field data of the mixture state in the specified space at the second moment is determined as a prediction deviation; the prediction model is trained based on at least the first deviation and the prediction deviation.

[0007] Optionally, in response to the fact that the mixture state also includes matter in a solid phase state; before training the prediction model based on the first deviation and the predicted deviation, the method also includes: determining, based on the predicted data and the multiple physical field data of the mixture state at the first moment, the deviation between the increment of the matter in the solid phase state in the specified space per unit time and the inflow of the matter in the solid phase state in the specified space per unit time, as a second deviation; training the prediction model at least based on the first deviation and the predicted deviation includes: training the prediction model with minimizing the sum of the first deviation, the second deviation and the predicted deviation as the optimization goal.

[0008] Optionally, the prediction model includes a first network layer and a second network layer; the multiple physical field data at the first moment are input into the prediction model to be trained so that the prediction model predicts the multiple physical field data of the mixture state in the specified space at the second moment based on the multiple physical field data at the first moment as prediction data, including: inputting the physical field data of the mixture state at the first moment into the first network layer in the prediction model so that the first network layer predicts the volume fraction field data of the mixture state in the specified space at the second moment based on the multiple physical field data of the mixture state at the first moment; inputting the volume fraction field data and the multiple physical field data of the mixture state at the first moment into the second network layer in the prediction model so that the second network layer predicts other physical field data of the mixture state in the specified space at the second moment based on the multiple physical field data of the mixture state at the first moment and the volume fraction field data, wherein the other physical field data are the physical field data other than the volume fraction field data of the multiple physical field data.

[0009] Optionally, the obtaining of multiple physical field data of the mixture state in the specified space at at least two moments includes: obtaining the multiple physical field data of the mixture state in the specified space at the at least two moments under set physical conditions; inputting the multiple physical field data of the first moment into the prediction model to be trained so that the prediction model predicts the multiple physical field data of the mixture state in the specified space at the second moment as the prediction data based on the multiple physical field data of the first moment, including: inputting the multiple physical field data of the mixture state at the first moment and the conditional parameters corresponding to the set physical conditions into the prediction model so that the prediction model predicts the multiple physical field data of the mixture state in the specified space at the second moment under the set physical conditions as the prediction data based on the multiple physical field data of the mixture state at the first moment.

[0010] Optionally, the condition parameters corresponding to the set physical conditions include one or more of the temperature of the designated space, the boundary conditions of the designated space, geometric parameters, and material properties.

[0011] Optionally, the plurality of physical field data include volume fraction field data, and one or more of velocity field data, pressure field data, and temperature field data of each phase in the mixture state.

[0012] Optionally, the multiple physical field data also include flow rate data and pressure data.

[0013] Optionally, the method also includes: evenly dividing the designated space into a plurality of unit spaces, the unit time being the time interval between the second moment and the first moment; obtaining a plurality of physical field data of the mixed state at at least two moments in the designated space includes: obtaining the plurality of physical field data of the mixed state in each of the unit spaces in the designated space at at least two moments.

[0014] In the second aspect, the present disclosure provides a task execution method, including: obtaining multiple physical field data of a target mixture state in a target designated space at a current moment, the target mixture state includes at least a substance in a fluid state, the target designated space includes an inlet and an outlet, the target mixture state enters the target designated space from the inlet and leaves the target designated space from the outlet; inputting the multiple physical field data into a pre-trained prediction model so that the prediction model predicts the multiple physical field data of the target mixture state in the target designated space at the next moment based on the multiple physical field data, and the prediction model is trained by the model training method in the above-mentioned first aspect; executing the task according to the predicted multiple physical field data of the target mixture state in the target designated space at the next moment.

[0015] In a third aspect, the present disclosure provides a device for model training, comprising: an acquisition module for acquiring multiple physical field data of a mixture state in a specified space at at least two moments, wherein the mixture state includes at least a substance in a fluid state, the specified space includes an inlet and an outlet, the mixture state enters the specified space from the inlet and leaves the specified space from the outlet, the at least two moments include a first moment and a second moment, and the second moment is a moment after the first moment; a prediction module for inputting the multiple physical field data of the first moment into a prediction model to be trained, so that the prediction model predicts the mixture state in the specified space based on the multiple physical field data of the first moment. The multiple physical field data of the mixture state at the second moment are used as prediction data; a determination module is used to determine, based on the prediction data and the multiple physical field data of the mixture state at the first moment, the deviation between the inflow of the substance in the fluid state in the specified space into each unit space per unit time and the increment of the mass of the substance contained in each unit space per unit time, as a first deviation, and to determine the deviation between the prediction data and the multiple physical field data of the mixture state in the specified space at the second moment as a prediction deviation; a training module is used to train the prediction model based on at least the first deviation and the prediction deviation.

[0016] Optionally, in response to the fact that the mixture state also includes matter in a solid phase state; before training the prediction model based on the first deviation and the predicted deviation, the determination module is also used to determine, based on the predicted data and the multiple physical field data of the mixture state at the first moment, the deviation between the increment of the matter in the solid phase state in the specified space per unit time and the inflow of the matter in the solid phase state in the specified space per unit time, as a second deviation; the training module is specifically used to train the prediction model with the optimization goal of minimizing the sum of the first deviation, the second deviation and the predicted deviation.

[0017] Optionally, the prediction model includes a first network layer and a second network layer; the prediction module is specifically used to input the physical field data of the mixture state at the first moment into the first network layer in the prediction model, so that the first network layer predicts the volume fraction field data of the mixture state in the specified space at the second moment based on the multiple physical field data of the mixture state at the first moment; input the volume fraction field data and the multiple physical field data of the mixture state at the first moment into the second network layer in the prediction model, so that the second network layer predicts other physical field data of the mixture state in the specified space at the second moment based on the multiple physical field data and the volume fraction field data of the mixture state at the first moment, wherein the other physical field data are the physical field data other than the volume fraction field data of the multiple physical field data.

[0018] Optionally, the acquisition module is specifically used to acquire the multiple physical field data of the mixture state in the designated space at the at least two moments under set physical conditions; the prediction module is specifically used to input the multiple physical field data of the mixture state at the first moment and the condition parameters corresponding to the set physical conditions into the prediction model, so that the prediction model predicts the multiple physical field data of the mixture state in the designated space at the second moment under the set physical conditions based on the multiple physical field data of the mixture state at the first moment, as the predicted data.

[0019] Optionally, the condition parameters corresponding to the set physical conditions include one or more of the temperature of the designated space, the boundary conditions of the designated space, geometric parameters, and material properties.

[0020] Optionally, the plurality of physical field data include volume fraction field data, and one or more of velocity field data, pressure field data, and temperature field data of each phase in the mixture state.

[0021] Optionally, the multiple physical field data also include flow rate data and pressure data.

[0022] Optionally, the acquisition module is also used to: evenly divide the designated space into a plurality of unit spaces, the unit time being the time interval between the second moment and the first moment; and acquire the plurality of physical field data of the mixed state in each of the unit spaces in the designated space at the at least two moments.

[0023] In a fourth aspect, the present disclosure provides a task execution device, comprising: an acquisition module for acquiring multiple physical field data of a target mixture state in a target designated space at a current moment, wherein the target mixture state includes at least a substance in a fluid state, and the target designated space includes an inlet and an outlet, and the target mixture state enters the target designated space from the inlet and leaves the target designated space from the outlet; a prediction module for inputting the multiple physical field data into a pre-trained prediction model so that the prediction model predicts the multiple physical field data of the target mixture state in the target designated space at the next moment based on the multiple physical field data, and the prediction model is trained by the model training method of the first aspect mentioned above; an execution module for executing a task based on the predicted multiple physical field data of the target mixture state in the target designated space at the next moment.

[0024] In a fifth aspect, the present disclosure provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the model training method of the first aspect or the task execution method of the second aspect.

[0025] In a sixth aspect, the present disclosure provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the model training method of the first aspect or the task execution method of the second aspect is implemented.

[0026] At least one of the above-mentioned technical solutions adopted in the present disclosure can achieve the following beneficial effects: As can be seen from the above-mentioned method, when applying the trained prediction model to the actual task execution process, compared with existing technologies, it does not require excessive time to step-by-step derive the various physical field data of the target mixed state in the target designated space at the next moment, thereby improving the efficiency of predicting physical field data. Because the prediction model is trained during the training phase based on the condition that the mass of the fluid state of the mixed state between the first moment and the second moment meets the mass constraint, the accuracy of the physical field data predicted by the prediction model in actual applications can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the drawings:

[0028] FIG1 is a flow chart of a model training method provided by the present disclosure;

[0029] FIG2 is a flow chart of a task execution method provided by the present disclosure;

[0030] FIG3 is a schematic diagram of a model training device provided by the present disclosure;

[0031] FIG4 is a schematic diagram of a task execution device provided by the present disclosure;

[0032] FIG5 is a schematic structural diagram of an electronic device provided by the present disclosure. DETAILED DESCRIPTION

[0033] To make the objectives, technical solutions, and advantages of the present disclosure more clear, the technical solutions of the present disclosure will be clearly and completely described below in conjunction with the specific embodiments of the present disclosure and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present disclosure.

[0034] The following is a detailed description of the technical solutions provided by various embodiments of the present disclosure in conjunction with the accompanying drawings. FIG1 is a flow chart of a model training method provided in the present disclosure, which includes the following steps.

[0035] S101: Acquire physical field data of a mixed state at at least two moments in a specified space.

[0036] The at least two moments include a first moment and a second moment, and the second moment is a moment after the first moment. For example, the second moment may be the next moment of the first moment. The time interval between the first moment and the second moment may be referred to as a unit time, which may be set according to actual conditions, and the present disclosure does not impose any restrictions on this. Exemplarily, the physical field data of the mixed state in a specified space under a certain time length (such as 1 minute) may be obtained. For example, each physical field data may be sampled with 1 second as the unit time to obtain 60 sampling results. These 60 sampling results are the physical field data at each moment, and every two adjacent sampling results (physical field data at two moments) may be used as a group of training samples.

[0037] In addition, in order to improve the generalization ability of the trained prediction model, physical field data of a variety of different mixture states can be obtained as training samples in the sample set. In some examples, physical field data of mixtures of gas-liquid two-phase flow, gas-solid two-phase flow, liquid-solid two-phase flow, and liquid-liquid two-phase flow can be obtained as training samples. Furthermore, mixture states of different proportions can be included in the training samples. For example, for a mixture state including gas and liquid, a training sample with a gas:liquid ratio of 1:10, a training sample with a gas:liquid ratio of 1:100, and so on can be included. The present disclosure is not limited to this.

[0038] The execution subject of the model training method provided in the present disclosure can be a terminal device such as a desktop computer, a laptop computer, or a server. For the sake of convenience, the present disclosure only uses the terminal device as the execution subject to illustrate the provided model training method.

[0039] Since the solution provided by the present disclosure is to train a prediction model to achieve fast and accurate prediction of physical field data through the trained prediction model, before starting to train the prediction model, data for training the prediction model must be obtained.

[0040] Therefore, in the present disclosure, a terminal device can obtain various physical field data of a mixed state in a specified space at a first moment. The specified space here can refer to a space having at least two openings, for example, at least one inlet and at least one outlet. Mixed states can flow into and out of the specified space. Therefore, this process can be regarded as continuously "blowing air" from the inlet into the specified space. A portion of the mixed state will flow into the specified space, and at the same time, a portion of the mixed state will flow out of the specified space.

[0041] Since the solution provided by the present disclosure is mainly intended to predict physical field data in the field of multiphase flow, the mixture state at least includes substances in a fluid state (such as gas or liquid), and in addition, the mixture state may also include substances in a solid state.

[0042] In physics, matter has four states or phases: solid, liquid, gas, and plasma. If electromagnetic properties are ignored, the plasma phase can also be included in the gas phase. The flow of a single-phase substance is called a single-phase flow, and the flow of two evenly mixed gases or liquids also belongs to a single-phase flow. The flow of a mixture of substances with two or more phases at the same time is a two-phase or multi-phase flow. In multiphase flow dynamics, the so-called phases are distinguished not only by the state of the substance, but also by chemical composition, size, and shape, that is, substances with different chemical compositions, different sizes, and different shapes may belong to different phases. The mixed state described in the present disclosure includes at least a mixture of substances with two or more phases.

[0043] Common two-phase and multiphase flows include gas-liquid two-phase flow, gas-solid two-phase flow, liquid-solid two-phase flow, liquid-liquid two-phase flow, gas-liquid-liquid, gas-liquid-solid, and liquid-liquid-solid multiphase flow. For example, processes or technologies widely used in the power, energy, metallurgy, grain processing, and chemical industries, such as pneumatic conveying, airflow drying, coal powder combustion, catalytic cracking of petroleum, and fluidized bed roasting of minerals, all utilize gas-solid two-phase flow. Engineering examples of liquid-liquid two-phase flow include oil-water two-phase flow in oilfield extraction and surface transportation, separation, and sewage disposal, the flow of emulsions in chemical processes, and the flow of large amounts of liquid-liquid mixtures during material purification and extraction.

[0044] It should also be pointed out that the mixture state mentioned in the present disclosure is not a physical state composed of fixed atoms or molecules. Just like the "blowing" process mentioned above, the mixture state in the present disclosure is a flow state, and the physical field data to be predicted in the present disclosure can be understood as the physical field data of this part of the mixture state enclosed in the specified space.

[0045] Among them, the physical field data mentioned here can be real data collected by measuring instruments, or it can be simulated data constructed through simulation. The reason why it is different physical field data is that considering the interaction between mixed states, two-phase and multiphase flow analysis is very complicated. Each physical field data corresponds to a different physical dimension, and the acquisition of different physical field data can help with subsequent analysis and research on mixed states. The physical field data mentioned in this disclosure may include temperature field data, velocity field data, pressure field data, volume fraction field, etc. Different physical field data actually reflect the physical properties of the mixture state at different positions in the specified space. For example, the physical field data can be 128ⅹ128ⅹ128 data. Therefore, when the physical field data is temperature field data, then the 128ⅹ128ⅹ128 temperature field data can be understood as the temperature value of each point on the 128ⅹ128ⅹ128 points uniformly distributed in the specified space. Similarly, for pressure field data, the 128ⅹ128ⅹ128 pressure field data can be understood as the pressure value of each point on the 128ⅹ128ⅹ128 points uniformly distributed in the specified space. It should be noted that the above 128×128×128 is just an example, and the designated space can also be divided into other sizes, such as 64×64×64, 256×256×128, etc., which is not limited in this application. The following description will be made using 128×128×128 as an example.

[0046] Exemplarily, the physical field data at each moment may include a 1-dimensional volume fraction field, a 6-dimensional velocity field (corresponding to 2 phases, 3 directions for each phase), a 1-dimensional pressure field, and a 2-dimensional temperature field (corresponding to two phases). Then the dimension of the physical field data at each moment is 10ⅹ128ⅹ128ⅹ128.

[0047] It's important to note that to improve subsequent prediction model training, the mass conservation principle must be adhered to by the mixture state within the specified space. That is, if no mixture state flows out of the specified space's outlet, then the amount of mixture state flowing into or out of the specified space should correspond to the increase or decrease in the mass of the mixture state within the specified space.

[0048] Furthermore, in practical applications, the physical field data acquired for the mixture state in a specified space at the first moment may be the physical field data under set physical conditions. The set physical conditions may take various specific forms, such as specific air pressure, temperature, flow rate, etc. For example, when the set physical condition is that the mixture state is at a certain temperature, the physical field data acquired in the specified space at the first moment when the mixture state is at that temperature.

[0049] S102: Inputting the physical field data at the first moment into the prediction model to be trained, so that the prediction model predicts the physical field data of the mixed state in the specified space at the second moment based on the physical field data at the first moment as prediction data.

[0050] After obtaining the above-mentioned physical field data, the terminal device can input the physical field data at the first moment into the prediction model to be trained. The prediction model will predict the physical field data of the mixed state in the specified space at the second moment as the prediction data.

[0051] When predicting the various physical field data at the second moment, the volume fraction field data of the mixture state in the specified space at the second moment can be first predicted. Based on the predicted volume fraction field data and the various physical field data at the first moment, other physical field data of the mixture state in the specified space at the second moment can be predicted. The volume fraction field data can be understood as representing the volume of the mixture state at each location in the specified space.

[0052] The prediction model in the present disclosure may include a first network layer and a second network layer, and the two network layers may specifically be convolutional neural networks.

[0053] On this basis, the specific process of the terminal device inputting the above-mentioned physical field data into the prediction model to be trained can be to input the physical field data of the mixture state at the first moment into the first network layer of the prediction model. The first network layer will predict the volume fraction field data of the mixture state in the specified space at the second moment based on the physical field data of the mixture state at the first moment. Then, the predicted volume fraction field data and the obtained physical field data of the mixture state at the first moment are input into the second network layer of the prediction model. The second network layer will predict other physical field data of the mixture state in the specified space at the second moment except the volume fraction field data based on the physical field data of the mixture state at the first moment and the volume fraction field data. The other physical field data mentioned here can be understood as physical field data other than volume fraction field data, such as temperature and pressure.

[0054] The dimension of the physical field data at the first moment mentioned above is 10ⅹ128ⅹ128ⅹ128 as an example for explanation. The data input to the first network layer can be data of this 10ⅹ128ⅹ128ⅹ128 dimension. The first network layer can include two channels, one channel is used to predict whether there is a solid phase, and the other channel is used to predict the volume fraction. Finally, the volume fraction field is obtained through the results of these two channels, with a dimension of 1ⅹ128ⅹ128ⅹ128. In addition to the physical field data at the first moment, the data input to the second network layer also includes the volume fraction field data output by the first network layer. Therefore, the corresponding data dimension is 11ⅹ128ⅹ128ⅹ128. The output of the second network layer is other physical fields except the volume fraction field at the next moment, with a dimension of 9ⅹ128ⅹ128ⅹ128.

[0055] In some cases, the data input to the prediction model to be trained may include other auxiliary physical field data such as flow rate and pressure in addition to the main physical field data such as volume fraction field, velocity field, pressure field, and temperature field. The physical field data acquired in step S101 includes at least the volume fraction field and another main physical field data.

[0056] S103: Based on the predicted data and the physical field data of the mixed state at the first moment, determine the deviation between the inflow of the substance in the fluid state in the designated space into each unit space in the designated space per unit time and the increase in the mass of the substance contained in each unit space per unit time as the first deviation, and determine the deviation between the predicted data and the physical field data of the mixed state in the designated space at the second moment as the predicted deviation.

[0057] In the process of training the above-mentioned prediction model, in addition to using conventional supervised learning (such as prediction deviation), the optimization goal of mass constraint can also be introduced. That is, since it has been stipulated above that the inflow and outflow of the mixture state in the specified space must satisfy the law of conservation of mass with the increase or decrease of the mass of the mixture state in the specified space, then in practice, the inflow / outflow of the substance in the fluid state in the specified space in each unit space of the specified space should be the same as the increase / decrease in the mass of the substance contained in each unit space.

[0058] After obtaining the above-mentioned prediction data, the deviation between the prediction data and the physical field data of each tag of the mixture state in the specified space at the second moment can be determined as the prediction deviation. At the same time, the terminal device also needs to determine, based on the prediction data and the physical field data of the mixture state at the first moment input into the prediction model, the deviation between the flow rate of the fluid substance in the specified space per unit time and the increase in the mass of the substance contained in each unit time in the specified space as the first deviation.

[0059] The physical field data of each tag mentioned here may refer to the real physical field data of a substance in a fluid state in a specified space at the second moment, such as the physical field data at the second moment obtained in step S101.

[0060] The unit spaces mentioned above can be understood as evenly dividing a designated space into various small spaces. That is, the unit spaces in this application can be understood as a concept similar to "microelement". For example, evenly dividing a designated space into various unit spaces can be the above-mentioned division of the designated space into 128×128×128 unit spaces.

[0061] S104: Training the prediction model at least based on the first deviation and the prediction deviation.

[0062] The terminal device can determine the loss value corresponding to the first deviation through the first deviation, and determine the corresponding loss value based on the predicted deviation, and then train the prediction model based on the determined loss value.

[0063] Specifically, the loss value determined can be minimized as the optimization goal to train the prediction model. Therefore, it can also be understood that minimizing the sum of the first deviation and the prediction deviation is the optimization goal to train the prediction model.

[0064] Among them, determine the loss value Loss corresponding to the first deviation f This can be achieved through the following formula:

[0065] Where ε f,iRepresents the volume ratio of the substance in the fluid state in the mixture state in the specified space, ρ f,i Represents the density of the substance in the fluid state in the specified space, u f,i Represents the velocity of the material flowing into the fluid state in the specified space.

[0066] It can be seen from the above formula that the inflow of material in a fluid state into each unit space in a specified space is consistent with the mass increment of the material contained in each unit space, that is, the first deviation should be close to 0. It should be pointed out that the inflow mentioned in this application can be positive or negative, that is, when the inflow is positive, it represents the net inflow of material into the unit space, and when the inflow is negative, it represents the net outflow of material from the unit space. Correspondingly, the mass increment mentioned in this application can also be positive or negative, that is, when the mass increment is positive, it indicates an increase in mass, and when the mass increment is negative, it indicates a decrease in mass.

[0067] In addition, since it has been stipulated above that the inflow and outflow of the mixed state in the specified space needs to be consistent with the mass increment of the substance contained in the specified space, then in actual applications, in some examples, the mixed state may include solid-phase substances, and the inflow and outflow of the substance in the solid-phase state in the specified space also needs to be consistent with the mass increment of the substance contained in the specified space.

[0068] On this basis, if the mixture includes solid matter, during model training, the deviation between the increment of the solid-phase matter in the specified space per unit time and the inflow of the solid-phase matter in the specified space per unit time can be determined based on the predicted data and the physical field data of the mixture at the first moment, as the second deviation. The prediction model is then trained with the optimization goal of minimizing the sum of the first deviation, the second deviation, and the predicted deviation.

[0069] Therefore, minimizing the second deviation is to ensure that the inflow and outflow of substances in the solid phase state are consistent with the mass increment of the solid phase substances contained in the specified space.

[0070] In the actual training process, the terminal device needs to determine the loss value corresponding to the second deviation, and then train the prediction model based on this loss value and the loss values ​​corresponding to other deviations (i.e., the first deviation and the prediction deviation). The loss value corresponding to the second deviation can be determined by the following formula: s =|∑ i m s,i -(M+ΔM)|

[0071] Where m s,iIt represents the mass of the substance in the solid phase at position i (unit space i) in the specified space at the second moment, M represents the total mass of the substance in the solid phase in the specified space at the first moment, and ΔM represents the change in the total mass of the substance in the solid phase in the specified space between the first moment and the second moment, which is determined by the flow rate at the inlet and outlet.

[0072] Furthermore, since the acquired physical field data is obtained under set physical conditions, the terminal device can actually input the acquired physical field data and the conditional parameters corresponding to the set physical conditions into a prediction model. The prediction model can then predict the physical field data of the mixed state in a specified space at a second moment under the set physical conditions based on the physical field data of the mixed state at a first moment. In other words, if the acquired physical field data is obtained under set physical conditions, the physical field data predicted by the prediction model will correspond to the set physical conditions.

[0073] The conditional parameters corresponding to the aforementioned set physical conditions can be determined by specific physical conditions. For example, if the set physical condition refers to a specific temperature of the mixture state in a specified space, then the conditional parameters can refer to the specific temperature value specified in the specified space. Conditional parameters can also include boundary conditions, geometric parameters, material properties, etc. of the specified space.

[0074] Furthermore, when the above-mentioned prediction model includes a first network layer and a second network layer, the terminal device can input the acquired physical field data of the mixture state at the first moment and the conditional parameters corresponding to the set physical conditions into the first network layer in the prediction model to predict the volume fraction field data of the mixture state in the specified space at the second moment.

[0075] Then, the volume fraction field data of the mixture state in the specified space at the second moment, the physical field data of the mixture state obtained at the first moment, and the conditional parameters corresponding to the set physical conditions are input into the second network layer in the prediction model. The second network layer will predict other physical field data of the mixture state in the specified space at the second moment except the volume fraction field data.

[0076] This process can be understood as the second network layer predicting other physical field data at the second moment based on the obtained physical field data of the mixed state at the first moment, referring to the conditional parameters corresponding to the set physical conditions and the predicted volume fraction field data.

[0077] In some examples, in addition to the physical field data, the data input into the prediction model may also include data such as flow rate and pressure. The dimension of the physical field data at the first moment mentioned above is still 10ⅹ128ⅹ128ⅹ128 as an example for explanation. The data input to the first network layer may include the physical field data, flow rate and pressure at the first moment, with a dimension of 12ⅹ128ⅹ128ⅹ128, and output the volume fraction field with a dimension of 1ⅹ128ⅹ128ⅹ128. In addition to the physical field data, flow rate and pressure at the first moment, the data input to the second network layer also includes the volume fraction field data output by the first network layer, so the corresponding data dimension is 13ⅹ128ⅹ128ⅹ128. The output of the second network layer is other physical fields except the volume fraction field at the next moment, with a dimension of 9ⅹ128ⅹ128ⅹ128.

[0078] After completing the above model training, the model training can be applied to the execution process of the actual task, which will be described in detail below. Figure 2 is a flow chart of a task execution method provided in the present disclosure, which includes the following steps.

[0079] S201: Obtaining the physical field data of the mixed state in the specified space at the current moment.

[0080] When executing a task, it is necessary to first obtain the physical field data of the mixed state in the specified space at the current moment. The specific task to be performed can be determined by actual needs. For example, when performing a dust removal task, it is necessary to analyze the physical field data of various physical states in the specified space to determine the final method for dust removal in the specified space.

[0081] It should be noted that the designated space in step S201 may or may not be the same designated space as the designated space in step S101, and the mixed state in step S201 may or may not be the same mixed state as the mixed state in step S101. The model trained by the method shown in FIG1 has strong generalization capabilities and can be applied to designated spaces and / or mixed states that are different from the training samples to predict corresponding physical field data. The designated space in step S201 can be referred to as the target designated space, and the mixed state in step S201 can be referred to as the target mixed state.

[0082] S202: Inputting the physical field data into a pre-trained prediction model, so that the prediction model predicts the physical field data of the mixed state in the designated space at the next moment based on the physical field data.

[0083] The terminal device can input the acquired physical field data into the trained prediction model to obtain the physical field data of the mixed state in the specified space at the next moment. The various specific implementation forms of this process have been explained in detail in the content of the training stage above, so I will not repeat them here.

[0084] S203: Execute the task according to the predicted physical field data of the mixed state in the designated space at the next moment.

[0085] After predicting the physical field data of the mixture state in the specified space at the next moment, the task can be executed.

[0086] Therefore, it can be seen from the above content that, since the sample data obtained in the present disclosure can be input into the prediction model to train the prediction model, the generalization of the prediction model is better. When the trained prediction model is applied to the actual task execution process, compared with the existing technology, it does not need to spend too much time to derive the physical field data of the target mixed state in the target specified space at the next moment step by step, which improves the efficiency of predicting the physical field data. Moreover, since in the training stage, the prediction model is trained based on the condition that the mass of the substance in the fluid state in the mixed state before and after the first moment and the second moment meets the mass constraint, the accuracy of the physical field data predicted by the prediction model in actual application can be improved.

[0087] Moreover, this application fully considers the discrete characteristics existing in multiphase materials, both in the model training stage and in the actual use stage, and in the prediction process, it is necessary to first predict the volume fraction field data of the mixed state, and then predict other physical field data based on this. Therefore, it fully considers the correlation between the various physical field data, thereby ensuring the accuracy of the model prediction data.

[0088] The above are one or more implementation methods of the present disclosure. Based on the same idea, the present disclosure also provides corresponding model training devices and task execution devices, as shown in Figures 3 and 4.

[0089] FIG3 is a schematic diagram of a model training device provided by the present disclosure, which includes the following modules:

[0090] An acquisition module 301 is configured to acquire a plurality of physical field data of a mixture state in a specified space at at least two moments in time, wherein the mixture state includes at least a substance in a fluid state, the specified space includes an inlet and an outlet, the mixture state enters the specified space from the inlet and leaves the specified space from the outlet, and the at least two moments in time include a first moment in time and a second moment in time, wherein the second moment in time is a moment in time after the first moment in time;

[0091] A prediction module 302 is configured to input the plurality of physical field data at the first moment into a prediction model to be trained, so that the prediction model predicts the plurality of physical field data of the mixed state in the specified space at the second moment based on the plurality of physical field data at the first moment as prediction data;

[0092] a determination module 303 for determining, based on the predicted data and the plurality of physical field data of the mixture state at the first moment, a deviation between an inflow of the substance in the fluid state in the designated space per unit time in each unit space of the designated space and an increment of the mass of the substance contained in each unit space per unit time, as a first deviation, and determining a deviation between the predicted data and the plurality of physical field data of the mixture state in the designated space at the second moment as a prediction deviation;

[0093] The training module 304 is configured to train the prediction model based at least on the first deviation and the prediction deviation.

[0094] Optionally, in response to the fact that the mixture state also includes matter in a solid phase state; before training the prediction model based on the first mass deviation and the prediction deviation, the determination module 303 is also used to determine, based on the prediction data and the multiple physical field data of the mixture state at the first moment, the deviation between the increment of the matter in the solid phase state in the specified space per unit time and the inflow of the matter in the solid phase state in the specified space per unit time, as a second deviation; the training module 304 is specifically used to train the prediction model with the optimization goal of minimizing the sum of the first deviation, the second deviation and the prediction deviation.

[0095] Optionally, the prediction model includes a first network layer and a second network layer; the prediction module 302 is specifically used to input the physical field data of the mixture state at the first moment into the first network layer in the prediction model, so that the first network layer predicts the volume fraction field data of the mixture state in the specified space at the second moment based on the multiple physical field data of the mixture state at the first moment; input the volume fraction field data and the multiple physical field data of the mixture state at the first moment into the second network layer in the prediction model, so that the second network layer predicts other physical field data of the mixture state in the specified space at the second moment based on the multiple physical field data and the volume fraction field data of the mixture state at the first moment, wherein the other physical field data are the physical field data other than the volume fraction field data of the multiple physical field data.

[0096] Optionally, the acquisition module 301 is specifically used to acquire the multiple physical field data of the mixture state in the designated space at the at least two moments under set physical conditions; the prediction module 302 is specifically used to input the multiple physical field data of the mixture state at the first moment and the condition parameters corresponding to the set physical conditions into the prediction model, so that the prediction model predicts the multiple physical field data of the mixture state in the designated space at the second moment under the set physical conditions based on the multiple physical field data of the mixture state at the first moment, as the predicted data.

[0097] Optionally, the condition parameters corresponding to the set physical conditions include one or more of the temperature of the designated space, the boundary conditions of the designated space, geometric parameters, and material properties.

[0098] Optionally, the plurality of physical field data include volume fraction field data, and one or more of velocity field data, pressure field data, and temperature field data of each phase in the mixture state.

[0099] Optionally, the multiple physical field data also include flow rate data and pressure data.

[0100] Optionally, the acquisition module 301 is also used to: evenly divide the designated space into a plurality of unit spaces, the unit time being the time interval between the second moment and the first moment; and acquire the plurality of physical field data of the mixed state in each unit space in the designated space at the at least two moments.

[0101] FIG4 is a schematic diagram of a task execution device provided by the present disclosure, comprising the following modules:

[0102] An acquisition module 401 is configured to acquire a plurality of physical field data of a target mixture state in a target designated space at a current moment, wherein the target mixture state includes at least a substance in a fluid state, and the target designated space includes an inlet and an outlet, the target mixture state enters the target designated space from the inlet and leaves the target designated space from the outlet;

[0103] A prediction module 402 is configured to input the plurality of physical field data into a pre-trained prediction model, so that the prediction model predicts the plurality of physical field data of the target mixed state in the target designated space at a next moment based on the plurality of physical field data, wherein the prediction model is trained using the above-mentioned model training method;

[0104] The execution module 403 is used to execute a task according to the multiple physical field data of the target mixed state in the target designated space at the next moment that are predicted.

[0105] The present disclosure also provides a computer-readable storage medium storing a computer program, which can be used to execute the model training method provided in FIG1 or the task execution method provided in FIG2. The computer-readable storage medium may include a non-volatile computer-readable storage medium.

[0106] The present disclosure also provides a schematic structural diagram of an electronic device corresponding to Figure 1 or Figure 2, as shown in Figure 5. As shown in Figure 5, at the hardware level, the electronic device includes a processor and a memory. Among them, the memory may include internal memory and non-volatile memory, and of course may also include hardware required for other services, such as internal buses, network interfaces, etc. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the model training method described in Figure 1 or the task execution method described in Figure 2.

[0107] Of course, in addition to software implementation, this disclosure does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0108] In the 1990s, technological improvements could be clearly distinguished as either hardware improvements (for example, improvements to circuit structures like diodes, transistors, and switches) or software improvements (improvements to process flows). However, with the advancement of technology, many process flow improvements today can now be considered direct improvements to hardware circuit structures. Designers almost always create the corresponding hardware circuit structure by programming the improved process flow into the hardware circuit. Therefore, it cannot be said that a process flow improvement cannot be implemented using hardware modules. For example, a programmable logic device (PLD), such as a field programmable gate array (FPGA), is an integrated circuit whose logical function is determined by user programming. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to hire a chip manufacturer to design and manufacture a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly done using "logic compiler" software. This is similar to the software compiler used when developing programs. Before compilation, the original code must also be written in a specific programming language, called a hardware description language (HDL). There is not just one HDL, but many, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art will also understand that by simply programming the method flow in one of these hardware description languages ​​and then programming it into an integrated circuit, a hardware circuit that implements the logic method flow can be easily obtained.

[0109] The controller can be implemented in any suitable manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to implementing the controller in a purely computer-readable program code format, the controller can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules that implement the method and structures within the hardware component.

[0110] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0111] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing the present disclosure, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0112] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0113] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0114] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0115] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0116] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0117] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0118] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0119] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0120] Those skilled in the art will appreciate that embodiments of the present disclosure may be provided as methods, systems, or computer program products. Thus, the present disclosure may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] The present disclosure may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present disclosure may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0122] The various embodiments of this disclosure are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so their description is relatively simple. For relevant portions, refer to the description of the method embodiments.

[0123] The foregoing is merely an embodiment of the present disclosure and is not intended to limit the present disclosure. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure are intended to be included within the scope of the claims of the present disclosure.

Claims

1. A model training method, characterized in that: include: Acquire a plurality of physical field data of a mixture state in a specified space at at least two moments, wherein the mixture state includes at least a substance in a fluid state, the specified space includes an inlet and an outlet, the mixture state enters the specified space from the inlet and leaves the specified space from the outlet, the at least two moments include a first moment and a second moment, and the second moment is a moment after the first moment; Inputting the plurality of physical field data at the first moment into a prediction model to be trained, so that the prediction model predicts the plurality of physical field data of the mixed state in the specified space at the second moment according to the plurality of physical field data at the first moment as prediction data; According to the predicted data and the multiple physical field data of the mixed state at the first moment, determining the deviation between the inflow of the substance in the fluid state in the designated space into each unit space in the designated space per unit time and the increment of the mass of the substance contained in each unit space per unit time as a first deviation, and determining the deviation between the predicted data and the multiple physical field data of the mixed state in the designated space at the second moment as a predicted deviation; The prediction model is trained based on at least the first deviation and the prediction deviation.

2. The method according to claim 1, characterized in that In response to the mixture state further comprising a substance in a solid phase; before training the prediction model according to the first deviation and the prediction deviation, the method further comprises: Determine, according to the predicted data and the plurality of physical field data of the mixed state at the first moment, a deviation between an increment per unit time of the substance in the solid phase in the designated space and an inflow per unit time of the substance in the solid phase in the designated space as a second deviation; The step of training the prediction model at least according to the first deviation and the prediction deviation includes: The prediction model is trained with the optimization goal of minimizing the sum of the first deviation, the second deviation and the prediction deviation.

3. The method according to claim 1, characterized in that The prediction model includes a first network layer and a second network layer; The step of inputting the plurality of physical field data at the first moment into a prediction model to be trained, so that the prediction model predicts the plurality of physical field data of the mixed state in the specified space at the second moment according to the plurality of physical field data at the first moment as prediction data, comprises: The physical field data of the mixed state at the first time is input into the first network layer in the prediction model, so that the first network layer can predict the mixed state according to the multiple physical field data of the mixed state at the first time. Physical field data, predicting volume fraction field data of the mixture state in the specified space at the second moment; The volume fraction field data and the multiple physical field data of the mixed state at the first moment are input into the second network layer in the prediction model, so that the second network layer predicts other physical field data of the mixed state in the specified space at the second moment based on the multiple physical field data of the mixed state at the first moment and the volume fraction field data, wherein the other physical field data are the physical field data other than the volume fraction field data of the multiple physical field data.

4. The method according to claim 1, characterized in that The obtaining of a plurality of physical field data of a mixed state at at least two moments in a specified space comprises: Acquire the plurality of physical field data of the mixed state in the specified space at the at least two moments under set physical conditions; The step of inputting the plurality of physical field data at the first moment into a prediction model to be trained, so that the prediction model predicts the plurality of physical field data of the mixed state in the specified space at the second moment according to the plurality of physical field data at the first moment as prediction data, comprises: The multiple physical field data of the mixed state at the first moment and the condition parameters corresponding to the set physical conditions are input into the prediction model, so that the prediction model predicts the multiple physical field data of the mixed state in the specified space under the set physical conditions at the second moment based on the multiple physical field data of the mixed state at the first moment as the prediction data.

5. The method according to claim 4, characterized in that The condition parameters corresponding to the set physical conditions include one or more of the temperature of the designated space, the boundary conditions of the designated space, geometric parameters, and material properties.

6. The method according to claim 1, characterized in that The plurality of physical field data include volume fraction field data, and one or more of velocity field data of each phase in the mixture state, pressure field data, and temperature field data of each phase in the mixture state.

7. The method according to claim 6, characterized in that The plurality of physical field data also includes flow rate data and pressure data.

8. The method according to claim 1, characterized in that The method further comprises: The designated space is evenly divided into a plurality of unit spaces, and the unit time is the time interval between the second moment and the first moment; The obtaining of a plurality of physical field data of a mixed state at at least two moments in a specified space comprises: Acquire the plurality of physical field data of the mixed state in each of the unit spaces in the designated space at the at least two moments.

9. A task execution method, characterized in that: include: Acquire multiple physical field data of a target mixed state in a target designated space at a current moment, wherein the target mixed state includes at least a substance in a fluid state, and the target designated space includes an inlet and an outlet, and the target mixed state enters the target designated space from the inlet and leaves the target designated space from the outlet; Inputting the multiple physical field data into a pre-trained prediction model so that the prediction model predicts the multiple physical field data of the target mixed state in the target specified space at the next moment based on the multiple physical field data, wherein the prediction model is trained by the method described in any one of claims 1 to 8 above; The task is executed according to the predicted multiple physical field data of the target mixed state in the target designated space at the next moment.

10. A device for model training, characterized in that: include: an acquisition module, used to acquire a plurality of physical field data of a mixture state in a specified space at at least two moments, wherein the mixture state includes at least a substance in a fluid state, the specified space includes an inlet and an outlet, the mixture state enters the specified space from the inlet and leaves the specified space from the outlet, the at least two moments include a first moment and a second moment, the second moment being a moment after the first moment; A prediction module, used for inputting the multiple physical field data at the first moment into a prediction model to be trained, so that the prediction model predicts the multiple physical field data of the mixed state in the specified space at the second moment according to the multiple physical field data at the first moment as prediction data; a determination module, for determining, based on the prediction data and the plurality of physical field data of the mixed state at the first moment, a deviation between an inflow of the substance in the fluid state in the designated space per unit time in each unit space of the designated space and an increment of the mass of the substance contained in each unit space per unit time, as a first deviation, and determining a deviation between the prediction data and the plurality of physical field data of the mixed state in the designated space at the second moment, as a prediction deviation; A training module is used to train the prediction model based on at least the first deviation and the prediction deviation.

11. The device according to claim 10, characterized in that In response to the mixed state also including a substance in a solid phase; Before training the prediction model according to the first deviation and the prediction deviation, the determination module is further used to determine, according to the prediction data and the multiple physical field data of the mixed state at the first moment, a deviation between the increment of the substance in the solid phase in the specified space per unit time and the inflow of the substance in the solid phase in the specified space per unit time as a second deviation; The training module is specifically used to minimize the first deviation, the second deviation and the predicted deviation. The prediction model is trained for the optimization objective.

12. The device according to claim 10, characterized in that The prediction model includes a first network layer and a second network layer; The prediction module is specifically used to input the physical field data of the mixed state at the first moment into the first network layer in the prediction model, so that the first network layer predicts the volume fraction field data of the mixed state in the specified space at the second moment according to the multiple physical field data of the mixed state at the first moment; The volume fraction field data and the multiple physical field data of the mixed state at the first moment are input into the second network layer in the prediction model, so that the second network layer predicts other physical field data of the mixed state in the specified space at the second moment based on the multiple physical field data of the mixed state at the first moment and the volume fraction field data, wherein the other physical field data are the physical field data other than the volume fraction field data of the multiple physical field data.

13. The device according to claim 10, characterized in that The acquisition module is specifically used to acquire the multiple physical field data of the mixed state in the specified space at the at least two moments under set physical conditions; The prediction module is specifically used to input the multiple physical field data of the mixed state at the first moment and the condition parameters corresponding to the set physical conditions into the prediction model, so that the prediction model predicts the multiple physical field data of the mixed state in the specified space under the set physical conditions at the second moment based on the multiple physical field data of the mixed state at the first moment, as the predicted data.

14. The device according to claim 13, characterized in that The condition parameters corresponding to the set physical conditions include one or more of the temperature of the designated space, the boundary conditions of the designated space, geometric parameters, and material properties.

15. The device according to claim 10, characterized in that The plurality of physical field data include volume fraction field data, and one or more of velocity field data of each phase in the mixture state, pressure field data, and temperature field data of each phase in the mixture state.

16. The device according to claim 15, characterized in that The plurality of physical field data also includes flow rate data and pressure data.

17. The device according to claim 10, characterized in that The acquisition module is also used for: The designated space is evenly divided into a plurality of unit spaces, and the unit time is the time interval between the second moment and the first moment; Acquire the plurality of physical field data of the mixed state in each of the unit spaces in the designated space at the at least two moments.

18. A task execution device, characterized in that: include: An acquisition module is used to acquire a plurality of physical field data of a target mixed state in a target designated space at a current moment, wherein the target mixed state includes at least a substance in a fluid state, and the target designated space includes an inlet and an outlet, and the target mixed state enters the target designated space from the inlet and leaves the target designated space from the outlet; A prediction module, used for inputting the multiple physical field data into a pre-trained prediction model, so that the prediction model predicts the multiple physical field data of the target mixed state in the target specified space at the next moment according to the multiple physical field data, wherein the prediction model is trained by the method according to any one of claims 1 to 8 above; An execution module is used to execute a task according to the predicted multiple physical field data of the target mixed state in the target specified space at the next moment.

19. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of claims 1 to 8 or 9 is implemented.

20. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method described in any one of claims 1 to 8 or 9 is implemented.

Images