Temperature calibration method and system for low-temperature liquid nitrogen biological container

Abstract

The invention discloses a temperature calibration method and system for a low-temperature liquid nitrogen biological container, and relates to the technical field of temperature sensors, and the method comprises the steps: inputting at least one frame of original array temperature into a pre-trained temperature calibration model group, and obtaining an output target calibration temperature; the temperature calibration model group comprises a basic calibration model and a physical calibration model, and the physical calibration model is used to obtain a target theoretical temperature set; determining a difference group based on an initial calibration temperature and the target theoretical temperature set, constructing a difference optimization model to obtain an optimization suggestion, and optimizing the basic calibration model and the physical calibration model according to the optimization suggestion; the method is established based on the physical principle and has good generalization ability for different container sizes, materials and liquid nitrogen flowing states, and the effect of improving temperature calibration in the low-temperature liquid nitrogen biological container is achieved. And meanwhile, the effects of quickly migrating the data module and improving the generalization of the data model are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature sensors, and in particular to a temperature calibration method and system for a low-temperature liquid nitrogen biological container. Background Art

[0002] In the field of low-temperature biological experiments, especially in biological containers using liquid nitrogen as a cooling medium, the accuracy of temperature control is crucial. Liquid nitrogen biological containers are often used to store biological samples, such as cells, tissues, and vaccines. These samples are extremely sensitive to temperature, and even slight temperature changes may cause irreversible damage to them. Therefore, accurate calibration of the temperature of low-temperature liquid nitrogen biological containers is the key to ensuring the safe preservation of biological samples. However, when the array temperature sensor is measuring temperature, there is a problem that the temperature data collected by different temperature measurement units at the same measurement temperature are not exactly the same, and the measurement temperatures obtained in different frame array temperatures for the same measurement temperature are also different, which leads to certain inaccuracies in the measurement results of the array temperature sensor.

[0003] The Chinese invention patent with application number 202010670631.X provides a temperature calibration method, device, equipment and storage medium. Obtain at least one frame of original array temperature collected by the temperature sensor; input the at least one frame of original array temperature into the pre-trained target temperature calibration model to obtain the output target calibration temperature; wherein the target temperature calibration model includes a target space calibration model and a target time calibration model, and performs temperature calibration from two dimensions of time and space, thereby improving the accuracy of temperature measurement.

[0004] Due to the low temperature environment of the cryogenic liquid nitrogen biological container, the original array temperature data contains noise or deviation, which reduces the data quality. Therefore, the accuracy of temperature calibration is reduced when calibrating in both time and space dimensions. In addition, due to the extremely narrow temperature range of liquid nitrogen, a large amount of training data cannot be generated. The original time and space calibration model is difficult to generalize to the low temperature liquid nitrogen environment, which further reduces the accuracy of temperature calibration. Summary of the invention

[0005] The present application solves the problem of low data quality and poor model generalization ability leading to reduced temperature calibration accuracy in the prior art by providing a temperature calibration method and system for a low-temperature liquid nitrogen biological container, thereby achieving the technical effect of improving temperature calibration in a low-temperature liquid nitrogen biological container.

[0006] The present application provides a temperature calibration method for a low-temperature liquid nitrogen biological container, the method comprising: Acquire at least one frame of original array temperature collected by a temperature sensor; input the at least one frame of original array temperature into a pre-trained temperature calibration model group to obtain an output target calibration temperature; The temperature calibration model group includes a basic calibration model and a physical calibration model, the at least one frame of original array temperature is input into the basic calibration model to obtain an output initial calibration temperature; a physical calibration model is constructed, and a target theoretical temperature set is obtained using the physical calibration model; A difference group is determined based on the initial calibration temperature and the target theoretical temperature set, a difference optimization model is constructed, the difference group is input into the difference optimization model to obtain optimization suggestions, and the basic calibration model and the physical calibration model are optimized according to the optimization suggestions.

[0007] Further, a physical calibration model is constructed, and a target theoretical temperature set is obtained using the physical calibration model, including: Obtaining basic parameters of the low-temperature liquid nitrogen biological container, establishing a three-dimensional geometric model according to the basic parameters, setting material properties and boundary conditions, establishing physical equations based on heat conduction theory and fluid mechanics theory, using a meshing tool to mesh the three-dimensional geometric model, and obtaining a target theoretical temperature value at each grid position through simulation calculation to form a target theoretical temperature set; The basic parameters include: the structure of the container, the material of the container, the basic physical properties of liquid nitrogen, the external environmental condition parameters, the initial temperature of the liquid nitrogen and the flow state parameters of the liquid nitrogen in the container.

[0008] Further, the at least one frame of original array temperature is input into a basic calibration model to obtain an output initial calibration temperature, including: The basic calibration model includes a spatial calibration model and a temporal calibration model. The spatial calibration model performs spatial temperature calibration on the at least one frame of original array temperature to obtain a spatial calibration temperature. The temporal calibration model performs temporal temperature calibration on the spatial calibration temperature to obtain the initial calibration temperature.

[0009] Further, the difference group includes a first difference group and a second difference group, and the optimization suggestion includes a first optimization suggestion and a second optimization suggestion; the first difference group is determined based on the initial calibration temperature and the target theoretical temperature set, the first difference group is input into the difference optimization model to obtain a first optimization suggestion, the physical calibration model is adjusted and optimized according to the first optimization suggestion, and the target theoretical temperature control set is obtained according to the adjusted physical calibration model; the second difference group is determined based on the initial calibration temperature, the target theoretical temperature set and the target theoretical temperature control set, the second difference group is input into the difference optimization model to obtain a second optimization suggestion; the basic calibration model and the physical calibration model are optimized based on the first optimization suggestion and the second optimization suggestion.

[0010] Furthermore, a first difference group is determined based on the initial calibration temperature and the target theoretical temperature set, including: calculating the difference between each theoretical temperature value in the target theoretical temperature set and the initial calibration temperature, comparing the similarity between the differences with a pre-set first similarity threshold, dividing the position corresponding to each theoretical temperature value whose similarity is not less than the first similarity threshold into the same area and marking it as a first difference area; dividing the position corresponding to each theoretical temperature value whose similarity is less than the first similarity threshold into the same area and marking it as a second difference area; and forming a triplet of the specific position, the target theoretical temperature value and the initial calibration temperature value in the first difference area and the second difference area to form a first difference group.

[0011] Further, determining a second difference group based on the initial calibration temperature, the target theoretical temperature set and the target theoretical temperature control set includes: calculating the difference between each theoretical temperature control value in the target theoretical temperature control set and the initial calibration temperature, and calculating the temperature change rate of the target theoretical temperature set and the target theoretical temperature control set corresponding area; The positions corresponding to each theoretical temperature comparison value whose similarity is not less than the second similarity threshold are divided into the same area and marked as the third similarity difference area; the positions corresponding to each theoretical temperature comparison value whose similarity is less than the second similarity threshold are divided into the same area and marked as the fourth similarity difference area; The positions corresponding to each theoretical temperature control value whose temperature change rate is not less than the change threshold are divided into the same area and marked as the third change difference area; the positions corresponding to each theoretical temperature control value whose temperature change rate is less than the change threshold are divided into the same area and marked as the fourth change difference area; The intersection area of ​​the third similar difference area and the third change difference area is marked as the third difference area, and the intersection area of ​​the fourth similar difference area and the fourth change difference area is marked as the fourth difference area; the specific positions, target theoretical temperature control values ​​and initial calibration temperature values ​​in the third difference area and the fourth difference area are grouped into a triplet to form a second difference group.

[0012] Furthermore, inputting the first difference group into the difference optimization model means inputting the triplet data corresponding to the first difference area in the first difference group into the difference optimization model; inputting the second difference group into the difference optimization model means inputting the triplet data corresponding to the third difference area and the fourth difference area in the second difference group into the difference optimization model.

[0013] Further, the physical calibration model is adjusted and optimized according to the first optimization suggestion, and a target theoretical temperature control set is obtained according to the adjusted physical calibration model, including: The first difference region is re-divided according to the first optimization suggestion and the basic parameters are adjusted to obtain all theoretical temperature values ​​in the re-divided first difference region to form a target theoretical temperature comparison set.

[0014] Furthermore, a difference optimization model is constructed, including: selecting a random forest regression model as a basic model for training, determining input features and output features of the difference optimization model, dividing the preprocessed historical data into a training set and a validation set, initializing model parameters using a random forest regressor in a machine learning library, and training the random forest regression model using the training set data; validating the random forest regression model using the validation set, and adjusting model parameters according to the validation results, and using the trained random forest regression model as the difference optimization model.

[0015] A temperature calibration system for a cryogenic liquid nitrogen biological container, the system comprising: A data acquisition and processing module is used to obtain the original array temperature data collected by the temperature sensor and perform preprocessing; A model group calibration module, the model group calibration module includes a basic calibration submodule and a physical calibration submodule, the basic calibration submodule is used to calibrate the original array temperature in space and time, and output the initial calibration temperature; the physical calibration submodule is used to establish a three-dimensional geometric model according to the basic parameters of the low-temperature liquid nitrogen biological container, set material properties and boundary conditions, perform simulation calculations based on heat conduction and fluid mechanics theory, and obtain a target theoretical temperature set; The difference analysis and optimization suggestion module is used to compare the initial calibration temperature and the target theoretical temperature set, determine the first difference group, input the first difference group into the difference optimization model, and obtain the first optimization suggestion; adjust the physical calibration model according to the first optimization suggestion, generate the target theoretical temperature control set, compare the initial calibration temperature, the target theoretical temperature set and the target theoretical temperature control set, determine the second difference group, input the second difference group into the difference optimization model, and obtain the second optimization suggestion; A model optimization and adjustment module, used for optimizing and adjusting the basic calibration model and the physical calibration model according to the first optimization suggestion and the second optimization suggestion; The difference optimization model building module is used to build the difference optimization model, including: selecting and training the random forest regression model as the basis of the difference optimization model, determining the input features and output features, using historical data for training and verification, adjusting the model parameters according to the verification results, and obtaining the final difference optimization model.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: A physical calibration model is constructed through the basic parameters of the low-temperature liquid nitrogen biological container to obtain the target theoretical temperature set. The initial calibration temperature and the target theoretical temperature set are compared to determine the difference group, and a difference optimization model is constructed. The difference group is input to obtain optimization suggestions. The basic calibration model and the physical calibration model are optimized according to the optimization suggestions, thereby achieving the effect of improving temperature calibration in the low-temperature liquid nitrogen biological container. It is based on physical principles and has good generalization capabilities for different container sizes, materials and liquid nitrogen flow states. When the container conditions or liquid nitrogen state change, the physical calibration model can quickly adjust and give a new target theoretical temperature set, thereby achieving the effect of quickly migrating data modules and improving the generalization of data models. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of the overall process of the temperature calibration method of the low-temperature liquid nitrogen biological container in an embodiment of the present invention; Figure 2 1 is a diagram showing the architecture of a temperature calibration system for a cryogenic liquid nitrogen biological container in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0020] Embodiment 1: Figure 1 As shown, a temperature calibration method for a low-temperature liquid nitrogen biological container, the method comprising: S100: acquiring at least one frame of original array temperature collected by a temperature sensor; inputting the at least one frame of original array temperature into a pre-trained temperature calibration model group to obtain an output target calibration temperature; The temperature calibration model group includes a basic calibration model and a physical calibration model, the at least one frame of original array temperature is input into the basic calibration model to obtain an output initial calibration temperature; a physical calibration model is constructed, and a target theoretical temperature set is obtained using the physical calibration model; S200: Determine a difference group based on the initial calibration temperature and the target theoretical temperature set, construct a difference optimization model, input the difference group into the difference optimization model, obtain optimization suggestions, and optimize the basic calibration model and the physical calibration model according to the optimization suggestions.

[0021] In some embodiments, the temperature sensor includes an array temperature sensor. Specifically, the array temperature sensor can be a thermal infrared array temperature sensor. A plurality of temperature measuring units are provided on the array temperature sensor, and the measured temperatures collected by the plurality of measuring units constitute the frame array temperature of the array temperature sensor. Each square is used to represent a temperature measuring unit, and the grayscale value in each square represents the original temperature value collected by each temperature measuring unit. It should be noted that the grayscale value is only a schematic explanation of the frame array temperature. In this embodiment, the frame array temperature can be represented in the form of an image, and the image can be a color image or a grayscale image, and the pixel value in the preset image area in the image represents the original temperature value. Of course, the frame array temperature can also be represented in the form of matrix data, and each matrix value represents the original temperature value.

[0022] In some embodiments, according to the size and shape of the liquid nitrogen biocontainer, a suitable low temperature temperature sensor is selected, and the sensor is ensured to have sufficient accuracy and stability, and the sensor is arranged at key positions of the liquid nitrogen biocontainer to obtain comprehensive temperature data. The sensor arrangement should take into account the factors of spatial distribution and time variation.

[0023] In some embodiments, constructing a physical calibration model includes: obtaining basic parameters of a low-temperature liquid nitrogen biological container, establishing a three-dimensional geometric model according to the basic parameters based on a modeling software, setting material properties and boundary conditions, establishing physical equations based on heat conduction theory and fluid mechanics theory, using a meshing tool to mesh the three-dimensional geometric model, and obtaining a target theoretical temperature value at each grid position through simulation calculation to form a target theoretical temperature set; The basic parameters include: the structure of the container, the material of the container, the basic physical properties of liquid nitrogen, the external environmental condition parameters, the initial temperature of the liquid nitrogen and the flow state parameters of the liquid nitrogen in the container.

[0024] In some embodiments, basic parameters of a low-temperature liquid nitrogen biological container are obtained, wherein the basic parameters include but are not limited to: structural parameters of the container, material of the container, basic physical properties of liquid nitrogen, external environmental condition parameters, initial temperature of liquid nitrogen, and flow state parameters of liquid nitrogen in the container; Specifically, the structural parameters of the container include but are not limited to: the shape, size, wall thickness, etc. of the container; the material of the container includes thermal conductivity, specific heat capacity and density at a specific temperature, etc., and these parameters are regarded as constants under specific conditions; the basic physical properties of the liquid nitrogen include boiling point, density at boiling point, specific heat capacity and thermal conductivity at a specific temperature; the external environmental condition parameters include the temperature and thermal radiation coefficient outside the container, etc. The temperature outside the container will change with time and space. When using the basic calibration model for temperature calibration, it is necessary to consider the change of the temperature outside the container to adjust the relevant parameters. The specific adjustment needs to be made according to the actual changes, and this application does not make specific restrictions; the initial temperature of the liquid nitrogen and the flow state parameters of the liquid nitrogen in the container need to be set and determined according to the actual situation. The flow state parameters of the liquid nitrogen in the container include flow velocity and turbulence intensity, etc., which change according to different experimental conditions or operating conditions. In a specific experiment, it is set according to the actual situation, and this application does not make specific restrictions.

[0025] In some embodiments, a three-dimensional geometric model is established according to the basic parameters based on modeling software (such as the Geometry module in ANSYS), and material properties and boundary conditions are set. In the ANSYS numerical simulation software, corresponding material properties such as thermal conductivity, specific heat capacity, etc. are set for the container and liquid nitrogen. Boundary conditions are applied according to actual conditions, such as the temperature outside the container, the initial temperature of the liquid nitrogen, thermal radiation conditions, etc. The thermal conductivity, specific heat capacity and other properties of the material will change with the change of temperature. For example, metals may show thermal conductivity properties that are completely different from those at room temperature at low temperatures; in the simulation of liquid nitrogen biological containers, the material properties of the container may be significantly affected due to the extremely low temperature of liquid nitrogen; therefore, it is necessary to set the material properties according to the actual working temperature. Different containers may be made of different materials, such as stainless steel, aluminum alloy, titanium alloy, etc., which have significant differences in thermal conductivity, specific heat capacity, etc. Therefore, the corresponding properties must be set according to the actual materials used during simulation. The purpose of numerical simulation is to predict actual physical phenomena as accurately as possible. In order to improve the accuracy of the simulation, all relevant parameters, including material properties, need to be input as accurately as possible. If the changes in material properties are ignored or inaccurate property values ​​are used, the simulation results may deviate greatly from the actual situation. You can also select or input material properties according to actual needs to ensure the accuracy of the simulation, allowing the nonlinear properties of materials to be defined according to changes in conditions such as temperature and pressure; in practical applications, experiments are needed to verify the accuracy of the simulation results. During the experiment, it may be found that some material properties deviate from expectations. At this time, corresponding adjustments need to be made in the simulation. Through continuous iterations of experiments and simulations, we can gradually approach the real physical phenomena.

[0026] In some embodiments, ANSYS numerical simulation software is a large-scale general-purpose finite element analysis software that integrates structure, fluid, electromagnetic field, acoustic field and thermal analysis. It is a widely used engineering simulation tool that can realize multi-physical field and multi-physical field coupling analysis, and provides data interfaces for multiple software to realize data sharing and exchange. Its pre-processing module provides powerful solid modeling and meshing tools. It is a prior art software well known to technicians in this field, and this application will not go into details here.

[0027] In some embodiments, based on the theory of heat conduction and the principles of fluid mechanics, physical equations are established to describe the temperature distribution and flow of liquid nitrogen in the container. These equations include the heat conduction equation, the Navier-Stokes equation (used to describe fluid flow) and the energy conservation equation; specifically, the main physical phenomena of liquid nitrogen in the container are clarified, including heat conduction and fluid flow, and based on the theory of heat conduction and the principles of fluid mechanics, corresponding equations are selected to describe these physical phenomena, and mathematical equations describing the temperature distribution and flow of liquid nitrogen are established, and the parameters required in the equations are identified, including material properties, geometric dimensions, fluid properties, etc. In this application, heat conduction equations, Navier-Stokes equations (used to describe fluid flow) and energy conservation equations are established.

[0028] Specifically, the heat conduction equation describes the relationship that the rate of change of temperature over time is proportional to the spatial second-order derivative of temperature (i.e., the rate of change of temperature gradient), and the specific parameters include thermal conductivity, mass density of the material, and specific heat capacity, wherein the specific heat capacity refers to the amount of heat required to increase the unit mass of the material by 1 degree; The Navier-Stokes equations describe the dynamic behavior of fluid flow, including the balance between inertial force, pressure gradient force, viscous force and external force. Specific parameters include the mass density, viscosity, flow rate and pressure of the fluid. The viscosity refers to the internal friction coefficient of the fluid, and the pressure refers to the pressure inside the fluid. The energy conservation equation describes the balance between the rate of change of energy in the system and the energy flow, heat conduction, external work and heat source. The specific parameters include all the parameters in the above heat conduction equation and Navier-Stokes equations as well as heat source terms (such as external heating or cooling).

[0029] In some embodiments, the geometric model is meshed using the meshing tool of ANSYS to ensure that the density and distribution of the mesh can accurately capture the characteristics of temperature changes and fluid flow; the previously established three-dimensional geometric model of the liquid nitrogen biological container is imported into ANSYS, and a meshing method (such as tetrahedron, hexahedron, sweep, etc.) is selected according to the actual calculation requirements of the geometric model, and the size, density and distribution of the mesh are determined. The mesh is generated using the meshing tool of ANSYS, and the cell quality, distortion, Jacobian and other indicators of the mesh are checked to ensure that the mesh quality meets the calculation requirements, and the temperature in each mesh is calculated to obtain the theoretical temperature value of each mesh position. Specifically, tetrahedral grid division is adopted, with about 100,000 units. A denser grid is adopted in areas where the temperature gradient and flow rate change greatly (such as near the container wall and the liquid nitrogen inlet / outlet). The unit mass shall be not less than 0.7 and the distortion shall be not more than 0.3. The specific division and the number of grids shall be determined according to the actual size of the container. Preferably, if the unit size of the grid division is too small, which affects the calculation of the unit temperature, grid fusion is required to fuse several adjacent grids into one. The specific setting and adjustment shall be made according to the actual situation, and this application does not impose any specific restrictions.

[0030] In some embodiments, the previously established physical equation is input into ANSYS, a transient thermal analysis solver is selected, and solution parameters such as time step, convergence criterion, initial condition and boundary condition are set, wherein the initial condition is the initial temperature of liquid nitrogen and the initial temperature of the container wall, and the external condition is that the temperature outside the container is constant, and the liquid nitrogen inlet flow rate and temperature are known; the solver is run to calculate the temperature distribution in the liquid nitrogen biological container; for example, after the solver is calculated, a temperature distribution diagram of each time step in the liquid nitrogen biological container is obtained, and as time goes by, the temperature of the liquid nitrogen in the container gradually tends to be stable. Near the container wall, due to the effect of heat conduction, the temperature gradient is large. In the area where liquid nitrogen flows, the temperature distribution is affected by the flow rate and turbulence intensity. In steady state, the temperature in the center area of ​​the container is about -196 degrees Celsius, and the temperature near the container wall is about -180 degrees Celsius (the specific values ​​depend on the material properties, geometric dimensions and boundary conditions); the solution process is monitored in real time to ensure the stability and accuracy of the solution.

[0031] For example, a cylindrical liquid nitrogen biological container is filled with liquid nitrogen, and the container wall is made of stainless steel. Multiple low-temperature temperature sensors are arranged in the container wall and the liquid nitrogen to obtain temperature data of the container wall and the inside of the liquid nitrogen. According to the size and material of the container and the physical properties of the liquid nitrogen, a heat conduction model is established to simulate the temperature distribution of the liquid nitrogen in the container. The sensor data is collected and pre-processed before being input into the space and time calibration model. The calibrated temperature data is compared with the theoretical temperature value calculated by the physical model, and the sensor calibration parameters are adjusted until satisfactory calibration accuracy is achieved.

[0032] In some embodiments, the at least one frame of original array temperature is input into a basic calibration model to obtain an output initial calibration temperature, including: the basic calibration model includes a spatial calibration model and a temporal calibration model, the spatial calibration model performs spatial temperature calibration on the at least one frame of original array temperature to obtain a spatial calibration temperature, and the temporal calibration model performs temporal temperature calibration on the spatial calibration temperature to obtain the initial calibration temperature.

[0033] Specifically, the types of the spatial calibration model and the temporal calibration model are convolutional neural network models. Exemplarily, the convolutional neural network model includes network parameters such as an input layer, a convolution layer, a pooling layer, a fully connected layer, and a convolution kernel. The convolution kernel is a function used to define the weights of a portion of the matrix data in the input matrix data, and the convolution kernel can realize filtering functions such as maximum value, minimum value, mean value, and smoothing.

[0034] In one embodiment, optionally, at least one spatial convolution layer in the spatial calibration model performs continuous convolution filtering on at least one frame of original array temperature in sequence to obtain a spatial calibration temperature corresponding to at least one frame of original array temperature; at least one temporal convolution layer in the temporal calibration model performs continuous convolution filtering on the one-dimensional intermediate array temperature in sequence to obtain an initial calibration temperature. Taking the original array temperature as an example, which includes H1×W1 original temperature values, after the original array temperature passes through the first convolution layer, H2×W2 intermediate spatial calibration temperatures are obtained. After passing through M convolution layers, 1×1 spatial calibration temperatures output by the fully connected convolution layer are obtained. In this embodiment, the number of convolution layers of the spatial calibration model can be customized. Using a convolutional neural network model to construct a spatial calibration model and a temporal calibration model is a prior art in the field, and this application does not impose specific restrictions.

[0035] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: The present application constructs a physical calibration model through the basic parameters of a low-temperature liquid nitrogen biological container, obtains a target theoretical temperature set, compares the initial calibration temperature with the target theoretical temperature set, determines a difference group, constructs a difference optimization model, inputs the difference group to obtain an optimization suggestion, optimizes the basic calibration model and the physical calibration model according to the optimization suggestion, and achieves the effect of improving temperature calibration in the low-temperature liquid nitrogen biological container; it is established based on physical principles and has good generalization capabilities for different container sizes, materials and liquid nitrogen flow states. When the container conditions or the liquid nitrogen state change, the physical calibration model can quickly adjust and give a new target theoretical temperature set, thereby achieving the effect of quickly migrating data modules and improving the generalization of data models.

[0036] Embodiment 2: In embodiment 1, the accuracy of temperature calibration in a low-temperature liquid nitrogen biological container is improved by constructing a physical calibration model. This embodiment makes further improvements on the basis of the above content.

[0037] In some embodiments, constructing a difference optimization model includes: selecting a random forest regression model as a basic model for training, determining input features and output features of the difference optimization model, dividing the preprocessed historical data into a training set and a validation set, initializing model parameters using a random forest regressor in a machine learning library, and training the random forest regression model using the training set data; validating the random forest regression model using the validation set, and adjusting the model parameters according to the validation results, and using the trained random forest regression model as the difference optimization model.

[0038] In some embodiments, a random forest regression model is selected as a difference optimization model. Random forest is an integrated learning method that improves the accuracy and stability of prediction by constructing multiple decision trees and synthesizing their prediction results. Random forest can handle nonlinear relationships, is insensitive to outliers, and has good generalization ability, which is suitable for temperature difference prediction; determine the input features, use the triple data in the difference group as input data, determine that the output feature is the adjusted theoretical temperature value or the predicted value of the temperature difference, that is, the target variable that the model needs to predict, and the data used in the training model and the verification model are all historical data, that is, the data of the difference group generated in the previous experimental process, and preprocess the historical data, including processing missing values, such as filling with the mean, median or interpolation method, detecting and processing outliers, such as deleting or replacing outliers, and standardizing or normalizing the data to ensure that the dimensions between different features are consistent. The preprocessed data is randomly divided into a training set and a verification set, usually divided in a ratio of 70%:30% or 80%:20%, to ensure that the training set and the verification set are consistent in feature distribution to avoid data bias. Initialize using the random forest regressor in the machine learning library and set the model parameters: the number of trees, which determines the number of decision trees in the random forest. A larger value is usually selected to improve the stability of the model; the maximum depth, which limits the maximum depth of the decision tree to prevent overfitting; other parameters, such as the minimum number of sample splits and the minimum number of leaf node samples, are adjusted according to the data characteristics and model performance. Use the training set data to train the random forest regression model, and adjust the model parameters to optimize the model's prediction accuracy; use the validation set to validate the random forest regression model, and adjust the model parameters based on the validation results. Use the trained random forest regression model as the differential optimization model.

[0039] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: This application analyzes the generated difference groups by constructing a difference optimization model and obtains optimization suggestions, which can more accurately predict and adjust the theoretical temperature value, thereby improving the process and results of temperature calibration, and achieving improved results of temperature calibration in low-temperature liquid nitrogen biological containers.

[0040] Embodiment 3: This embodiment makes further improvements on the basis of the above contents.

[0041] In some embodiments, the difference group includes a first difference group and a second difference group, each of the first difference group and the second difference group includes a first difference area and a second difference area; the optimization suggestion includes a first optimization suggestion and a second optimization suggestion; the first difference group is determined based on the initial calibration temperature and the target theoretical temperature set, the first difference group is input into the difference optimization model to obtain a first optimization suggestion, the physical calibration model is adjusted and optimized according to the first optimization suggestion, and the target theoretical temperature control set is obtained according to the adjusted physical calibration model; the second difference group is determined based on the initial calibration temperature, the target theoretical temperature set and the target theoretical temperature control set, the second difference group is input into the difference optimization model to obtain a second optimization suggestion; the basic calibration model and the physical calibration model are optimized based on the first optimization suggestion and the second optimization suggestion.

[0042] In some embodiments, the first optimization suggestion is proposed based on the prediction results of the difference optimization model, aiming to adjust the parameters or settings in the physical calibration model to reduce the temperature difference in the first difference area. Specific suggestions include adjusting material properties (such as thermal conductivity), changing boundary conditions (such as external temperature, thermal radiation coefficient), optimizing geometric models (such as adjusting container structure, wall thickness), etc., which are comprehensively obtained based on the output results of the difference optimization model and the actual situation of the physical calibration model. The first difference area is those areas with significant temperature differences and similarities not less than the similarity threshold. They are the focus of most attention in temperature calibration because they have a greater impact on the overall temperature distribution. By prioritizing the optimization of the first difference area, the accuracy of temperature calibration can be more effectively improved. Concentrating limited optimization resources on the areas that need optimization most, improving optimization efficiency, and optimizing areas with significant differences can see improvements in calibration effects more quickly. Through regional division, the differences in temperature distribution can be more clearly understood, providing accurate directions for subsequent optimization.

[0043] In some embodiments, determining a first difference group based on the initial calibration temperature and the set of target theoretical temperatures includes: calculating the difference between each theoretical temperature value in the set of target theoretical temperatures and the initial calibration temperature, comparing the similarity between the differences with a preset first similarity threshold, dividing the positions corresponding to each theoretical temperature value with a similarity not less than the first similarity threshold into the same region and labeling it as the first difference region; dividing the positions corresponding to each theoretical temperature value with a similarity less than the first similarity threshold into the same region and labeling it as the second difference region; forming a triple with the specific positions, target theoretical temperature values, and initial calibration temperature values in the first difference region and the second difference region to form the first difference group.

[0044] In some embodiments, the purpose of region division is to identify the similar and different parts between the set of target theoretical temperatures and the initial calibration temperature. Through region division, it is possible to focus on those regions with significant temperature differences, which are the key points for subsequent optimization and adjustment. Calculate the difference or relative difference between the temperature value at each grid position in the set of target theoretical temperatures and the temperature value at the corresponding position of the initial calibration temperature. The difference group includes those grid positions with significant differences between the set of target theoretical temperatures and the initial calibration temperature and their corresponding temperature value pairs. Region division helps to quickly locate the regions with significant temperature differences, so as to conduct targeted optimization and adjustment. Through region division, limited resources can be concentrated on the places where optimization is most needed, improving the optimization efficiency. By dividing regions, it is possible to more clearly understand the differences in temperature distribution.

[0045] In some embodiments, adjusting and optimizing the physical calibration model according to the first optimization suggestion, and obtaining a set of target theoretical temperature references according to the adjusted physical calibration model includes: re-dividing the first difference region and adjusting the basic parameters according to the first optimization suggestion, and obtaining all the theoretical temperature values in the re-divided first difference region to form the set of target theoretical temperature references.

[0046] In some embodiments, carefully analyze the temperature data in the first difference region to understand the specific situation and distribution law of the temperature differences, combine the parameters and settings of the physical calibration model, analyze the possible reasons for the temperature differences, optimize the prediction results of the model according to the differences, propose optimization suggestions for the first difference region, adjust the relevant parameters or settings in the physical calibration model according to the first optimization suggestion, and re-run the physical calibration model to obtain the adjusted set of target theoretical temperature references.

[0047] In some embodiments, the set of target theoretical temperatures is the temperature distribution calculated based on the original physical calibration model, and the set of target theoretical temperature references is the temperature distribution recalculated after adjusting the physical calibration model.

[0048] In some embodiments, determining the second difference group based on the initial calibration temperature, the target theoretical temperature set, and the target theoretical temperature control set includes: calculating the difference between each theoretical temperature control value in the target theoretical temperature control set and the initial calibration temperature, and calculating the temperature change rate of the target theoretical temperature set and the target theoretical temperature control set corresponding to the area; The positions corresponding to each theoretical temperature comparison value whose similarity is not less than the second similarity threshold are divided into the same area and marked as the third similarity difference area; the positions corresponding to each theoretical temperature comparison value whose similarity is less than the second similarity threshold are divided into the same area and marked as the fourth similarity difference area; The positions corresponding to each theoretical temperature control value whose temperature change rate is not less than the change threshold are divided into the same area and marked as the third change difference area; the positions corresponding to each theoretical temperature control value whose temperature change rate is less than the change threshold are divided into the same area and marked as the fourth change difference area; The intersection area of ​​the third similar difference area and the third change difference area is marked as the third difference area, and the intersection area of ​​the fourth similar difference area and the fourth change difference area is marked as the fourth difference area; the specific positions, target theoretical temperature control values ​​and initial calibration temperature values ​​in the third difference area and the fourth difference area are grouped into a triplet to form a second difference group.

[0049] In some embodiments, the sum of the first difference area and the second difference area is the entire area of ​​the container, the sum of the third difference area and the fourth difference area is less than or equal to the first difference area, and the first similarity threshold, the second similarity threshold and the change threshold are all pre-set based on historical data and historical experimental data, and need to be adjusted according to actual conditions. This application does not impose any specific restrictions.

[0050] In some embodiments, inputting the first difference group into the difference optimization model means inputting the triplet data corresponding to the first difference area in the first difference group into the difference optimization model; inputting the second difference group into the difference optimization model means inputting the triplet data corresponding to the third difference area and the fourth difference area in the second difference group into the difference optimization model.

[0051] In some embodiments, the basic calibration model and the physical calibration model are optimized based on the first optimization suggestion and the second optimization suggestion. The basic calibration model and the physical calibration model need to be optimized according to actual conditions, and this application does not impose specific restrictions.

[0052] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: Through regional division and difference group construction, this application can more accurately identify the difference parts in temperature calibration, generate targeted optimization suggestions based on the difference optimization model, and improve the accuracy of calibration; through the generation and application of two optimization suggestions, the basic calibration model and the physical calibration model are gradually improved, so that the temperature calibration is closer to the target theoretical temperature, and the effect of temperature calibration in the low-temperature liquid nitrogen biological container is further improved.

[0053] Embodiment 4: Figure 2 As shown, the embodiment of the present application also provides a temperature calibration system for a low-temperature liquid nitrogen biological container, the system comprising: A data acquisition and processing module is used to obtain the original array temperature data collected by the temperature sensor and perform preprocessing; A model group calibration module, the model group calibration module includes a basic calibration submodule and a physical calibration submodule, the basic calibration submodule is used to calibrate the original array temperature in space and time, and output the initial calibration temperature; the physical calibration submodule is used to establish a three-dimensional geometric model according to the basic parameters of the low-temperature liquid nitrogen biological container, set material properties and boundary conditions, perform simulation calculations based on heat conduction and fluid mechanics theory, and obtain a target theoretical temperature set; The difference analysis and optimization suggestion module is used to compare the initial calibration temperature and the target theoretical temperature set, determine the first difference group, input the first difference group into the difference optimization model, and obtain the first optimization suggestion; adjust the physical calibration model according to the first optimization suggestion, generate the target theoretical temperature control set, compare the initial calibration temperature, the target theoretical temperature set and the target theoretical temperature control set, determine the second difference group, input the second difference group into the difference optimization model, and obtain the second optimization suggestion; A model optimization and adjustment module, used for optimizing and adjusting the basic calibration model and the physical calibration model according to the first optimization suggestion and the second optimization suggestion; The difference optimization model building module is used to build the difference optimization model, including: selecting and training the random forest regression model as the basis of the difference optimization model, determining the input features and output features, using historical data for training and verification, adjusting the model parameters according to the verification results, and obtaining the final difference optimization model.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A temperature calibration method for a cryogenic liquid nitrogen biological container, characterized in that: include: Acquire at least one frame of raw array temperature collected by a temperature sensor; Inputting the at least one frame of original array temperature into a pre-trained temperature calibration model group to obtain an output target calibration temperature; The temperature calibration model group includes a basic calibration model and a physical calibration model, the at least one frame of original array temperature is input into the basic calibration model to obtain an output initial calibration temperature; a physical calibration model is constructed, and a target theoretical temperature set is obtained using the physical calibration model; A difference group is determined based on the initial calibration temperature and the target theoretical temperature set, a difference optimization model is constructed, the difference group is input into the difference optimization model to obtain optimization suggestions, and the basic calibration model and the physical calibration model are optimized according to the optimization suggestions.

2. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 1, characterized in that: Constructing a physical calibration model and using the physical calibration model to obtain a target theoretical temperature set includes: Obtaining basic parameters of the low-temperature liquid nitrogen biological container, establishing a three-dimensional geometric model according to the basic parameters, setting material properties and boundary conditions, establishing physical equations based on heat conduction theory and fluid mechanics theory, using a meshing tool to mesh the three-dimensional geometric model, and obtaining a target theoretical temperature value at each grid position through simulation calculation to form a target theoretical temperature set; The basic parameters include: the structure of the container, the material of the container, the basic physical properties of liquid nitrogen, the external environmental condition parameters, the initial temperature of the liquid nitrogen and the flow state parameters of the liquid nitrogen in the container.

3. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 1, characterized in that: Inputting the at least one frame of original array temperature into a basic calibration model to obtain an output initial calibration temperature includes: The basic calibration model includes a spatial calibration model and a temporal calibration model. The spatial calibration model performs spatial temperature calibration on the at least one frame of original array temperature to obtain a spatial calibration temperature. The temporal calibration model performs temporal temperature calibration on the spatial calibration temperature to obtain the initial calibration temperature.

4. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 1, characterized in that: The difference group includes a first difference group and a second difference group, and the optimization suggestion includes a first optimization suggestion and a second optimization suggestion; Determine a first difference group based on the initial calibration temperature and the target theoretical temperature set, input the first difference group into a difference optimization model to obtain a first optimization suggestion, adjust and optimize the physical calibration model according to the first optimization suggestion, and obtain a target theoretical temperature control set according to the adjusted physical calibration model; Determine a second difference group based on the initial calibration temperature, the target theoretical temperature set and the target theoretical temperature control set, input the second difference group into a difference optimization model, and obtain a second optimization suggestion; The basic calibration model and the physical calibration model are optimized based on the first optimization suggestion and the second optimization suggestion.

5. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 4, characterized in that: Determining a first difference group based on the initial calibration temperature and the target theoretical temperature set includes: Calculating the difference between each theoretical temperature value in the target theoretical temperature set and the initial calibration temperature, comparing the differences with a preset first similarity threshold according to the similarity between the differences, dividing the positions corresponding to each theoretical temperature value whose similarity is not less than the first similarity threshold into the same area and marking them as first difference areas; Divide the position corresponding to each theoretical temperature value whose similarity is less than the first similarity threshold into the same area and mark it as a second difference area; The specific positions in the first difference area and the second difference area, the target theoretical temperature value and the initial calibration temperature value are grouped into a triplet to form a first difference group.

6. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 4, characterized in that: Determining a second difference group based on the initial calibration temperature, the target theoretical temperature set, and the target theoretical temperature control set includes: Calculate the difference between each theoretical temperature comparison value in the target theoretical temperature comparison set and the initial calibration temperature, and calculate the temperature change rate of the target theoretical temperature set and the area corresponding to the target theoretical temperature comparison set; The positions corresponding to each theoretical temperature comparison value whose similarity is not less than the second similarity threshold are divided into the same area and marked as the third similarity difference area; the positions corresponding to each theoretical temperature comparison value whose similarity is less than the second similarity threshold are divided into the same area and marked as the fourth similarity difference area; The positions corresponding to each theoretical temperature control value whose temperature change rate is not less than the change threshold are divided into the same area and marked as the third change difference area; the positions corresponding to each theoretical temperature control value whose temperature change rate is less than the change threshold are divided into the same area and marked as the fourth change difference area; The intersection area of ​​the third similar difference area and the third change difference area is marked as the third difference area, and the intersection area of ​​the fourth similar difference area and the fourth change difference area is marked as the fourth difference area; the specific positions, target theoretical temperature control values ​​and initial calibration temperature values ​​in the third difference area and the fourth difference area are grouped into a triplet to form a second difference group.

7. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 5, characterized in that: Inputting the first difference group into the difference optimization model means inputting the triplet data corresponding to the first difference area in the first difference group into the difference optimization model; inputting the second difference group into the difference optimization model means inputting the triplet data corresponding to the third difference area and the fourth difference area in the second difference group into the difference optimization model.

8. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 4, characterized in that: The physical calibration model is adjusted and optimized according to the first optimization suggestion, and a target theoretical temperature control set is obtained according to the adjusted physical calibration model, including: The first difference region is re-divided according to the first optimization suggestion and the basic parameters are adjusted to obtain all theoretical temperature values ​​in the re-divided first difference region to form a target theoretical temperature comparison set.

9. The temperature calibration method of the low-temperature liquid nitrogen biological container according to claim 1, characterized in that: Constructing a differential optimization model includes: selecting a random forest regression model as a basic model for training, determining the input features and output features of the differential optimization model, dividing the preprocessed historical data into a training set and a validation set, initializing model parameters using a random forest regressor in a machine learning library, and training the random forest regression model using the training set data; validating the random forest regression model using the validation set, and adjusting the model parameters according to the validation results, and using the trained random forest regression model as the differential optimization model.

10. A temperature calibration system for a low-temperature liquid nitrogen biological container, applied to a temperature calibration method for a low-temperature liquid nitrogen biological container according to any one of claims 1 to 9, characterized in that: The system comprises: A data acquisition and processing module is used to obtain the original array temperature data collected by the temperature sensor and perform preprocessing; A model group calibration module, the model group calibration module includes a basic calibration submodule and a physical calibration submodule, the basic calibration submodule is used to calibrate the original array temperature in space and time, and output the initial calibration temperature; the physical calibration submodule is used to establish a three-dimensional geometric model according to the basic parameters of the low-temperature liquid nitrogen biological container, set material properties and boundary conditions, perform simulation calculations based on heat conduction and fluid mechanics theory, and obtain a target theoretical temperature set; The difference analysis and optimization suggestion module is used to compare the initial calibration temperature and the target theoretical temperature set, determine the first difference group, input the first difference group into the difference optimization model, and obtain the first optimization suggestion; adjust the physical calibration model according to the first optimization suggestion, generate the target theoretical temperature control set, compare the initial calibration temperature, the target theoretical temperature set and the target theoretical temperature control set, determine the second difference group, input the second difference group into the difference optimization model, and obtain the second optimization suggestion; A model optimization and adjustment module, used for optimizing and adjusting the basic calibration model and the physical calibration model according to the first optimization suggestion and the second optimization suggestion; The difference optimization model building module is used to build the difference optimization model, including: selecting and training the random forest regression model as the basis of the difference optimization model, determining the input features and output features, using historical data for training and verification, adjusting the model parameters according to the verification results, and obtaining the final difference optimization model.

Images