Method and device for determining gas collision cross section based on electron swarm parameters, terminal equipment and storage medium

Abstract

The application discloses a gas collision cross section determination method and device based on electron group parameters, a terminal equipment and a storage medium. The method comprises the following steps: inputting an electron group parameter to be solved into a preset cross section prediction model to obtain a collision cross section; the training of the cross section prediction model comprises the following steps: obtaining a first electron group parameter according to a historical collision cross section; taking the first electron group parameter as the input of the cross section prediction model to be trained and taking the predicted collision cross section as the output; repeatedly performing an iterative training operation until the loss function converges to obtain the cross section prediction model; the iterative training operation comprises the following steps: obtaining a current predicted collision cross section according to the current first electron group parameter; calculating a loss function according to the current first electron group parameter and a current second electron group parameter of the current predicted collision cross section; generating disturbance data according to the current first electron group parameter and the current predicted collision cross section and updating the current first electron group parameter. Through the implementation of the application, the mapping relationship from the electron group parameter to the collision cross section is determined.

Description

Technical Field

[0001] The present invention relates to the technical field of gas collision cross sections, and in particular to a method, device, terminal equipment and storage medium for determining a gas collision cross section based on electron group parameters. Background Art

[0002] Electron transport models are crucial for the predictive control of low-temperature plasma systems. Collision cross sections are typically derived through experimental and theoretical techniques, and verified through electron group experiments to ensure their effectiveness. Generally, obtaining collision cross section sets through calculation and experiment is difficult, and they are inconsistent with the electron group parameters, requiring manual correction. Electron group parameters can be measured through electron group experiments or calculated by solving the Boltzmann equation for the collision cross section. However, existing methods do not have a direct mapping from electron group data to collision cross sections, and rely solely on manual intuition and experience for correction, resulting in inefficient prediction of collision cross sections. Summary of the Invention

[0003] The embodiments of the present invention provide a method, apparatus, terminal device, and storage medium for determining a gas collision cross section based on electron group parameters, which can effectively solve the problem that the existing technology has no direct mapping relationship from electron group data to collision cross section, and relies only on human intuition and experience for correction, resulting in low collision cross section prediction efficiency.

[0004] An embodiment of the present invention provides a method for determining a gas collision cross section based on electron group parameters, comprising:

[0005] Obtaining the electron group parameters of the gas to be determined;

[0006] Inputting the electron group parameters into a preset cross-section prediction model for prediction to obtain the collision cross-section of the gas to be determined;

[0007] The training of the cross-section prediction model includes:

[0008] Obtain historical collision cross sections for multiple gases;

[0009] Solving according to the historical collision cross section to obtain a first electron group parameter for representing the historical parameter;

[0010] Using the first electron group parameter as the input of the cross-section prediction model to be trained and the predicted collision cross section as the output, repeatedly performing an iterative training operation until a preset loss function converges, thereby obtaining a trained cross-section prediction model;

[0011] The iterative training operation includes:

[0012] The collision cross section is predicted according to the current first electron group parameter to obtain the current predicted collision cross section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter;

[0013] Solve according to the current predicted collision cross section to obtain the corresponding current second electron group parameters;

[0014] Calculate according to the current first electron group parameter and the current second electron group parameter to obtain a loss function;

[0015] Calculation is performed based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data; and the current first electron group parameters are updated based on the disturbance data.

[0016] Furthermore, a solution is performed based on the historical collision cross section to obtain a first electron group parameter for representing the historical parameter, including:

[0017] Determining a first threshold energy corresponding to each historical collision cross section according to the historical collision cross sections;

[0018] Randomly selecting every two historical collision cross sections as a collision cross section pair, and generating cross section training data according to the collision cross section pair and the corresponding first threshold energy;

[0019] A solution is performed according to the cross-section training data and a preset solver to obtain first electron group parameters for representing historical parameters.

[0020] Furthermore, a collision cross section prediction is performed based on the current first electron group parameters to obtain a current predicted collision cross section, including:

[0021] Normalizing the current first electron group parameter to obtain the target electron group parameter;

[0022] Performing vector conversion according to the target electron group parameter to obtain a current first electron group parameter vector;

[0023] The collision cross section is predicted based on the current first electron group parameter vector to obtain the current predicted collision cross section.

[0024] Furthermore, calculations are performed based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data, including:

[0025] Determining a second threshold energy corresponding to the current predicted collision cross section according to the current predicted collision cross section;

[0026] Calculating the similarity between every two current first electron group parameters according to each current first electron group parameter, and using the similarity as a disturbance judgment value;

[0027] Making a judgment based on the disturbance judgment value and a preset disturbance threshold;

[0028] The first electronic group parameter corresponding to the disturbance judgment value being greater than the preset disturbance threshold is used as the first target parameter; and first disturbance data is obtained by performing calculation based on the first target parameter and the corresponding second threshold energy;

[0029] The current first electronic group parameter whose disturbance judgment value is less than or equal to the preset disturbance threshold is used as the second target parameter; and second disturbance data is obtained by performing calculation based on the second target parameter and the corresponding second threshold energy;

[0030] Disturbance data is generated according to the first disturbance data and the second disturbance data.

[0031] Furthermore, the current first electron group parameters are updated according to the disturbance data, including:

[0032] Updating the current first electron group parameters according to the first disturbance data;

[0033] updating the current first electron group parameters according to the second perturbation data;

[0034] The first disturbance data and the second disturbance data are used alternately to update the current first electron group parameters.

[0035] Furthermore, the cross-section prediction model is a multi-branch neural network model.

[0036] Furthermore, before inputting the electron group parameters into a preset cross-section prediction model for prediction, the method further includes: normalizing the electron group parameters to obtain final electron group parameters.

[0037] As an improvement to the above solution, another embodiment of the present invention provides a device for determining a gas collision cross section based on electron group parameters, comprising:

[0038] A parameter acquisition module, used to obtain the electron group parameters of the gas to be determined;

[0039] A collision cross section determination module is used to input the electron group parameters into a preset cross section prediction model for prediction to obtain a collision cross section of the gas to be determined;

[0040] The model training module is used to train the cross-section prediction model, including:

[0041] A historical cross-section acquisition unit is used to obtain historical collision cross-sections of multiple gases;

[0042] an electron group parameter solving unit, configured to solve according to the historical collision cross section to obtain a first electron group parameter for representing the historical parameter;

[0043] an iterative operation unit, configured to use the first electron group parameter as input of the cross-section prediction model to be trained and the predicted collision cross section as output, repeatedly perform an iterative training operation until a preset loss function converges, thereby obtaining a trained cross-section prediction model;

[0044] The iterative training operation includes:

[0045] The collision cross section is predicted according to the current first electron group parameter to obtain the current predicted collision cross section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter;

[0046] Solve according to the current predicted collision cross section to obtain the corresponding current second electron group parameters;

[0047] Calculate according to the current first electron group parameter and the current second electron group parameter to obtain a loss function;

[0048] Calculation is performed based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data; and the current first electron group parameters are updated based on the disturbance data.

[0049] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a method for determining a gas collision cross section based on electron group parameters as described in the above embodiment.

[0050] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program. When the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the gas collision cross-section determination method based on electron group parameters described in the above embodiment.

[0051] By implementing the present invention, at least the following beneficial effects are achieved:

[0052] The present invention provides a method, apparatus, terminal device and storage medium for determining gas collision cross sections based on electron group parameters. The method can obtain the electron group parameters of the gas to be determined; input the electron group parameters into a preset cross section prediction model for prediction to obtain the collision cross section of the gas to be determined; wherein the training of the cross section prediction model includes: obtaining historical collision cross sections of multiple gases; solving according to the historical collision cross sections to obtain a first electron group parameter for representing the historical parameter; using the first electron group parameter as the input of the cross section prediction model to be trained and the predicted collision cross section as the output, repeatedly performing iterative training operations until the preset collision cross section is obtained. The loss function converges to obtain a trained cross-section prediction model; the iterative training operation includes: predicting the collision cross-section based on the current first electron group parameter to obtain the current predicted collision cross-section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter; solving according to the current predicted collision cross-section to obtain the corresponding current second electron group parameter; calculating according to the current first electron group parameter and the current second electron group parameter to obtain the loss function; calculating according to the current first electron group parameter and the current predicted collision cross-section to generate perturbation data; and updating the current first electron group parameter according to the perturbation data. The cross-section prediction model is optimized through iterative training operations, so that the model learns the relationship between the electron group parameter and the collision cross-section. At the same time, by adding perturbation data, the model gradually approaches the optimal collision cross-section. No manual cross-section correction or model correction is required. A mapping relationship from the electron group parameter to the collision cross-section is formed, which improves the prediction efficiency of the collision cross-section while also improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 1 is a flow chart of a method for determining a gas collision cross section based on electron group parameters provided by one embodiment of the present invention;

[0054] Figure 2 1 is a schematic diagram of a multi-branch neural network for a method for determining a gas collision cross section based on electron group parameters provided by one embodiment of the present invention;

[0055] Figure 3 1 is a schematic diagram of an iterative training operation flow provided by an embodiment of the present invention;

[0056] Figure 4 1 is a schematic diagram of a gas collision cross section provided by an embodiment of the present invention;

[0057] Figure 5 1 is a schematic structural diagram of a gas collision cross section determination device based on electron group parameters provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0059] See also Figure 1 , is a flow chart of a method for determining a gas collision cross section based on electron group parameters provided by one embodiment of the present invention, comprising:

[0060] S1. Obtaining the electron group parameters of the gas to be determined;

[0061] S2. Inputting the electron group parameters into a preset cross-section prediction model to perform prediction to obtain a collision cross section of the gas to be determined;

[0062] The training of the cross-section prediction model includes:

[0063] Obtain historical collision cross sections for multiple gases;

[0064] Solving according to the historical collision cross section to obtain a first electron group parameter for representing the historical parameter;

[0065] Using the first electron group parameter as the input of the cross-section prediction model to be trained and the predicted collision cross section as the output, repeatedly performing an iterative training operation until a preset loss function converges, thereby obtaining a trained cross-section prediction model;

[0066] The iterative training operation includes:

[0067] The collision cross section is predicted according to the current first electron group parameter to obtain the current predicted collision cross section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter;

[0068] Solve according to the current predicted collision cross section to obtain the corresponding current second electron group parameters;

[0069] Calculate according to the current first electron group parameter and the current second electron group parameter to obtain a loss function;

[0070] Calculation is performed based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data; and the current first electron group parameters are updated based on the disturbance data.

[0071] Specifically, the electron population parameters include the electron drift velocity W e , density-normalized longitudinal electron diffusion coefficient ND L and the effective ionization rate coefficient k effThe collision cross sections include: adsorption collision cross section, ionization collision cross section, vibration excitation collision cross section, electron excitation collision cross section and elastic collision cross section.

[0072] Specifically, electron group parameters can be measured through electron group experiments, such as using pulsed Thomson (PT) experiments to measure the electron group parameters of the gas to be determined, or they can be calculated by solving the Boltzmann equation for the collision cross section. However, there is no direct mapping relationship between electron group data and collision cross sections. Therefore, this embodiment focuses on determining the collision cross section set from electron group data, also known as the inverse group problem. Currently, there are two main numerical techniques for solving the inverse group problem: iterative group technology and the recent application of artificial neural networks. However, these two methods have the following significant disadvantages: a major limitation of the solution to the inverse group problem is its pathological nature. Under this limitation, iterative group technology relies on the intuition and experience of experts. Coupled with the trial-and-error nature of this method, the program is inefficient and difficult to replicate. Compared with equivalent networks designed using physical information, traditional artificial neural networks limit the network's ability to reliably and self-consistently simulate multiple independent cross-section regressions. The degenerate nature of the inverse group problem limits the prediction of multiple collision cross sections simultaneously.

[0073] In a preferred embodiment of the present invention, the electron group parameters of the gas to be determined are first obtained; then the electron group parameters are input into a preset cross-section prediction model for prediction to obtain the collision cross-section of the gas to be determined. The training of the cross-section prediction model includes: obtaining the historical collision cross-sections of multiple gases from a preset database, which is the LXCat database; solving according to the historical collision cross-sections to obtain the first electron group parameters used to represent the historical parameters; using the first electron group parameters as the input of the cross-section prediction model to be trained and the predicted collision cross-section as the output, repeatedly performing iterative training operations until the preset loss function converges to obtain a trained cross-section prediction model; the iterative training operation includes: predicting the collision cross-section according to the current first electron group parameters to obtain the current predicted collision cross-section; and the current first electron group parameters of the first iterative training operation are the initial first electron group parameters; solving according to the current predicted collision cross-section to obtain the corresponding current second electron group parameters, and the corresponding electron group parameters can be obtained from the collision cross-section according to the Boltzmann equation; calculating according to the current first electron group parameters and the current second electron group parameters to obtain the loss function, which is an error loss function. x i 、y i They are the first electron group parameter and the second electron group parameter respectively; according to the current first electron group parameter and the current predicted collision cross section, calculation is performed to generate disturbance data; and the current first electron group parameter is updated according to the disturbance data.

[0074] Preferably, solving the historical collision cross section to obtain the first electron group parameter for representing the historical parameter includes:

[0075] Determining a first threshold energy corresponding to each historical collision cross section according to the historical collision cross sections;

[0076] Randomly selecting every two historical collision cross sections as a collision cross section pair, and generating cross section training data according to the collision cross section pair and the corresponding first threshold energy;

[0077] A solution is performed according to the cross-section training data and a preset solver to obtain first electron group parameters for representing historical parameters.

[0078] In a preferred embodiment of the present invention, based on the historical collision cross sections, the first threshold energy corresponding to each historical collision cross section is determined, and every two historical collision cross sections σ1 and σ2 are randomly selected as collision cross section pairs. Based on the collision cross section pairs and the corresponding first threshold energies ε1 and ε2, cross section training data are generated. For the same type of cross sections, multiple cross section training data are generated between each two, such as:

[0079]

[0080] Among them, σs(ε) is the cross-section training data, σ1 and σ2 are collision cross-section pairs, the corresponding first threshold energies are ε1 and ε2, and the parameter r' is a random mixing ratio.

[0081] Then, the Boltzmann equation is solved based on the cross-section training data and the preset solver to obtain the first electron group parameter used to represent the historical parameter. The preset solver is bol sig+ solver. The cross-section training data is input into the solver to obtain the first electron group parameter corresponding to the training data. The first electron parameter represents the historical electron group parameter corresponding to the historical collision cross section. For example, the multi-point Boltzmann equation solver is used to calculate the historical electron group parameter corresponding to the historical collision cross section between 70 and 860Td (1Td = 1Townsend = 10-21 Vm 2 ) under the reduced electric field of We and ND with 150 logarithmic spacing L and keff.

[0082] Schematically, the collision cross section is predicted based on the current first electron group parameters to obtain the current predicted collision cross section, including:

[0083] Normalizing the current first electron group parameter to obtain the target electron group parameter;

[0084] Performing vector conversion according to the target electron group parameter to obtain a current first electron group parameter vector;

[0085] The collision cross section is predicted based on the current first electron group parameter vector to obtain the current predicted collision cross section.

[0086] In a preferred embodiment of the present invention, the electron drift velocity We, the density normalized longitudinal electron diffusion coefficient ND L The effective ionization rate coefficient keff must be normalized. All data except keff are normalized to [-1,1] after taking the logarithm. keff needs special processing before normalization. 3 / s,10- 25 m 3 / s] is equivalent to 10- 25 m 3 / s, in [-10- 25 m 3 / s,0 m 3 / s) is equivalent to -10- 25 m 3 / s, and then perform the following calculation to obtain the target electron group parameters:

[0087]

[0088] Then, a vector conversion is performed according to the target electron group parameter to obtain the current first electron group parameter vector. and They are the first electron group parameter vectors calculated under several reduced electric fields E / N, and the collision cross section is predicted for the current first electron group parameter vector to obtain the current predicted collision cross section.

[0089] Preferably, the calculation is performed based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data, including:

[0090] Determining a second threshold energy corresponding to the current predicted collision cross section according to the current predicted collision cross section;

[0091] Calculating the similarity between every two current first electron group parameters according to each current first electron group parameter, and using the similarity as a disturbance judgment value;

[0092] Making a judgment based on the disturbance judgment value and a preset disturbance threshold;

[0093] The first electronic group parameter corresponding to the disturbance judgment value being greater than the preset disturbance threshold is used as the first target parameter; and first disturbance data is obtained by performing calculation based on the first target parameter and the corresponding second threshold energy;

[0094] The current first electronic group parameter whose disturbance judgment value is less than or equal to the preset disturbance threshold is used as the second target parameter; and second disturbance data is obtained by performing calculation based on the second target parameter and the corresponding second threshold energy;

[0095] Disturbance data is generated according to the first disturbance data and the second disturbance data.

[0096] In a preferred embodiment of the present invention, the prediction of a large number of similar cross sections poses a substantial challenge to the solution of the inverse group problem. The prediction ability of the cross section prediction model can be improved by limiting the training data to perturbations around the reference cross section set. To this end, a perturbation data, i.e., a weighted mixture of the predicted cross section data, is proposed to train the model. The model includes three stages: initialization stage, exploration stage, and refinement stage. Figure 3 First, according to the current predicted collision cross section, the second threshold energy ε corresponding to the current predicted collision cross section is determined. s and ε c，i ; ε s and ε c，i Respectively represent the threshold energy of the current predicted collision cross section; according to each current first electron group parameter, calculate the similarity r between each two current first electron group parameters, and use the similarity as the disturbance judgment value. A value close to 1 will cause the current first electron group parameter σ c and σ c，i Small perturbations around , while values ​​close to 0 lead to large perturbations. Values ​​of r greater than 1 can be used to generate the current first electron group parameter σ c and σ c，i The surrounding enhanced disturbances are used to expand the solution space beyond the available data. The degree of these disturbances defines the ability of the network to explore the solution space or improve the existing solution. The judgment is made according to the disturbance judgment value and the preset disturbance threshold; the current first electron group parameter corresponding to the disturbance judgment value being greater than the preset disturbance threshold is used as the first target parameter; and the first disturbance data is obtained by calculation based on the first target parameter and the corresponding second threshold energy; the current first electron group parameter with a disturbance judgment value less than or equal to the preset disturbance threshold is used as the second target parameter; and the second disturbance data is obtained by calculation based on the second target parameter and the corresponding second threshold energy; disturbance data is generated based on the first disturbance data and the second disturbance data.

[0097] In another preferred embodiment of the present invention, if the training data, i.e., the first electron group parameter, is restricted to a small perturbation, the solution may fall into a local minimum. On the contrary, a large perturbation may cause the model to be unable to determine a sufficiently accurate set of cross sections. The exploration phase and refinement phase of this embodiment are intended to strike a balance between these two regimes. In the exploration phase, the σ cMake large perturbations to help the network traverse the solution space outside the current fitting range; in the refinement stage, c Small perturbations are made to further refine the solution. Two iterations are performed during the exploration phase, while five iterations are performed during the refinement phase to help ensure that a particular solution is sufficiently refined after each exploration phase. For high-energy (>10 eV) processes, such as electronic excitation and ionization, r is sampled from the range [0.5, 0.8] in each iteration of the exploration phase, while in the refinement phase, r = 0.8 in this example. For low-energy processes, such as vibrational and elastic processes, r is sampled from the range [0.5, 1.5] during the exploration phase and from the range [0.8, 1.2] during the refinement phase. To reduce the impact of non-uniqueness in determining cross-section sets with multiple similar collision processes, the iterative process incrementally explores the solution space by using perturbations.

[0098] Preferably, updating the current first electron group parameters according to the disturbance data includes:

[0099] Updating the current first electron group parameters according to the first disturbance data;

[0100] updating the current first electron group parameters according to the second perturbation data;

[0101] The first disturbance data and the second disturbance data are used alternately to update the current first electron group parameters.

[0102] Specifically, the cross-section prediction model is a multi-branch neural network model (multi-branch ANN network model) that bridges the gap between the self-consistency requirement and the independent feature map of each cross-section. That is, for each cross-section, there is an independent dense layer block, and each dense layer block extends from a dense layer block. Each parallel branch is then allowed to develop a feature set specific to a single cross-section, while still ensuring that each regression is performed in the context of the complete cross-section set. This embodiment uses a multi-branch ANN of the following form:

[0103]

[0104] in, is a mapping defined by a dense weight matrix Wi and a bias vector Bi, and mish(x) = xtanh(ln(1+e x )) is a nonlinear activation function, and An array of n parallel branches is formed, each of which independently represents the nth section using the output of A2. and Each contains 64 elements, Each output n contains 1 element, while the first two layers contain 256 elements. The size of the weight matrix is ​​corresponding. The architecture diagram of the multi-branch ANN is as follows Figure 2 As shown, for this embodiment, the output end includes 5 neurons (N=5), representing the adsorption collision cross section, ionization collision cross section, vibration excitation collision cross section, electron excitation collision cross section and elastic collision cross section respectively, and multiple collision cross sections can be predicted simultaneously.

[0105] Preferably, before inputting the electron group parameters into a preset cross-section prediction model for prediction, the method further includes: normalizing the electron group parameters to obtain final electron group parameters.

[0106] In a preferred embodiment of the present invention, the gas C5F 10 The proposed iterative process is demonstrated by the prediction of the O-section set. 64 neurons are selected for the hidden layer of each parallel branch. Figure 4 , the collision cross section set obtained for the gas in this embodiment through the above process is shown.

[0107] By implementing this embodiment, the electron group parameters of the gas to be determined are obtained; the electron group parameters are input into a preset cross-section prediction model for prediction to obtain the collision cross-section of the gas to be determined; wherein, the training of the cross-section prediction model includes: obtaining historical collision cross-sections of multiple gases; solving according to the historical collision cross-sections to obtain first electron group parameters for representing historical parameters; using the first electron group parameters as input of the cross-section prediction model to be trained and the predicted collision cross-section as output, repeatedly performing iterative training operations until the preset loss function converges to obtain a trained cross-section prediction model; the iterative training operation includes: predicting the collision cross-section according to the current first electron group parameters to obtain the current predicted collision cross-section; and, the current first electron group parameters of the first iterative training operation are the initial first electron group parameters; solving according to the current predicted collision cross-section to obtain the corresponding current second electron group parameters; calculating according to the current first electron group parameters and the current second electron group parameters to obtain the loss function; calculating according to the current first electron group parameters and the current predicted collision cross-section to generate perturbation data; and updating the current first electron group parameters according to the perturbation data. The cross-section prediction model is optimized through iterative training operations, so that the model learns the relationship between electron group parameters and collision cross-sections. At the same time, by adding perturbation data, the model gradually approaches the optimal collision cross-section. There is no need for manual cross-section correction and model correction. While forming a mapping relationship from electron group parameters to collision cross-sections, the prediction efficiency of collision cross-sections is improved while also improving the accuracy of predictions.

[0108] See also Figure 5, is a schematic structural diagram of a gas collision cross section determination device based on electron group parameters provided by one embodiment of the present invention, comprising:

[0109] A parameter acquisition module, used to obtain the electron group parameters of the gas to be determined;

[0110] A collision cross section determination module is used to input the electron group parameters into a preset cross section prediction model for prediction to obtain a collision cross section of the gas to be determined;

[0111] The model training module is used to train the cross-section prediction model, including:

[0112] A historical cross-section acquisition unit is used to obtain historical collision cross-sections of multiple gases;

[0113] an electron group parameter solving unit, configured to solve according to the historical collision cross section to obtain a first electron group parameter for representing the historical parameter;

[0114] an iterative operation unit, configured to use the first electron group parameter as input of the cross-section prediction model to be trained and the predicted collision cross section as output, repeatedly perform an iterative training operation until a preset loss function converges, thereby obtaining a trained cross-section prediction model;

[0115] The iterative training operation includes:

[0116] The collision cross section is predicted according to the current first electron group parameter to obtain the current predicted collision cross section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter;

[0117] Solve according to the current predicted collision cross section to obtain the corresponding current second electron group parameters;

[0118] Calculate according to the current first electron group parameter and the current second electron group parameter to obtain a loss function;

[0119] Calculation is performed based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data; and the current first electron group parameters are updated based on the disturbance data.

[0120] The present invention provides a gas collision cross section determination device based on electron group parameters, which obtains the electron group parameters of the gas to be determined according to the parameter acquisition module; inputs the electron group parameters into a preset cross section prediction model through the collision cross section determination module for prediction to obtain the collision cross section of the gas to be determined; wherein, in the model training module, the cross section prediction model is trained, including: according to the historical cross section acquisition unit, obtaining the historical collision cross sections of multiple gases; in the electron group parameter solving unit, solving according to the historical collision cross sections to obtain a first electron group parameter for representing the historical parameter; through the iterative operation unit, using the first electron group parameter as the input of the cross section prediction model to be trained to predict the collision cross section of the gas to be determined. The cross section is output, and the iterative training operation is repeatedly performed until the preset loss function converges to obtain a trained cross section prediction model; the iterative training operation includes: predicting the collision cross section based on the current first electron group parameter to obtain the current predicted collision cross section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter; solving according to the current predicted collision cross section to obtain the corresponding current second electron group parameter; calculating according to the current first electron group parameter and the current second electron group parameter to obtain the loss function; calculating according to the current first electron group parameter and the current predicted collision cross section to generate perturbation data; and updating the current first electron group parameter according to the perturbation data. The cross section prediction model is optimized through iterative training operations, so that the model learns the relationship between the electron group parameter and the collision cross section. At the same time, by adding perturbation data, the model gradually approaches the optimal collision cross section. There is no need for manual cross section correction and model correction. While forming a mapping relationship from the electron group parameter to the collision cross section, the prediction efficiency of the collision cross section is improved while also improving the accuracy of the prediction.

[0121] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0122] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0123] Another embodiment of the present invention provides a terminal device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the method for determining a gas collision cross section based on electron group parameters as described in the above embodiment. The terminal device can be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device can include, but is not limited to, a processor and a memory.

[0124] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and connects various parts of the entire terminal device using various interfaces and lines.

[0125] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med i aCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0126] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program. When the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the gas collision cross-section determination method based on electron group parameters described in the above embodiment.

[0127] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiments can be implemented. The computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium.

[0128] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for determining gas collision cross sections based on electron group parameters, characterized in that: include: Obtaining the electron group parameters of the gas to be determined; Inputting the electron group parameters into a preset cross-section prediction model for prediction to obtain the collision cross-section of the gas to be determined; The training of the cross-section prediction model includes: Obtain historical collision cross sections for multiple gases; Solving according to the historical collision cross section to obtain a first electron group parameter for representing the historical parameter; Using the first electron group parameter as the input of the cross-section prediction model to be trained and the predicted collision cross section as the output, repeatedly performing an iterative training operation until a preset loss function converges, thereby obtaining a trained cross-section prediction model; The iterative training operation includes: The collision cross section is predicted according to the current first electron group parameter to obtain the current predicted collision cross section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter; Solve according to the current predicted collision cross section to obtain the corresponding current second electron group parameters; Calculate according to the current first electron group parameter and the current second electron group parameter to obtain a loss function; Calculating based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data; and updating the current first electron group parameters based on the disturbance data; Solving the historical collision cross section to obtain the first electron group parameters used to represent the historical parameters includes: Determining a first threshold energy corresponding to each historical collision cross section according to the historical collision cross sections; Randomly selecting every two historical collision cross sections as a collision cross section pair, and generating cross section training data according to the collision cross section pair and the corresponding first threshold energy; Solving according to the cross-section training data and a preset solver to obtain a first electronic group parameter for representing a historical parameter; Calculate based on the current first electron group parameters and the current predicted collision cross section to generate perturbation data, including: Determining a second threshold energy corresponding to the current predicted collision cross section according to the current predicted collision cross section; Calculating the similarity between every two current first electron group parameters according to each current first electron group parameter, and using the similarity as a disturbance judgment value; Making a judgment based on the disturbance judgment value and a preset disturbance threshold; The first electronic group parameter corresponding to the disturbance judgment value being greater than the preset disturbance threshold is used as the first target parameter; and first disturbance data is obtained by performing calculation based on the first target parameter and the corresponding second threshold energy; The current first electronic group parameter whose disturbance judgment value is less than or equal to the preset disturbance threshold is used as the second target parameter; and second disturbance data is obtained by performing calculation based on the second target parameter and the corresponding second threshold energy; generating disturbance data according to the first disturbance data and the second disturbance data; The cross-section prediction model is a multi-branch neural network model; the multi-branch neural network model is: in, is composed of a dense weight matrix W i and the bias vector B i The defined mapping, mish(x) = xtanh(ln(1+e x )) is a nonlinear activation function, and Form an array of n parallel branches, each of which independently represents the nth section using the output of A2; and Each contains 64 elements, Each output n contains 1 element, and the first two layers contain 256 elements; the size of the weight matrix is ​​corresponding; the output end contains 5 neurons, representing the adsorption collision cross section, ionization collision cross section, vibration excitation collision cross section, electron excitation collision cross section and elastic collision cross section, respectively, and can predict multiple collision cross sections at the same time.

2. The method for determining a gas collision cross section based on electron group parameters according to claim 1, wherein: The collision cross section is predicted based on the current first electron group parameters to obtain the current predicted collision cross section, including: Normalizing the current first electron group parameter to obtain the target electron group parameter; Performing vector conversion according to the target electron group parameter to obtain a current first electron group parameter vector; The collision cross section is predicted based on the current first electron group parameter vector to obtain the current predicted collision cross section.

3. The method for determining gas collision cross section based on electron group parameters according to claim 1, characterized in that: Update the current first electron group parameters according to the perturbation data, including: Updating the current first electron group parameters according to the first disturbance data; updating the current first electron group parameters according to the second perturbation data; The first disturbance data and the second disturbance data are used alternately to update the current first electron group parameters.

4. The method for determining gas collision cross section based on electron group parameters according to claim 1, wherein: Before inputting the electron group parameters into a preset cross-section prediction model for prediction, the method further includes: normalizing the electron group parameters to obtain final electron group parameters.

5. A gas collision cross section determination device based on electron group parameters, characterized in that: include: A parameter acquisition module, used to obtain the electron group parameters of the gas to be determined; A collision cross section determination module is used to input the electron group parameters into a preset cross section prediction model for prediction to obtain a collision cross section of the gas to be determined; The model training module is used to train the cross-section prediction model, including: A historical cross-section acquisition unit is used to obtain historical collision cross-sections of multiple gases; an electron group parameter solving unit, configured to solve according to the historical collision cross section to obtain a first electron group parameter for representing the historical parameter; an iterative operation unit, configured to use the first electron group parameter as input of the cross-section prediction model to be trained and the predicted collision cross section as output, repeatedly perform an iterative training operation until a preset loss function converges, thereby obtaining a trained cross-section prediction model; The iterative training operation includes: The collision cross section is predicted according to the current first electron group parameter to obtain the current predicted collision cross section; and the current first electron group parameter of the first iterative training operation is the initial first electron group parameter; Solve according to the current predicted collision cross section to obtain the corresponding current second electron group parameters; Calculate according to the current first electron group parameter and the current second electron group parameter to obtain a loss function; Calculating based on the current first electron group parameters and the current predicted collision cross section to generate disturbance data; and updating the current first electron group parameters based on the disturbance data; The electron group parameter solving unit is used to solve according to the historical collision cross section to obtain a first electron group parameter used to represent the historical parameter, including: Determining a first threshold energy corresponding to each historical collision cross section according to the historical collision cross sections; Randomly selecting every two historical collision cross sections as a collision cross section pair, and generating cross section training data according to the collision cross section pair and the corresponding first threshold energy; Solving according to the cross-section training data and a preset solver to obtain a first electronic group parameter for representing a historical parameter; Calculating based on the current first electron group parameter and the current predicted collision cross section to generate disturbance data includes: determining a second threshold energy corresponding to the current predicted collision cross section based on the current predicted collision cross section; Calculating the similarity between every two current first electron group parameters according to each current first electron group parameter, and using the similarity as a disturbance judgment value; Making a judgment based on the disturbance judgment value and a preset disturbance threshold; The first electronic group parameter corresponding to the disturbance judgment value being greater than the preset disturbance threshold is used as the first target parameter; and first disturbance data is obtained by performing calculation based on the first target parameter and the corresponding second threshold energy; The current first electronic group parameter whose disturbance judgment value is less than or equal to the preset disturbance threshold is used as the second target parameter; and second disturbance data is obtained by performing calculation based on the second target parameter and the corresponding second threshold energy; generating disturbance data according to the first disturbance data and the second disturbance data; The cross-section prediction model is a multi-branch neural network model; the multi-branch neural network model is: in, is composed of a dense weight matrix W i and the bias vector B i The defined mapping, mish(x) = xtanh(ln(1+e x )) is a nonlinear activation function, and Form an array of n parallel branches, each of which independently represents the nth section using the output of A2; and Each contains 64 elements, Each output n contains 1 element, and the first two layers contain 256 elements; the size of the weight matrix is ​​corresponding; the output end contains 5 neurons, representing the adsorption collision cross section, ionization collision cross section, vibration excitation collision cross section, electron excitation collision cross section and elastic collision cross section, respectively, and can predict multiple collision cross sections at the same time.

6. A terminal device, characterized in that: The invention comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a method for determining a gas collision cross section based on electron group parameters according to any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for determining a gas collision cross section based on electron group parameters according to any one of claims 1 to 4.

Images