A method and apparatus for purifying helium

Abstract

The present application relates to helium purification technical field, disclose a kind of helium purification method, comprising the following steps: S1: the pressure of mixed gas is obtained;S2: when the pressure of mixed gas exceeds preset value, then the mixed gas is transported to purification system and is purified;S3: purification system is according to mixed gas and the concentration of pure helium gas needed to obtain, and the purification parameter of purification system is predicted by regression model, and purification system is according to the purification parameter of prediction to the mixed gas and is purified;S4: to the mixed gas after purification is analyzed, if gas purity is not less than preset value, then the mixed gas after purification is output;If gas purity is less than preset value, then return step S3 and continue to purify;Regression model is updated simultaneously.The regression model plays the role of different purification parameters suitable for different output mixed gas, improves the efficiency of purification, and the purification parameter is predicted by regression model, and the degree of automation is high.The present application also provides a kind of helium purification equipment.

Description

Technical Field

[0001] This invention relates to the field of helium purification technology, and in particular to a helium purification method and equipment. Background Technology

[0002] With the rapid development of technology, the demand for high-purity gases in the electronics and chemical industries is increasing. Taking the semiconductor industry as an example, it has enabled significant progress in highly integrated, high-quality products. High integration requires materials with high purity, including high-purity gases. Gas purity is a crucial factor directly affecting device quality. High-purity gases are the most important basic materials in industries such as integrated circuits, photovoltaics, optical fibers, and light-emitting diodes. As production demands increasingly higher purity gases (electronic gases), the sensitivity for impurity analysis in electronic gases needs to be at the ppb level. Harmful impurities in electronic gases are mainly compounds of oxygen and carbon, such as water, oxygen, carbon monoxide, carbon dioxide, and methane. Therefore, it is necessary to maximize the purity of helium through technological means. Currently, helium purification systems have low automation levels, wasting manpower and resources. Furthermore, the semiconductor chip industry now requires helium impurity standards of one part per billion (ppm), while traditional industries only require one part per million (ppm). Therefore, the purity of helium should be improved. Currently, helium purification removes impurity gas atoms and molecules using getters. However, the purification parameters are manually set, and the input gas mixture is different each time. The purification process is only tested; if it passes, it is output; otherwise, it is purified again, resulting in multiple purification cycles and low efficiency. Furthermore, the manual setting of purification parameters leads to low automation.

[0003] Existing technology provides a purification device for ppb-level ultra-high purity argon / helium gas, including an inlet pipeline, a first heat exchanger, a getter reactor, a second heat exchanger, a dehydrogenation tank, and an outlet pipeline. The inlet pipeline is connected to the refrigerant inlet of the first heat exchanger, and the refrigerant outlet of the first heat exchanger is connected to the getter reactor. The getter reactor contains a getter agent and is equipped with a heating element for heating the getter agent. The outlet of the getter reactor is connected to the heat medium inlet of the first heat exchanger, and the heat medium outlet of the first heat exchanger is connected to the heat medium inlet of the second heat exchanger. The heat medium outlet of the second heat exchanger is connected to the dehydrogenation tank, and the dehydrogenation tank is connected to the outlet pipeline. This patent removes impurities from the raw helium gas using the getter agent in the getter reactor. However, the heating temperature of the getter reactor and the purification parameters such as the getter agent are manually set. Since the mixed gas input is different each time, the purification process is only tested; if it passes the test, it is output; otherwise, it is purified again. This results in multiple purification cycles and low efficiency. Furthermore, the purification parameters are manually set, leading to low automation. Summary of the Invention

[0004] The purpose of this invention is to provide a highly efficient and automated helium purification method and equipment.

[0005] To achieve the above objectives, the present invention provides a helium purification method, comprising the following steps:

[0006] S1: Obtain the pressure of the mixed gas;

[0007] S2: When the pressure of the mixed gas exceeds the preset value, the mixed gas is transported to the purification system for purification;

[0008] S3: The purification system uses a regression model to predict the purification parameters of the purification system based on the concentration, temperature, composition of the mixed gas and the required concentration of pure helium. The purification system then purifies the mixed gas according to the predicted purification parameters.

[0009] S4: Perform gas purity and component analysis on the purified mixed gas. If the gas purity is not less than the preset value, output the purified mixed gas; if the gas purity is less than the preset value, return to step S3 to continue purification; at the same time, update the regression model using the concentration, temperature, purification parameters, gas purity and components of the mixed gas.

[0010] As a preferred embodiment, the regression model is a Gaussian process regression model, and the process of obtaining the regression model is as follows:

[0011] The collected data are compiled into a dataset, which is then divided into a training set and a test set. The collected data includes gas concentration, component ratio, temperature, and purification parameters during purification. Each data point is collected at a sampling point of the Gaussian process.

[0012] Functions for constructing Gaussian processes Where x represents the gas concentration, component ratio, and temperature parameters during the purification process, f(x) represents the purification parameters, m(x) represents the mean function, and k(x, x′) represents the covariance function, with x and x′ representing different sampling points; the Gaussian process is determined by the mean function m(x) and the covariance function k(x, x′). Assuming the mean function m(x) is 0, the covariance function k(x, x′) and the mean of the point conditional probability are obtained through Bayesian inference and continuously corrected;

[0013] Construct an expression to determine the prediction point based on the posterior probability: Where f(p,s,z..) represents the observed values ​​in the training set, f * The predicted values ​​are the purification parameters. The input vector (p, s, z...) represents the gas concentration, component ratio, temperature, etc., parameter values ​​during the purification process, respectively. k(p, s, z... p * ,s * ,z *...) represents the covariance function in the training set and the prediction point (p). * ,s * ,z * The covariance function of ...);

[0014] The model is trained using training and testing sets to obtain a trained regression model. For the trained regression model, a new vector (p, s, z...) is input, and the algorithm derives a new predicted value f* based on the prediction model.

[0015] As a preferred approach, k-fold cross-validation is used to evaluate the regression model during training, and the accuracy of the regression model is calculated based on the mean absolute error formula. When the accuracy of the regression model reaches the preset requirement, the calculation is stopped and the final regression model is output directly. Otherwise, the corresponding value is returned and the ratio of the training set to the test set is adjusted until the accuracy meets the requirements.

[0016] As a preferred approach, when correcting the covariance function k(x, x′) and the mean of the point conditional probability, Bayes' theorem is used to perform incremental calculations for each point in the sample. For the initial point, the prior and likelihood are assumed, and its posterior is calculated. Then, the posterior estimate of the initial point is used as the prior for the next estimate (two sample points). This process is repeated until the entire dataset is calculated. As the number of samples increases, the estimate becomes more and more accurate, thus achieving correction.

[0017] As a preferred embodiment, the purification system employs a metal getter reaction, and the purification parameters include the specific dosage of the getter, the proportion of each active element in the getter, and the proportion of the active element in the total getter.

[0018] The present invention also provides a helium purification device, comprising:

[0019] A switching system is used for the input of mixed gas and controls the input of mixed gas into the purification system according to the pressure of the mixed gas;

[0020] A purification system used to purify mixed gases;

[0021] The detection and analysis system is used to detect pipelines in the purification system and analyze the quality of the purified gas.

[0022] The purification control system records data including the concentration, temperature, and composition of the mixed gas input to the switching system, the purification parameters of the purification system, and the purity and composition of the purified gas. The recorded data is used to train a regression model. The trained regression model outputs predicted purification parameters of the purification system based on the concentration, temperature, and composition of the mixed gas input to the switching system, and the purification parameters of the purification system are adjusted according to the predicted purification parameters.

[0023] As a preferred embodiment, the switching system includes a main gas supply device, a secondary gas supply device, and an automatic switching valve group. The main gas supply device and the secondary gas supply device are respectively connected to one end of the automatic switching valve group, and the other end of the automatic switching valve group is connected to the purification system. Both the main gas supply device and the secondary gas supply device are equipped with pressure detection devices.

[0024] As a preferred embodiment, the automatic switching valve group includes a first switching branch, a second switching branch, a connecting branch, a first pressure regulating valve, and a second pressure regulating valve. The main gas supply device is connected to the connecting branch through the first switching branch, and the auxiliary gas supply device is connected to the connecting branch through the second switching branch. The connecting branch is connected to the purification system. The first pressure regulating valve is located on the first switching branch, and the second pressure regulating valve is located on the second switching branch. A first flow switch is provided between the main gas supply device and the first switching branch, and a second flow switch is provided between the auxiliary gas supply device and the second switching branch.

[0025] As a preferred embodiment, the purification system includes a purification pipeline and a metal gas suction device. The purification pipeline is connected to the switching system, the metal gas suction device is located inside the purification pipeline, and the detection and analysis system is connected to the purification pipeline.

[0026] As a preferred embodiment, the system also includes an internal blowing gas source, which is connected to the purification pipeline via a high-purity pipeline.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] This invention establishes a regression model that outputs predicted purification parameters based on the concentration, temperature, and composition of the mixed gas and the desired concentration of pure helium. The purification system adjusts these parameters according to the model's predictions, allowing for the application of different purification parameters to different mixed gases, thus improving purification efficiency. Furthermore, since the purification parameters are predicted by the regression model, the system boasts a high degree of automation. This invention also provides a helium purification device with a purification control system. This system records data from the purification system to train the regression model, which then outputs predicted purification parameters to control the purification system, resulting in high efficiency and automation. Attached Figure Description

[0029] Figure 1 This is a flowchart of a helium purification method according to an embodiment of the present invention.

[0030] Figure 2 This is a flow principle block diagram of the helium purification device according to an embodiment of the present invention.

[0031] Figure 3This is a structural diagram of the helium purification device according to an embodiment of the present invention.

[0032] In the diagram, 1-Switching system; 101-Main gas supply device; 102-Auxiliary gas supply device; 103-Automatic switching valve group; 1031-First switching branch; 1032-Second switching branch; 1033-Connecting branch; 1034-First pressure regulating valve; 1035-Second pressure regulating valve; 1036-First flow switch; 1037-Second flow switch; 1038-Third switching branch; 1039-Third pressure regulating valve; 1040-Third flow switch; 104-Pressure detection device; 2-Purification system; 201-Purification pipeline; 202-Metal suction device; 3-Detection and analysis system; 4-Purification control system; 5-Internal blowing gas source; 6-High purity pipeline. Detailed Implementation

[0033] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0034] like Figure 1 As shown, a preferred embodiment of the helium purification method of the present invention includes the following steps:

[0035] S1: Obtain the pressure of the mixed gas;

[0036] S2: When the pressure of the mixed gas exceeds the preset value, the mixed gas is transported to the purification system for purification;

[0037] S3: The purification system uses a regression model to predict the purification parameters of the purification system based on the concentration, temperature, composition of the mixed gas and the required concentration of pure helium. The purification system then purifies the mixed gas according to the predicted purification parameters.

[0038] S4: Perform gas purity and component analysis on the purified mixed gas. If the gas purity is not less than the preset value, output the purified mixed gas; if the gas purity is less than the preset value, return to step S3 to continue purification; at the same time, update the regression model using the concentration, temperature, purification parameters, gas purity and components of the mixed gas.

[0039] This embodiment sets up a regression model, which outputs predicted purification parameters based on the concentration, temperature, composition of the mixed gas and the required concentration of pure helium. The purification system adjusts the purification parameters according to the predicted values ​​of the regression model, thereby applying different purification parameters to different mixed gases, improving purification efficiency. Furthermore, since the purification parameters are predicted by the regression model, the system has a high degree of automation.

[0040] Furthermore, the regression model in this implementation example is a Gaussian process regression model, and the process of obtaining the regression model is as follows:

[0041] (1) The collected data constitutes a dataset, and the dataset is divided into a training set and a test set. The collected data includes the gas concentration, component ratio, temperature and purification parameters during purification. Each data point is collected at the sampling point of the Gaussian process.

[0042] (2) Construct the Gaussian process function f(x)~GP[m(x),k(x,x')], where x is the gas concentration, component ratio, and temperature parameters in the purification process, f(x) is the purification parameter, m(x) is the mean function, k(x,x′) is the covariance function, and x and x′ represent different sampling points; the Gaussian process is determined by the mean function m(x) and the covariance function k(x,x′). Assume that the mean function m(x) is 0. The covariance function k(x,x′) and the mean of the point conditional probability are obtained by Bayesian inference and continuously corrected.

[0043] The covariance function k(x, x′) generates a correlation coefficient matrix during Gaussian regression to measure the distance between any two points, ultimately determining the continuity (smoothness) of the sampling. The point conditional probability mean is used to derive the point expression, thereby solving for the hyperparameters using high likelihood estimation for prediction. Both the covariance function k(x, x′) and the point conditional probability mean are used in the following (3) when constructing the posterior probability expression to determine the predicted points for prediction.

[0044] In addition, in this embodiment, when correcting the covariance function k(x, x′) and the mean of the point conditional probability, according to Bayes' theorem, each point of the sample is incrementally calculated. For the initial point, the prior and likelihood are assumed, and its posterior is calculated. Then, the posterior estimate of the initial point is used as the prior for the next estimate (two sample points). This process is repeated until the entire dataset is calculated. As the number of samples increases, the estimate becomes more and more accurate, thus achieving correction.

[0045] (3) Construct an expression for determining the prediction point based on the posterior probability: Where f(p,s,z..) represents the observed values ​​in the training set, f * The predicted values ​​are the purification parameters. The input vector (p, s, z...) represents the gas concentration, component ratio, temperature, etc., parameter values ​​during the purification process, respectively. k(p, s, z... p * ,s * ,z * ...) represents the covariance function in the training set and the prediction point (p). * ,s * ,z * The covariance function of ...).

[0046] (4) The model is trained using training and test sets to obtain the trained regression model. For the trained regression model, a new vector (p, s, z...) is input, and the algorithm derives a new predicted value f* based on the prediction model. Further, during the training of the regression model, k-fold cross-validation is used to evaluate the regression model, and the accuracy of the regression model is calculated using the mean absolute error formula. When the accuracy of the regression model reaches the preset requirement, the computation stops and the final regression model is output directly; otherwise, the corresponding value is returned, and the ratio of the training set to the test set is adjusted until the accuracy meets the requirement. In this embodiment, the accuracy requirement is mainly characterized by parameters such as the goodness of fit. Generally, a goodness of fit between 0.9 and 1 indicates a high model accuracy.

[0047] Furthermore, specifically, the purification system of this embodiment adopts a metal getter reaction, and the purification parameters include the specific dosage of the getter, the proportion of each active element in the getter, and the proportion of the active element in the total getter.

[0048] Example 2

[0049] like Figure 2 As shown, a preferred embodiment of the helium purification device of the present invention includes:

[0050] Switching system 1 is used for the input of mixed gas and controls the input of mixed gas into the purification system according to the pressure of the mixed gas;

[0051] Purification system 2 is used to purify the mixed gas;

[0052] The detection and analysis system 3 is used to detect the pipelines in the purification system and analyze the quality of the purified gas.

[0053] The purification control system 4 is used to record data, including the concentration, temperature and composition of the mixed gas input to the switching system, the purification parameters of the purification system, and the purity and composition of the purified gas. The recorded data is used to train a regression model. The trained regression model outputs predicted purification parameters of the purification system based on the concentration, temperature and composition of the mixed gas input to the switching system, and the purification parameters of the purification system are adjusted according to the predicted purification parameters.

[0054] The helium purification equipment in this embodiment is equipped with a purification control system 4, which can record data from the purification system 2 to train a regression model. The trained regression model outputs predicted purification parameters to control the purification system 2, resulting in high efficiency and a high degree of automation.

[0055] Specifically, the mixed gas is input into switching system 1, and the pressure and temperature sensors in switching system 1 determine whether the values ​​exceed the preset pressure and temperature values. If the preset values ​​are exceeded, another pressure loop in switching system 1 is activated, and the gas is transported to purification system 2. Purification system 2 uses chemical methods to improve the purity of helium. The resulting helium of a certain purity is then analyzed in real time by detection and analysis system 3, which compares the gas purity and composition with the preset helium concentration and transmits the data to purification control system 4. If the value is lower than the preset helium concentration, a signal is fed back to purification system 2, which adjusts the system parameters in the chemical reaction and transmits these parameters to purification control system 4. Purification control system 4 automatically records the system parameters into a database and trains these data using machine learning to obtain a regression model. This allows for direct adjustment of purification system parameters during subsequent helium purification, achieving time-saving and convenient results. The adjusted parameters continue until the purity exceeds the preset helium concentration value, and finally, a quality report is output. When the trained regression model is used for the next purification of the mixed gas input device, the purification control system 4 outputs purification parameters to the purification system 2 for purification, and the detection and analysis system 3 detects and analyzes the output purified gas. Combining the detection and analysis of the input mixed gas, the purification control system 4 collects the parameters of each system to update the regression model.

[0056] The purification control system 4 in this embodiment uses a microcontroller.

[0057] Implementation Three

[0058] like Figure 3 As shown, the difference between this embodiment and embodiment two is that, based on embodiment two, this embodiment provides a further description of the helium purification equipment.

[0059] In this embodiment, the switching system 1 includes a main gas supply device 101, a secondary gas supply device 102, and an automatic switching valve group 103. The main gas supply device 101 and the secondary gas supply device 102 are each connected to one end of the automatic switching valve group 103, and the other end of the automatic switching valve group 103 is connected to the purification system 2. Both the main gas supply device 101 and the secondary gas supply device 102 are equipped with pressure detection devices 104. The main gas supply device 101 and the secondary gas supply device 102 ensure a continuous gas supply. In this embodiment, the pressure detection device 104 includes a local pressure gauge and a remote pressure sensor. The local pressure gauge is used for on-site observation, and the remote pressure sensor can remotely transmit the on-site pressure value from the central control room and has an alarm function.

[0060] The automatic switching valve assembly 103 includes a first switching branch 1031, a second switching branch 1032, a connecting branch 1033, a first pressure regulating valve 1034, and a second pressure regulating valve 1035. The main gas supply device 101 is connected to the connecting branch 1033 via the first switching branch 1031, and the auxiliary gas supply device 102 is connected to the connecting branch 1033 via the second switching branch 1032. The connecting branch 1033 is connected to the purification system 2. The first pressure regulating valve 1034 is located on the first switching branch 1031, and the second pressure regulating valve 1035 is located on the second switching branch 1032. A first flow switch 1036 is provided between the main gas supply device 101 and the first switching branch 1031, and a second flow switch 1037 is provided between the auxiliary gas supply device 102 and the second switching branch 1032. The first flow switch 1036 and the second flow switch 1037 are overflow switches; an alarm is triggered when the flow rate exceeds a set value. In use, the preset pressure is input. When the gas supply device in use senses that the gas pressure is lower than the set value through the pressure detection device 104 (i.e., the local pressure gauge and the remote pressure sensor), the remote pressure sensor outputs an alarm signal and automatically adjusts the flow switch to switch to another gas supply device. The pressure difference between the two switching devices is generally set to 0.5 bar. The first pressure regulating valve 1034 and the second pressure regulating valve 1035 have pressure regulating functions to ensure reliable gas supply. In addition, the automatic switching valve group 103 of this embodiment is equipped with another manual pressure regulating bypass, which is the third switching branch 1038. The third switching branch 1038 is equipped with a third pressure regulating valve 1039 and a third flow switch 1040. The pressure setting of the third switching branch 1038 is generally set 1 bar lower than that of the automatic switching pressure regulating valve to ensure 100% availability.

[0061] The purification system 2 in this embodiment includes a purification pipeline 201 and a metal getter 202. The purification pipeline 201 is connected to the switching system 1, and the metal getter 202 is located inside the purification pipeline 201. The detection and analysis system 3 is also connected to the purification pipeline 201. The purification system 2 mainly purifies the gas transported in the switching system 1. The metal getter 202 in the purification system 2 mainly adopts the principle of metal getter reaction, removing impurity gas atoms and molecules from the gas stream through active elements or surface chemical reactions. It uses high-performance getter purification materials and can detect various parameters during purification in real time, adjusting the parameter settings of the purification system 2 accordingly. Finally, through chemical reaction at a high temperature of 300°C, impurities such as water, oxygen, carbon monoxide, carbon dioxide, hydrogen, methane, and nitrogen in the gas are efficiently removed to below the PPB level, using helium gas with a purity of 99.999% as raw material, and purified to 99.9999999%.

[0062] The detection and analysis system 3 primarily monitors the purification pipeline 201 in the purification system 1 and analyzes the quality of the purified gas. During the stainless steel smelting and manufacturing process, approximately 200g of gas can be absorbed per ton. After the stainless steel is processed, not only are various contaminants adhering to its surface, but a certain amount of gas is also retained within its metal lattice. When airflow passes through the pipeline, this gas retained by the metal will re-enter the airflow, contaminating the pure gas. When the airflow inside the pipeline is discontinuous, the pipeline exerts pressure adsorption on the passing gas. When the airflow stops, the adsorbed gas undergoes depressurization and desorption, and the desorbed gas also enters the pure gas inside the pipeline as impurities. Simultaneously, the adsorption and desorption cycle repeats, causing a certain amount of powder to be generated on the inner surface of the pipeline. This metal dust particles also contaminate the pure gas inside the pipeline, thus requiring detection and analysis of the purification pipeline. The system receives purified gas and automatically activates gas analysis software to detect the gas composition in the pipeline and filling bottle online (the gas purified by purification system 2 is filled into the filling bottle, and the gas in the filling bottle is the purified gas), and promptly feeds back the detection value to the purification control system 4. Therefore, the purification equipment in this embodiment also includes an internal purging gas source 5, which is connected to the purification pipeline 201 through a high-purity pipeline 6. The purification pipeline 201 is purged by the internal purging gas source 5. In this embodiment, the internal purging gas source 5 includes an argon gas source and a helium gas source.

[0063] The other parts of this embodiment are the same as those in Embodiment 2, and will not be repeated here.

[0064] The working process of this invention is as follows: (1) When the pressure of the gas supply device in use is lower than that set by the automatic switching valve group 103 in the automatic switching system 1, the automatic switching valve group 103 automatically switches to another gas supply device to supply gas; the automatic switching valve group 103 also has a pressure reducing function, which can reduce the pressure of the helium gas supply device from 200 bar to the pressure required by the customer. (2) During purification: First, check whether the purifier is leaking gas, whether the function key is working, etc., then during the start-up of the purification system 2, use argon to purge the purification pipeline 201. Before reaching the operating temperature, exhaust gas with argon as the main component and other components such as water, CO, CO2 and other impurities is discharged through the exhaust port. Then, after the purification system 2 reaches the operating temperature, the argon is emptied, and helium is used for multiple replacements and purging. Next, after the purification system 2 is started up, it needs to be at least 30% of the rated flow (≥3Nm). 3Purification system 2 should be purged with helium gas at a rate of 1 / hr for 12-24 hours. Finally, after the purification system 2 has been debugged, it should be purged with 50%-70% helium gas for 6-8 weeks until the impurity index, such as CH4, is less than 1 ppb. (3) During testing and analysis: a. First, confirm that the gas source to be used for internal purging must meet or exceed the contract specifications; b. Before analysis, the system should be purged for at least 12 hours; c. Once the internal purging gas source has been confirmed to be qualified, the pipeline can be purged. Only high-purity pipelines and connectors can be used to connect the analyzer to the end of the pipeline for analysis; d. Confirm that the appropriate valves are open / closed; e. The analyzer must receive 100% analytical gas (without bypass) and maintain this for at least 20 minutes; f. Once the pipeline has passed the analysis, reduce the gas flow rate and remove the analyzer; g. Open / close the appropriate valves and set the regulator to a maximum of 20 psig; h. Fill out the test report and have it signed by a suitable person. The report must be kept in a suitable place for easy inspection at any time; During analysis: First, calibrate and activate the analyzer; then, connect the analyzer pipeline; next, provide real-time feedback information to the purification system based on the analysis results, so that it can change the corresponding parameters and ultimately achieve the required gas purity.

[0065] In summary, this invention provides a helium purification method that uses a regression model to predict purification parameters based on the concentration, temperature, composition of the mixed gas, and the desired concentration of pure helium. The purification system adjusts these parameters according to the model's predictions, allowing for the application of different purification parameters to different mixed gases, thus improving purification efficiency. Furthermore, since the purification parameters are predicted by the regression model, the method is highly automated. This invention also provides a helium purification device with a purification control system. This system records data from the purification system to train a regression model. The trained regression model outputs predicted purification parameters to control the purification system, resulting in high efficiency and automation.

[0066] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method of purifying helium gas, characterized by, The method comprises the following steps: S1: obtaining the pressure of the mixed gas; S2: when the pressure of the mixed gas exceeds a preset value, the mixed gas is transported to a purification system for purification; S3: the purification system predicts the purification parameters of the purification system according to the concentration, temperature, composition of the mixed gas and the required concentration of the purified helium gas, and purifies the mixed gas according to the predicted purification parameters; S4: analyzing the purity and composition of the purified mixed gas, and if the purity of the gas is not less than a preset value, the purified mixed gas is output; if the purity of the gas is less than the preset value, the step S3 is returned to continue purification; meanwhile, the regression model is updated by using the concentration, temperature, purification parameters, gas purity and composition of the mixed gas.

2. The helium purification method according to claim 1, wherein, The regression model is a Gaussian process regression model, and the acquisition process of the regression model is as follows: The collected data is formed into a data set, and the data set is divided into a training set and a test set. The collected data includes the gas concentration, composition ratio, temperature and purification parameters during purification, and each data collected is collected at a sampling point of the Gaussian process; Functions constructing a Gaussian process where are the gas concentration, the component ratio, the temperature parameter in the purification process, are the purification parameters, is the mean function, is the covariance function, denote different sampling points; the Gaussian process is determined by the mean function and the covariance function , assuming that the mean function is 0, the covariance function and the point conditional probability mean are obtained according to Bayesian inference and are constantly corrected; The expression of posterior probability determines the prediction point: where f(p,s,z..) is the observed value in the training set, f * is the predicted value, i.e. the purification parameter, and the input vector (p,s,z...) represents the parameter values of gas concentration, component ratio, temperature... in the purification process, respectively, k(p,s,z...p * ,s * ,z * ...) represents the covariance function of the training set and the covariance function of the prediction point (p * ,s * ,z * ...). The model is trained by the training set and the test set to obtain the trained regression model; for the trained regression model, a new vector (p, s, z...) is input, and the algorithm obtains a new predicted value f* according to the prediction model.

3. The helium purification method according to claim 2, wherein, In the training of the regression model, the regression model is evaluated by using k-fold cross-validation, and the accuracy of the regression model is calculated according to the mean absolute error formula; when the accuracy of the regression model reaches a preset requirement, the operation is stopped and the final regression model is directly output; otherwise, the proportion of the training set and the test set is returned to continue adjusting until the accuracy reaches the requirement.

4. The helium purification method according to claim 2, wherein, In the correction of the covariance function When correcting the covariance function and the point conditional probability mean, according to the Bayes formula, each point of the sample is incrementally calculated. For the initial point, the prior and likelihood are assumed, the posterior is calculated, and then the posterior estimate of the initial point is taken as the prior of the next estimate. This is repeated until the entire data set is calculated. With the increase of the sample, the estimate will be more and more accurate, and the correction is realized.

5. The helium purification method according to any one of claims 1 to 4, characterized by, The purification system adopts a metal getter reaction, and the purification parameters include the specific dosage of the getter, the proportion of each active element in the getter, and the proportion of the active element in the whole getter.

6. A helium purification apparatus based on the method of any one of claims 1 to 5, characterized in that, It comprises: A switching system (1) for inputting the mixed gas and controlling the input of the mixed gas into a purification system (2) according to the pressure of the mixed gas; A purification system (2) for purifying the mixed gas; A detection and analysis system (3) for detecting the pipeline in the purification system (2) and analyzing the quality of the purified gas; A purification control system (4) for recording data, including the concentration, temperature and composition of the mixed gas input into the switching system (1), the purification parameters of the purification system (2), the purity and composition of the purified gas, and using the recorded data to train the regression model, outputting the predicted purification parameters of the purification system (2) according to the concentration, temperature and composition of the mixed gas input into the switching system (1) through the trained regression model, and adjusting the purification parameters of the purification system (2) according to the predicted purification parameters of the purification system (2).

7. The helium purification apparatus of claim 6, wherein, The switching system (1) comprises a main gas supply device (101), a secondary gas supply device (102) and an automatic switching valve group (103), the main gas supply device (101) and the secondary gas supply device (102) are connected with one end of the automatic switching valve group (103) respectively, the other end of the automatic switching valve group (103) is connected with the purification system (2), and the main gas supply device (101) and the secondary gas supply device (102) are both provided with a pressure detection device (104).

8. The helium purification apparatus of claim 7, wherein, The automatic switching valve group (103) comprises a first switching branch (1031), a second switching branch (1032), a connecting branch (1033), a first pressure regulating valve (1034) and a second pressure regulating valve (1035), the main gas supply device (101) is communicated with the connecting branch (1033) through the first switching branch (1031), the secondary gas supply device (102) is communicated with the connecting branch (1033) through the second switching branch (1032), the connecting branch (1033) is connected with the purification system (2), the first pressure regulating valve (1034) is arranged on the first switching branch (1031), the second pressure regulating valve (1035) is arranged on the second switching branch (1032), a first flow switch (1036) is arranged between the main gas supply device (101) and the first switching branch (1031), and a second flow switch (1037) is arranged between the secondary gas supply device (102) and the second switching branch (1032).

9. The helium purification apparatus of claim 6, wherein, The purification system (2) comprises a purification pipeline (201) and a metal getter device (202), the purification pipeline (201) is communicated with the switching system (1), the metal getter device (202) is arranged in the purification pipeline (201), the detection and analysis system (3) is communicated with the purification pipeline (201).

10. The helium purification apparatus of claim 9, wherein, An internal blowing gas source (5) is further included, the internal blowing gas source (5) is connected with the purification pipeline (201) through a high-purity pipeline (6).

Images