Multi-component high-precision gas filling system

Abstract

The invention relates to the technical field of gas filling, and discloses a multi-component high-precision gas filling system which comprises a data module used for obtaining the residual gas content, the environment temperature and humidity and the filled gas pressure of the gas filling system, a collaboration module used for receiving the residual gas content, establishing a component interaction matrix and obtaining the component interaction matrix. The cross reaction amount of the residual gas and the newly-filled gas is calculated, the environment temperature and humidity are received, an environment compensation matrix is established, the cleaning effect weight is output through the environment compensation matrix, the gas filling efficiency is obtained by receiving the filled gas pressure in real time through an effect matrix, and the system health weight is output; the pollution module receives the cross reaction amount, the cleaning effect weight and the system health weight to establish a pollution model, the pollution model outputs and predicts the pollution amount of the gas filling system, the predicted pollution amount is input into the execution module, and the adjustment and optimization module is used for dynamically adjusting the component interaction matrix, the environment compensation matrix and the efficacy matrix; and pollution caused by multi-group gas filling is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of multi-component gas filling pollution detection and discloses a multi-component high-precision gas filling system. Background Art

[0002] Demand for high-precision, multi-component gas filling systems is growing in fields such as semiconductor manufacturing, medical gas supply, and chemical synthesis. For example, semiconductor wafer manufacturing requires the use of electronic specialty gases with specified purity requirements, and the ratio accuracy of multi-component mixed gases must reach the ppm level. The purity requirements for ternary gas mixtures in the medical field are equally stringent. However, existing technologies still have many shortcomings. Traditional systems lack quantitative analysis of cross-reactions between residual gases and newly filled gases, resulting in chemical reaction contamination during multi-component gas filling. Changes in ambient temperature and humidity during the filling process can affect physical properties such as gas density and compressibility. Existing systems lack dynamic compensation mechanisms, and cylinder cleaning relies on natural pressure release and mechanical pumping, which takes up to several hours. Residual gas in the pipeline cannot be effectively handled, resulting in a waste of resources and the potential for secondary contamination. There is also a lack of real-time monitoring of filling efficiency and equipment health. For example, membrane presses have long mixing times and high labor intensity, and compressor leaks can lead to reduced gas purity, making it difficult to meet the high-frequency, high-precision industrial needs. Summary of the Invention

[0003] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0004] In order to solve the above technical problems, the main purpose of the present invention is to provide a multi-component high-precision gas filling system, comprising: Data module, collaboration module, pollution module, execution module and tuning module; The data module is used to obtain the residual gas content, ambient temperature and humidity, and filling gas pressure of the gas filling system; The collaborative module receives the residual gas content, establishes a component interaction matrix, calculates the cross-reaction amount between the residual gas and the newly filled gas, and receives the ambient temperature and humidity to establish an environmental compensation matrix. The environmental compensation matrix outputs the cleaning effect weight. The effectiveness matrix obtains the gas filling efficiency by receiving the filling gas pressure in real time and outputs the system health weight. The pollution module receives the cross-reaction amount, the cleaning effect weight and the system health weight to establish a pollution model, outputs the predicted pollution amount of the gas filling system from the pollution model, and inputs the predicted pollution amount into the execution module; The tuning module is used to dynamically adjust the component interaction matrix, the environmental compensation matrix and the effectiveness matrix.

[0005] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: The data module includes a laser spectrum sensor, a temperature and humidity coupling sensor and a pressure transmitter; The laser spectrum sensor is used to detect the composition and concentration of residual gas in the pipeline; The temperature and humidity coupling sensor is used to collect environmental temperature and humidity data; The pressure transmitter is used to monitor the gas pressure changes during the filling process in real time; The data module transmits the composition and concentration of the residual gas, ambient temperature and humidity data, and gas pressure changes to the coordination module.

[0006] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: The collaborative module establishes a component interaction matrix, with all chemical components in the residual gas as row vectors and all chemical components of the new gas components to be filled as column vectors, constructing a two-dimensional interaction matrix, with the blanks filled with zeros. The value of each cell in the matrix represents the interaction intensity level of physical adsorption and chemical reaction between the corresponding row and column; The interaction strength level is matched with the molecular polarity, chemical bond activity and steric hindrance characteristics of each component through the gas physicochemical property library to identify the type of intermolecular force; the interaction strength is divided into discrete quantitative levels; the component combination with catalytic effect is The component interaction matrix uses the concentration of each component in the residual gas as a row weight vector and the concentration of the newly filled gas component as a column weight vector. The component interaction matrix establishes a fully connected topological network of residual and newly filled components, encodes the binary interaction relationship between any two components through matrix units, and autonomously captures unexpected cross-contamination through the binary interaction relationship.

[0007] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: The collaborative module establishes an environmental compensation matrix, constructing a two-dimensional environmental state matrix with temperature gradient intervals as row dimensions and humidity gradient intervals as column dimensions, where each matrix unit corresponds to an environmental action coefficient under a temperature and humidity combination; The environmental effect coefficient includes temperature effect, humidity effect and temperature and humidity synergistic effect; The calculation method of the cleaning effect weight is to map the real-time collected ambient temperature and humidity data to matrix units, obtain the basic environmental effect coefficient by calculating the value of each matrix unit, and compensate for the sensor measurement deviation through the adjacent unit gradient interpolation algorithm.

[0008] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: Constructing an effectiveness matrix with the pressure operating range as the row dimension and the filling stage as the column dimension through the collaborative module; The generation of the system health weight includes pressure fluctuation feature extraction and a dual-path evaluation mechanism; The pressure fluctuation feature extraction is performed by extracting real-time pressure data and separating the steady-state pressure value and the dynamic fluctuation component through time-frequency analysis; The dual-path evaluation mechanism includes the sharing of steady-state pressure values ​​and dynamic fluctuation components, wherein the steady-state pressure values ​​are mapped to the row dimension of the effectiveness matrix. An effectiveness matrix is ​​composed of multiple groups of steady-state pressure values, and the value of the calculation matrix is ​​the basic working condition health coefficient; The dynamic fluctuation component is calculated by spectrum entropy to generate a pressure fluctuation anomaly index; The calculation expression of the system health weight is: System health weight = α × (basic working condition health coefficient) + β × (pressure fluctuation abnormality index), where α is the basic working condition health weight and β is the pressure fluctuation abnormality index weight; When the system health weight is abnormal, nonlinear compensation is triggered; By establishing a CNN network, the system health weight is input into the output layer, the convolution layer and maximum pooling layer of the CNN network calculate the system health weight prediction value, and the output layer outputs the health weight prediction value.

[0009] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: The pollution model includes base value analysis layer, environmental regulation layer and equipment coupling layer; The base value analysis layer receives the cross-reaction amount input and uses the cross-reaction amount as the chemical pollution base value to represent the theoretical total amount of pollutants generated by the interaction of gas components; The environmental adjustment layer is used to receive the cleaning effect weight, dynamically correct the chemical pollution base value through the environmental attenuation function, and obtain the corrected pollution amount; The device coupling layer is used to receive the system health weight input and execute the segmented degradation amplification mechanism. When the system health weight is higher than the safety threshold, the linear amplification mode is adopted; when the system health weight is lower than the safety threshold, the amplification mode is activated.

[0010] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: The pollution model also includes a pollution risk index synthesis layer, which integrates the output through third-order coupling calculation; The third-order coupling calculation method is: S1, the chemical pollution base value of the receiving base value analysis layer, the corrected pollution value of the environmental adjustment layer, and the equipment amplification factor of the equipment coupling layer; S2. Map the chemical pollution base value to the pollution source intensity factor through the chemical pollution database, map the corrected pollution amount to the environmental inhibition factor through the pollution amount database, and map the equipment amplification factor to the system degradation factor through the system aging database. The comprehensive pollution risk index is calculated based on the weight distribution. The calculation method is: Comprehensive pollution risk index = pollution source intensity factor × × system degradation factor, where γ is the environmental inhibition sensitivity index, which is generated from the historical pipeline pollution data.

[0011] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: The pollution model also includes a temporal risk prediction layer, which implements dynamic risk prediction through a convolutional neural network architecture; The input layer reconstructs the comprehensive pollution risk index sequence within the continuous time window into a two-dimensional matrix of time features, where: The time dimension includes the comprehensive pollution risk index of the current moment and the previous N historical moments; The characteristic dimension embeds pipeline pressure fluctuations and temperature gradients; The convolution layer includes short-term convolution kernel, medium-term convolution kernel and long-term convolution kernel; The short-time convolution kernel is used to extract the burst pollution pattern; The mid-time convolution kernel is used to capture the periodic fluctuations of pollution; The long-term convolution kernel is used to identify the trend evolution of pollution; The feature fusion layer performs feature fusion by receiving sudden pollution patterns, periodic fluctuations of pollution, and pollution trend evolution. The output of the feature fusion layer is mapped into the comprehensive pollution risk prediction value for the future Δt period and the confidence weight of the current system health status through a fully connected network.

[0012] As a preferred solution of the multi-component high-precision gas filling system of the present invention, wherein: The tuning module dynamically adjusts the component interaction matrix. The pollution module outputs the deviation between the actual pollution amount and the predicted value for tuning. The component interaction matrix quantifies the reaction possibility between the residual gas and the new gas through the interaction intensity level. When the actual pollution amount deviates from the predicted pollution amount, the tuning module corrects the interaction intensity level of the corresponding component combination in the component interaction matrix. The environmental compensation matrix outputs the cleaning effect weight, reflecting the cleaning efficiency under different temperature and humidity conditions. The tuning module combines the residual gas concentration data detected by the monitoring unit after cleaning. If the actual cleaning effect does not meet expectations, the environmental effect coefficient of the corresponding temperature and humidity range in the environmental compensation matrix is ​​adjusted; The effectiveness matrix generates the system health weight, which is used to evaluate the filling efficiency and equipment health. When the correlation coefficient between the predicted health weight and the actual pollution amount is lower than the system rated value, the tuning module re-divides the pressure operating range, adjusts the weight distribution of the steady-state pressure value and the dynamic fluctuation component, and reorganizes the frequency bands for spectral entropy calculation.

[0013] A multi-component high-precision gas filling device comprises: A multi-component high-precision gas filling cabinet comprises: a first exhaust pipe, a controller, multiple gas monitoring screens, a gas charging regulator, a pressure gauge, a first vent, a second vent, a PID controller, and a gas inlet heat dissipation port; The humidity detector is installed on one side of the gas filling cabinet; The negative pressure switch detection end is connected to the filling chamber inside the cabinet through the air path; the control end is connected to the controller and PID controller through the bus to provide real-time feedback on the negative pressure status of the chamber; The flame detector's detection end faces the filling chamber inside the cabinet, and the warning end is directly connected to the sprinkler and controller through the control bus. Once the flame characteristics are identified, the sprinkler is immediately triggered and the gas filling regulator is linked to shut down. The third ventilation port is arranged at the bottom of the control cabinet.

[0014] Beneficial effects of the present invention: The present invention sets a collaborative module to calculate the chemical reaction amount of residual gas and new gas in real time through the component interaction matrix, combines the pollution model to predict the pollution amount, and guides the execution module to start the blocking unit to intercept the reaction products. The present invention also sets an environmental compensation matrix to adjust the cleaning effect weight according to real-time temperature and humidity data; The present invention sets up a multi-level cleaning system, in which the basic unit performs the standard cleaning process, the enhancement unit shortens the cleaning time by heating displacement and molecular pumping, the monitoring unit detects the content of multi-component gases in real time, reduces the residual gas concentration, and the effectiveness matrix evaluates the health status of the system through filling pressure data; The present invention sets a tuning module to iteratively update the component interaction matrix, environmental compensation matrix and effectiveness matrix according to historical data, thereby improving the intelligent performance of the multi-group gas filling system, while reducing manual detection errors and reducing the risk of multi-group gas filling residues affecting subsequent gas filling. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them: Figure 1This is a system composition diagram of a multi-component high-precision gas filling system of the present invention; Figure 2 This is a working schematic diagram of a multi-component high-precision gas filling system of the present invention; Figure 3 This is a flow chart of a method for determining the cleaning effect of a multi-component high-precision gas filling system according to the present invention; Figure 4 This is a front view of a multi-component high-precision gas filling device of the present invention; Figure 5 This is a side view of a multi-component high-precision gas filling device of the present invention; Figure 6 This is an oblique view of a multi-component high-precision gas filling device of the present invention; Figure 7 This is a top view of a multi-component high-precision gas filling device of the present invention.

[0016] Figure numerals: 1. First exhaust pipe; 2. Controller; 3. Multiple gas monitoring screens; 4. Gas charging regulator; 5. Pressure gauge; 6. First ventilation port; 7. Second ventilation port; 8. PID controller; 9. Gas inlet and heat dissipation port; 10. Temperature and humidity detector; 11. Negative pressure switch; 12. Flame detector; 13. Sprinkler; 14. Third ventilation port. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0020] Example 1 like Figure 1 As shown, a multi-component high-precision gas filling system includes: Data module, collaboration module, pollution module, execution module and tuning module.

[0021] like Figure 2As shown, the data module is used to obtain the residual gas content, ambient temperature and humidity, and filling gas pressure of the gas filling system.

[0022] Specifically, the data module includes a laser spectrum sensor, a temperature and humidity coupling sensor, and a pressure transmitter; The laser spectrum sensor is used to detect the composition and concentration of residual gas in the pipeline; The temperature and humidity coupling sensor is used to collect ambient temperature and relative humidity data to obtain ambient temperature and humidity data; The pressure transmitter monitors the gas pressure changes during the filling process in real time; The data module transmits the composition and concentration of the residual gas, ambient temperature and humidity data, and gas pressure changes to the coordination module.

[0023] The collaborative module establishes a component interaction matrix by receiving the residual gas content, calculates the cross-reaction amount of the residual gas and the newly filled gas, establishes an environmental compensation matrix by receiving the ambient temperature and humidity, and outputs the cleaning effect weight through the environmental compensation matrix. The effectiveness matrix obtains the efficiency of gas filling by receiving the filling gas pressure in real time and outputs the system health weight.

[0024] Specifically, the collaborative module establishes a component interaction matrix, with all chemical components in the residual gas as row vectors and all chemical components of the new gas components to be filled as column vectors, constructing a two-dimensional interaction matrix, filling the blanks with 0, and the value of each cell in the matrix represents the interaction intensity level of physical adsorption and chemical reaction between the corresponding row and column.

[0025] The interaction strength level is determined by matching the molecular polarity, chemical bond activity, and steric hindrance characteristics of each component with the gas physicochemical property library to identify the types of intermolecular forces that may occur. The interaction strength is divided into discrete quantitative levels based on the reaction thermal tendency and kinetic rate threshold. For component combinations with catalytic effects, their interaction strength level is improved.

[0026] Among them, through the gas physicochemical property library, the molecular polarity, chemical bond activity and steric hindrance characteristics of each component are matched from public databases or industry standard databases.

[0027] The component interaction matrix uses the concentration of each component in the residual gas as a row weight vector and the concentration of the newly filled gas component as a column weight vector, and calculates the overall cross-reaction amount through matrix bilinear weighting.

[0028] The calculation process preferentially couples the interacting units of high-concentration residual components with highly active new components and suppresses the contribution of inert component combinations.

[0029] The initial value of the matrix is ​​generated by the gas characteristics database. During operation, the deviation between the actual pollution amount output by the pollution module and the predicted pollution amount is automatically corrected to achieve online self-calibration of the matrix parameters.

[0030] Furthermore, the component interaction matrix establishes a fully connected topological network of residual and newly added components, encodes the binary interaction relationship between any two components through matrix units, and autonomously captures unexpected cross-contamination through the binary interaction relationship, avoiding the limitation of the traditional method that requires pre-set reaction paths.

[0031] Furthermore, the multiplication and addition operation of the weight vector and the matrix is ​​actually row weighting and column weighting; Among them, row weighting is used to reflect the pollution contribution of high-concentration residual components; column weighting is used to enhance the inducing ability of highly active new components; by setting a nonlinear inhibition function, such as Sigmoid threshold, low-risk interactions are filtered to reduce computational complexity.

[0032] The collaborative module establishes an environmental compensation matrix, constructing a two-dimensional environmental state matrix with temperature gradient intervals as row dimensions and humidity gradient intervals as column dimensions, and each matrix unit corresponds to an environmental action coefficient under a temperature and humidity combination.

[0033] Specifically, the environmental effect coefficient includes temperature effect, humidity effect and temperature-humidity synergistic effect; Furthermore, this application provides a preferred method for generating an environmental effect coefficient: Specifically, the temperature effect quantifies the effect of temperature on the desorption efficiency of residual gas based on the nonlinear relationship between the thermal motion intensity of gas molecules and the adsorption energy of metal surfaces at different temperatures; Furthermore, the humidity effect is based on the competitive adsorption characteristics of water molecules on the metal surface, and the influence of humidity on the inhibition / promotion of the desorption of residual gas components is established; Furthermore, the temperature-humidity synergistic effect sets an enhanced attenuation coefficient for high temperature and high humidity areas, reflecting the exponential amplification effect of water vapor thermal activation on the corrosion rate.

[0034] The calculation method of the cleaning effect weight is to map the real-time collected ambient temperature and humidity data to the corresponding matrix unit, obtain the basic environmental effect coefficient, and compensate the sensor measurement deviation through the adjacent unit gradient interpolation algorithm; In this application, a preferred specific implementation method of the neighboring unit gradient interpolation algorithm includes: The continuous temperature and humidity space is divided into gradient intervals. For example, since the gas adsorption / desorption process has a phase change inflection point at a specific temperature and humidity threshold, and the discretization processing enhances the robustness of the system to nonlinear responses, the temperature and humidity intervals are divided in this way. Those skilled in the art can divide the temperature and humidity intervals according to the actual production needs in this field.

[0035] The matrix unit stores the environmental action coefficient calibrated by experiments, which represents the ratio of the theoretical cleaning efficiency to the standard working condition under the environmental combination.

[0036] Furthermore, the filling pipeline material property correction factor and the historical cleanliness attenuation memory factor are superimposed to output the final cleaning effect weight.

[0037] In this application, a preferred specific implementation method for outputting the final cleaning effect weight includes: According to the surface energy database of pipeline materials, for example, stainless steel is AJ / m 2 , aluminum alloy is BJ / m 2 , dynamically scale the material correction factor based on different pipeline materials.

[0038] The surface energy database of pipeline materials is derived from calibration methods or industry reference standards.

[0039] A time decay function is introduced to record the recent cleaning effect deviation and suppress weight oscillation caused by environmental mutations.

[0040] In this application, a preferred specific implementation method of the environmental compensation matrix also includes: Based on temperature and humidity, the influencing factors of cleaning time and cleaning agent type can be increased. When the cleaning time is extended, the weight of the cleaning effect is strengthened; when high-purity cleaning agents are used, the weight is increased, and when ordinary cleaning agents are used, the weight is appropriately reduced.

[0041] The collaborative module constructs an effectiveness matrix, which is a three-dimensional effectiveness tensor (pressure × filling stage × health weight) with the pressure condition interval as the row dimension and the filling stage as the column dimension, where: The pressure operating condition range is divided into low pressure operating condition, standard operating condition and overpressure operating condition according to the filling pressure threshold; The charging stage is divided into four stages: initialization, boost, voltage stabilization, and termination; The generation of system health weight includes pressure fluctuation feature extraction and dual-path evaluation mechanism.

[0042] Specifically, pressure fluctuation feature extraction extracts real-time pressure data and separates steady-state pressure values ​​from dynamic fluctuation components through time-frequency analysis; Furthermore, the dual-path evaluation mechanism includes the sharing of steady-state pressure values ​​and dynamic fluctuation components, where the steady-state pressure values ​​are mapped to the row dimension of the matrix to obtain the basic working condition health coefficient.

[0043] Furthermore, the dynamic fluctuation component is calculated through spectral entropy to generate the pressure fluctuation anomaly index.

[0044] In this application, a preferred specific implementation method for spectrum entropy calculation includes: The pressure fluctuation time domain signal is decomposed into frequency band energy by wavelet transform, and the Shannon entropy of the energy distribution is calculated. The increase in entropy indicates the enhancement of randomness.

[0045] In this application, a preferred method for calculating the system health weight specifically includes: System health weight = α × (basic operating condition health coefficient) + β × (pressure fluctuation abnormality index), where α is the basic operating condition health weight and β is the pressure fluctuation abnormality index weight. Both α and β are set by the system rated parameters.

[0046] Specifically, the dynamic weight allocation (α, β) is as follows: α=0.7, β=0.3 in the voltage stabilization stage; α=0.4, β=0.6 in the voltage boost stage; when the abnormality index is greater than 0.8, nonlinear compensation is triggered:

[0047] in, is the nonlinear compensation value, is the previous nonlinear compensation value, e is the exponential constant, k is the linear compensation coefficient, and INDEX is the standard index.

[0048] By building a CNN network, the system health weight is input from the output layer, and the convolution layer of the CNN network The health weight prediction value of the system is calculated by the maximum pooling layer, and the health weight prediction value is output by the output layer.

[0049] In this application, a preferred CNN network implementation method includes: The convolutional neural network consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0050] The input layer receives the preprocessed system health weights; The convolution layer uses multiple convolution kernels of different sizes to perform convolution operations on the system health weight to extract features at different levels. For example, convolution kernels of sizes such as 3×3 and 5×5 are used to extract local features of the system health weight.

[0051] The pooling layer uses the maximum pooling operation to downsample the features output by the convolutional layer.

[0052] The fully connected layer flattens the features output by the pooling layer and uses them as input, and further integrates and processes the features through fully connected neurons.

[0053] The output layer consists of a single neuron, which outputs the final predicted value of the system health weight through an activation function. During the training phase, the mean square error is selected as the loss function, and the stochastic gradient descent optimization algorithm is used to iteratively update the network parameters to minimize the error between the predicted value and the true value, while also avoiding overfitting.

[0054] The correlation analysis between the predicted value of the system health weight and the actual pollution amount is carried out in each cycle. When the correlation coefficient is less than the rated pollution amount of the grouped gas filling system, the pressure operating condition range optimization, basic operating condition health weight and abnormal index weight coefficient redistribution, and spectrum entropy band reorganization are automatically started to output the system health nonlinear compensation value of different component gas fillings to optimize the efficiency of multi-group gas filling.

[0055] The compensation coefficient of the pressure fluctuation anomaly index amplification mechanism in the effectiveness matrix is ​​dynamically adjusted according to whether the system detects a physical abnormality signal, enhancing the response sensitivity in abnormal situations and reducing the probability of misjudgment in normal situations.

[0056] In this application, a preferred specific implementation method of the effectiveness matrix also includes: Pressure fluctuation signals are classified according to their frequency characteristics to distinguish fluctuations caused by different reasons such as valve switching, unstable gas source, and equipment leakage. Differentiated weights are set for each type of fluctuation to avoid misjudging normal operation fluctuations as equipment abnormalities or ignoring key fault signals, thereby improving the accuracy of system health assessment.

[0057] The pollution module receives the cross-reaction amount, the cleaning effect weight and the system health weight to establish a pollution model, and the pollution model outputs a predicted pollution amount of the gas filling system, and the predicted pollution amount is input into the execution module; Specifically, the pollution model includes a base value analysis layer, an environmental adjustment layer, and a device coupling layer; Specifically, the base value analysis layer receives the cross-reaction amount input and uses the cross-reaction amount as the chemical pollution base value to represent the theoretical total amount of pollutants produced by the interaction of gas components; The environmental adjustment layer is used to receive the cleaning effect weight and dynamically correct the chemical pollution base value through the environmental attenuation function to obtain the corrected pollution amount; Specifically, a preferred implementation method of the environmental attenuation function includes: The environmental attenuation function is used to establish a quantitative mapping model between cleaning ability and pollution retention. The principle is: the cleaning effect weight is used as the input variable, with a definition domain of [0,1], and the pollution retention ratio coefficient is output, with a value range of [0,2], where: the cleaning effect weight = 1.0 (optimal cleaning state) mapping output coefficient = 1.0 (zero pollution attenuation); the cleaning effect weight = 0.0 (no cleaning at all) mapping output coefficient = 2.0 (double pollution retention).

[0058] like Figure 3 As shown, the cleaning status is judged by the cleaning effect weight. If the cleaning effect weight is not less than the system rated cleaning weight, it enters the cleaning linear attenuation area and the gas filling system is cleaned by the cleaning agent coverage or the filling of inert gas; If the cleaning effect weight is less than the system rated cleaning weight, the system cleaning capacity is insufficient and the pollution changes suddenly.

[0059] Among them, the higher the cleaning effect weight, the stronger the environmental attenuation effect, and the lower the actual retained pollution amount, ultimately achieving a negative correlation mapping between cleaning ability and pollution residue; The device coupling layer is used to receive the system health weight input and execute the segmented degradation amplification mechanism: when the system health weight is higher than the safety threshold, the linear amplification mode is adopted; when the system health weight is lower than the safety threshold, the amplification mode is activated.

[0060] In linear amplification mode, when the health weight is no less than the safety threshold, the system is considered healthy, and the impact of equipment degradation on pollution is linear. The device coupling layer converts the system health weight into an equipment amplification factor, which is used to correct the base pollution level of the pollution model. Good system health includes, but is not limited to, stable equipment operation, no significant anomalies, and mild and predictable degradation.

[0061] Exponential amplification mode: when the health weight is lower than the safety threshold, the impact of equipment degradation on pollution shows a nonlinear aggravation trend. The equipment coupling layer activates the amplification mode and increases the equipment amplification factor corresponding to the system health weight according to the exponential law.

[0062] Equipment degradation includes abnormal pressure fluctuations, increased risk of pipeline corrosion, a rapid increase in the degree of degradation, and a sharp expansion of pollution risks, which strengthen pollution warnings under high-risk conditions through exponential amplification.

[0063] The pollution model also includes a pollution risk index synthesis layer, which integrates the outputs of each layer through a third-order coupling calculation. The specific third-order coupling calculation method is: S1, the chemical pollution base value of the receiving base value analysis layer, the corrected pollution value of the environmental adjustment layer, and the equipment amplification factor of the equipment coupling layer; S2. Map the chemical pollution base value to the pollution source intensity factor through the chemical pollution database, map the corrected pollution amount to the environmental inhibition factor through the pollution amount database, and map the equipment amplification factor to the system degradation factor through the system aging database. The comprehensive pollution risk index is calculated based on the weight distribution. The calculation method is: Comprehensive pollution risk index = pollution source intensity factor × × system degradation factor, where γ is the environmental inhibition sensitivity index, which is generated from the historical pipeline pollution data.

[0064] Specifically, the chemical pollution base value is converted into the pollution source intensity factor and directly mapped to the equivalent value (1:1); The corrected pollution amount is converted into an environmental inhibition factor, and the conversion logic is the reciprocal operation (1 / corrected pollution amount); The equipment amplification factor is converted into a direct equivalent mapping (1:1) of the system degradation factor.

[0065] The pollution model also includes a temporal risk prediction layer, which realizes dynamic risk prediction through a convolutional neural network architecture.

[0066] Specifically, the input layer reconstructs the comprehensive pollution risk index sequence within the continuous time window into a time-feature two-dimensional matrix, where: The time dimension includes the comprehensive pollution risk index of the current moment and the previous N historical moments; The feature dimension embeds auxiliary working condition features such as pipeline pressure fluctuation and temperature gradient; The convolution layer includes short-time convolution kernels, medium-time convolution kernels and long-time convolution kernels. Among them, the short-time convolution kernel is used to extract sudden pollution patterns, the medium-time convolution kernel is used to capture the periodic fluctuations of pollution, and the long-time convolution kernel is used to identify the trend evolution of pollution. The sudden pollution patterns, periodic fluctuations of pollution and trend evolution of pollution are received through the feature fusion layer, and the output of the feature fusion layer is mapped into the comprehensive pollution risk prediction value for the future Δt period and the confidence weight of the current system health status through a fully connected network.

[0067] Furthermore, an environmental inhibition sensitivity index is set in the pollution model, and the pollution model is pre-trained based on historical data of nearly M fillings.

[0068] The execution module includes a basic unit for executing a standard cleaning process, an enhancement unit for handling physical adsorption, a blocking unit for intercepting chemical reactions, and a monitoring unit for detecting the content of multiple groups of gases; Furthermore, the execution module receives control instructions through PID to control the basic unit, the enhancement unit and the blocking unit.

[0069] Specifically, in the execution module, when the comprehensive pollution risk prediction value is lower than the minimum risk threshold, the basic unit is started first, and based on the preset standard cleaning process, the basic content of residual gas is reduced through routine pipeline cleaning operations.

[0070] When the comprehensive pollution risk prediction value is greater than the minimum risk threshold and less than the maximum risk threshold, the enhancement unit and the monitoring unit operate in coordination. The enhancement unit adopts heating replacement, enhanced exhaust and other enhancement measures for physical adsorption to break the adsorption balance of residual gas on the pipeline surface and increase the cleaning depth. The monitoring unit detects the content changes of multi-component gases in real time and ensures the effect of enhanced cleaning through dynamic feedback to avoid pollution caused by incomplete treatment of physical adsorption residues.

[0071] If the comprehensive pollution risk prediction value exceeds the maximum risk threshold, the blocking unit will be activated immediately to actively intercept the pollutants generated during the reaction process in response to the violent cross-reactions that occur under high risks. Physical or chemical means will be used to prevent the reaction products from diffusing in the pipeline to prevent them from entering the new gas system and affecting the purity. At the same time, the real-time data of the monitoring unit will be used to ensure that the interception measures are accurate and effective.

[0072] Specifically, the risk threshold is determined based on the purity requirements of the industry application scenario, the system's own rated pollution tolerance, and the statistical results of pollution data accumulated during the historical filling process. The minimum and maximum risk thresholds are defined. The minimum risk threshold is the basic residual level acceptable to the system, and the maximum risk threshold is the critical value for initiating emergency interception. The threshold will be dynamically calibrated by the tuning module according to the actual operating data of the filling cycle to ensure the accuracy of the response.

[0073] The tuning module is used to dynamically adjust the component interaction matrix, the environmental compensation matrix and the effectiveness matrix.

[0074] Specifically, the tuning module dynamically adjusts the component interaction matrix based on the deviation between the actual pollution amount output by the pollution module and the predicted pollution amount. The component interaction matrix quantifies the reaction possibility between the residual gas and the newly filled gas through the interaction intensity level. Its initial value comes from the gas property database. When the actual pollution amount deviates from the predicted pollution amount, the tuning module corrects the interaction intensity level of the corresponding component combination in the component interaction matrix. For example, if the actual reaction pollution of a group of residual and newly filled gases exceeds the prediction, the interaction intensity level of the combination is increased, otherwise it is reduced, thereby realizing online self-calibration of the matrix parameters and enhancing the accuracy of the cross-reaction amount calculation.

[0075] For the environmental compensation matrix, the adjustment of the tuning module is based on the difference between the actual feedback of the cleaning effect weight and the expected effect. The environmental compensation matrix outputs the cleaning effect weight through the environmental effect coefficient, the pipeline material correction factor and the historical cleanliness attenuation memory factor to reflect the cleaning efficiency under different temperature and humidity conditions; further, the tuning module combines the actual data such as the residual gas concentration after cleaning detected by the monitoring unit. If the actual cleaning effect does not meet expectations, the environmental effect coefficient of the corresponding temperature and humidity range in the environmental compensation matrix is ​​adjusted to ensure that the cleaning effect weight can truly reflect the impact of the environment on the cleaning process and improve the adaptability of the cleaning strategy.

[0076] For the adjustment of the effectiveness matrix, the tuning module is based on the correlation analysis between the predicted value of the system health weight and the actual pollution amount. The effectiveness matrix generates the system health weight based on the pressure condition and the filling stage, which is used to evaluate the filling efficiency and equipment health. When the correlation coefficient between the health weight predicted value and the actual pollution amount is lower than the system rated value, the tuning module will start the optimization mechanism, re-divide the pressure condition interval to better fit the actual pressure change characteristics, adjust the weight distribution of the steady-state pressure value and the dynamic fluctuation component, and reorganize the frequency band of the spectral entropy calculation, so that the health weight can more accurately capture the abnormal status of the equipment and enhance the system's ability to predict filling efficiency and health risks.

[0077] Furthermore, the tuning module can also define a reasonable deviation range between the actual pollution amount and the predicted value, and start matrix correction only when the deviation exceeds the range and remains stable, so as to avoid frequent parameter fluctuations caused by measurement errors.

[0078] Receive multiple gas filling amounts through PID controller and control multiple gas filling; Furthermore, the gas filling amounts of multiple groups are obtained from the gas component ratios. The gas component ratio calculation method is:

[0079] in, is the gas component ratio, is the component pressure, is the cylinder volume, is the component molecular weight, R is the gas constant, T is the temperature, is the compression factor, The filling gas identifier is used to identify the type of gas currently being filled.

[0080] Example 2: like Figure 4 As shown, this embodiment provides a multi-component high-precision gas filling cabinet, including The first exhaust pipe 1 is installed on the left and right sides of the cabinet top. It is a pipe structure extending vertically upward and is used to connect to an external exhaust system to achieve directional extraction of gas in the cabinet.

[0081] The controller 2 is fixed in the right area of ​​the top of the cabinet. It is an independent control unit and is connected to various functional modules in the cabinet, such as the PID controller 8 and the gas monitoring screen, through a cable bus. It is used for overall system logic control and command scheduling.

[0082] Multiple sets of gas monitoring screens are embedded in the upper part of the functional area on the right side of the front of the cabinet. They are visual display components and are distributed in a side-by-side array. They are used to display real-time monitoring data such as gas composition and concentration.

[0083] The gas filling regulator 4 is located in the middle of the functional area on the right side of the cabinet front, right below the multiple gas monitoring screens 3. It is a knob-button adjustment component used to manually and automatically adjust the flow rate, pressure and other parameters of the filling gas to adapt to gas filling needs.

[0084] The pressure gauge 5 is integrated in the lower part of the functional area on the right side of the cabinet front, corresponding to the gas charging regulator 4 above and below. It is a pressure detection and display component, which provides real-time feedback on the pressure status of the charged gas or the cabinet cavity.

[0085] The first ventilation opening 6 and the second ventilation opening 7 are respectively embedded in the lower left and middle functional areas of the cabinet front. They are grille louver ventilation structures and are arranged horizontally. They are used for gas exchange and heat dissipation inside and outside the cabinet to ensure the stability of the internal environment.

[0086] The PID controller 8 is embedded in the middle of the left and middle functional areas on the front of the cabinet. It is an independent control module that works with the gas filling regulator 4, the gas monitoring screen 3, etc. to achieve precise PID adjustment of the gas filling process.

[0087] The gas inlet and heat dissipation port 9 is embedded in the upper left and middle functional areas of the cabinet front. It is a gas input and heat dissipation integrated component, which has both the gas inlet interface function and the heat dissipation grille structure to realize the gas diversion and internal heat dissipation.

[0088] The gas enters the cabinet through the gas heat dissipation port 9, and the PID controller 8 cooperates with the gas charging regulator 4 to accurately control the charging parameters; Multiple groups of gas monitoring screens 3 collect gas data in real time and feed it back to the controller 2. Abnormal conditions trigger control instructions; The first exhaust duct 1 cooperates with the first ventilation port 6 and the second ventilation port 7 to maintain air circulation and pressure balance in the cabinet, thereby ensuring stable operation of the system.

[0089] The third ventilation opening 14 is provided at the bottom of the control cabinet.

[0090] like Figure 5 As shown, the temperature sensor 10 is arranged on one side of the cabinet.

[0091] like Figure 6 As shown, the negative pressure switch 11 is embedded in the middle of the functional area on the right side of the cabinet front, immediately below the multiple groups of gas monitoring screens 3, and is horizontally aligned with the pressure gauge 5.

[0092] The detection end of the negative pressure switch 11 is connected to the filling chamber inside the cabinet through an air path; the control end is linked with the controller 2 and the PID controller 8 through a bus to provide real-time feedback on the negative pressure status of the chamber.

[0093] The flame detector 12 is integrated into the upper part of the functional area on the right side of the cabinet front, attached to the right side of the multiple gas monitoring screens 3, and perpendicular to the projection area of ​​the first exhaust pipe 1 on the top.

[0094] The detection end of the flame detector 12 faces the filling chamber inside the cabinet, and the early warning end is directly connected to the sprinkler 13 and the controller 2 through the control bus. Once the flame characteristics are identified, the sprinkler 13 is immediately triggered for emergency spraying, and the gas filling regulator 4 is linked to shut down.

[0095] like Figure 7 As shown, the sprayer 13 is arranged on the top of the cabinet.

[0096] It is important to note that the configuration and arrangement of the present application, as illustrated in various exemplary embodiments, are illustrative only. Although only two embodiments are described in detail in this disclosure, those reading this disclosure will readily appreciate that numerous modifications are possible without materially departing from the novel teachings and advantages of the subject matter described herein. For example, variations in the size, scale, structure, shape, and proportions of various components, as well as parameter values ​​(e.g., temperature, pressure, etc.), mounting arrangements, use of materials, color, orientation, and the like, are possible. For example, an element shown as being region-shaped may be comprised of multiple parts or elements, the position of an element may be inverted or otherwise altered, and the nature, number, or position of discrete elements may be modified or changed. Therefore, all such modifications are intended to be encompassed within the scope of this invention. The order or sequence of any process or method steps may be altered or reordered according to alternative embodiments. Any "means-plus-function" clause is intended to cover structures that perform the functions described herein, and not only structural equivalence but also structural equivalents. Other substitutions, modifications, changes, and omissions may be made in the design, operating conditions, and arrangement of the exemplary embodiments without departing from the scope of this invention. Therefore, the invention is not limited to the specific embodiments, but extends to various modifications that still fall within the scope of the appended claims.

[0097] Additionally, in order to provide a concise description of exemplary embodiments, all features of an actual embodiment may not be described (ie, those features that are not relevant to the best mode presently contemplated for carrying out the invention or those that are not relevant to implementing the invention).

[0098] It should be understood that in the development of any actual embodiment, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but for those of ordinary skill having the benefit of this disclosure, the development effort will be a routine task of design, fabrication, and production without undue experimentation.

[0099] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multi-component high-precision gas filling system, characterized in that: include: Data module, collaboration module, pollution module, execution module and tuning module; The data module is used to obtain the residual gas content, ambient temperature and humidity, and filling gas pressure of the gas filling system; The collaborative module receives the residual gas content, establishes a component interaction matrix, calculates the cross-reaction amount between the residual gas and the newly filled gas, and receives the ambient temperature and humidity to establish an environmental compensation matrix. The environmental compensation matrix outputs the cleaning effect weight. The effectiveness matrix obtains the gas filling efficiency by receiving the filling gas pressure in real time and outputs the system health weight. The pollution module receives the cross-reaction amount, the cleaning effect weight and the system health weight to establish a pollution model, outputs the predicted pollution amount of the gas filling system from the pollution model, and inputs the predicted pollution amount into the execution module; The tuning module is used to dynamically adjust the component interaction matrix, the environmental compensation matrix and the effectiveness matrix.

2. A multi-component high-precision gas filling system according to claim 1, characterized in that: The data module includes a laser spectrum sensor, a temperature and humidity coupling sensor and a pressure transmitter; The laser spectrum sensor is used to detect the composition and concentration of residual gas in the pipeline; The temperature and humidity coupling sensor is used to collect environmental temperature and humidity data; The pressure transmitter is used to monitor the gas pressure changes during the filling process in real time; The data module transmits the composition and concentration of the residual gas, ambient temperature and humidity data, and gas pressure changes to the coordination module.

3. A multi-component high-precision gas filling system according to claim 1, characterized in that: The collaborative module establishes a component interaction matrix, with all chemical components in the residual gas as row vectors and all chemical components of the new gas components to be filled as column vectors, constructing a two-dimensional interaction matrix, with the blanks filled with zeros. The value of each cell in the matrix represents the interaction intensity level of physical adsorption and chemical reaction between the corresponding row and column; The interaction strength level is matched with the molecular polarity, chemical bond activity and steric hindrance characteristics of each component through the gas physicochemical property library to identify the type of intermolecular force; the interaction strength is divided into discrete quantitative levels; the component combination with catalytic effect is The component interaction matrix uses the concentration of each component in the residual gas as a row weight vector and the concentration of the newly filled gas component as a column weight vector. The component interaction matrix establishes a fully connected topological network of residual and newly filled components, encodes the binary interaction relationship between any two components through matrix units, and autonomously captures unexpected cross-contamination through the binary interaction relationship.

4. A multi-component high-precision gas filling system according to claim 1, characterized in that: The collaborative module establishes an environmental compensation matrix, constructing a two-dimensional environmental state matrix with temperature gradient intervals as row dimensions and humidity gradient intervals as column dimensions, where each matrix unit corresponds to an environmental action coefficient under a temperature and humidity combination; The environmental effect coefficient includes temperature effect, humidity effect and temperature and humidity synergistic effect; The calculation method of the cleaning effect weight is to map the real-time collected ambient temperature and humidity data to matrix units, obtain the basic environmental effect coefficient by calculating the value of each matrix unit, and compensate for the sensor measurement deviation through the adjacent unit gradient interpolation algorithm.

5. The multi-component high-precision gas filling system according to claim 1, characterized in that: Constructing an effectiveness matrix with the pressure operating range as the row dimension and the filling stage as the column dimension through the collaborative module; The generation of the system health weight includes pressure fluctuation feature extraction and a dual-path evaluation mechanism; The pressure fluctuation feature extraction is performed by extracting real-time pressure data and separating the steady-state pressure value and the dynamic fluctuation component through time-frequency analysis; The dual-path evaluation mechanism includes the sharing of steady-state pressure values ​​and dynamic fluctuation components, wherein the steady-state pressure values ​​are mapped to the row dimension of the effectiveness matrix. An effectiveness matrix is ​​composed of multiple groups of steady-state pressure values, and the value of the calculation matrix is ​​the basic working condition health coefficient; The dynamic fluctuation component is calculated by spectrum entropy to generate a pressure fluctuation anomaly index; The calculation expression of the system health weight is: System health weight = α × (basic working condition health coefficient) + β × (pressure fluctuation abnormality index), where α is the basic working condition health weight and β is the pressure fluctuation abnormality index weight; When the system health weight is abnormal, nonlinear compensation is triggered; By establishing a CNN network, the system health weight is input into the output layer, the convolution layer and maximum pooling layer of the CNN network calculate the system health weight prediction value, and the output layer outputs the health weight prediction value.

6. A multi-component high-precision gas filling system according to claim 5, characterized in that: The pollution model includes base value analysis layer, environmental regulation layer and equipment coupling layer; The base value analysis layer receives the cross-reaction amount input and uses the cross-reaction amount as the chemical pollution base value to represent the theoretical total amount of pollutants generated by the interaction of gas components; The environmental adjustment layer is used to receive the cleaning effect weight, dynamically correct the chemical pollution base value through the environmental attenuation function, and obtain the corrected pollution amount; The device coupling layer is used to receive the system health weight input and execute the segmented degradation amplification mechanism. When the system health weight is higher than the safety threshold, the linear amplification mode is adopted; When the system health weight falls below the safety threshold, the amplification mode is activated.

7. A multi-component high-precision gas filling system according to claim 6, characterized in that: The pollution model also includes a pollution risk index synthesis layer, which integrates the output through third-order coupling calculation; The third-order coupling calculation method is: S1, the chemical pollution base value of the receiving base value analysis layer, the corrected pollution value of the environmental adjustment layer, and the equipment amplification factor of the equipment coupling layer; S2. Map the chemical pollution base value to the pollution source intensity factor through the chemical pollution database, map the corrected pollution amount to the environmental inhibition factor through the pollution amount database, and map the equipment amplification factor to the system degradation factor through the system aging database. The comprehensive pollution risk index is calculated based on the weight distribution. The calculation method is: Comprehensive pollution risk index = pollution source intensity factor × × system degradation factor, where γ is the environmental inhibition sensitivity index, which is generated from the historical pipeline pollution data.

8. A multi-component high-precision gas filling system according to claim 7, characterized in that: The pollution model also includes a temporal risk prediction layer, which implements dynamic risk prediction through a convolutional neural network architecture; The input layer reconstructs the comprehensive pollution risk index sequence within the continuous time window into a two-dimensional matrix of time features, where: The time dimension includes the comprehensive pollution risk index of the current moment and the previous N historical moments; The characteristic dimension embeds pipeline pressure fluctuations and temperature gradients; The convolution layer includes short-term convolution kernel, medium-term convolution kernel and long-term convolution kernel; The short-time convolution kernel is used to extract the burst pollution pattern; The mid-time convolution kernel is used to capture the periodic fluctuations of pollution; The long-term convolution kernel is used to identify the trend evolution of pollution; The feature fusion layer performs feature fusion by receiving sudden pollution patterns, periodic fluctuations of pollution, and pollution trend evolution. The output of the feature fusion layer is mapped into the comprehensive pollution risk prediction value for the future Δt period and the confidence weight of the current system health status through a fully connected network.

9. The multi-component high-precision gas filling system according to claim 1, characterized in that: The tuning module dynamically adjusts the component interaction matrix. The pollution module outputs the deviation between the actual pollution amount and the predicted value for tuning. The component interaction matrix quantifies the reaction possibility between the residual gas and the new gas through the interaction intensity level. When the actual pollution amount deviates from the predicted pollution amount, the tuning module corrects the interaction intensity level of the corresponding component combination in the component interaction matrix. The environmental compensation matrix outputs the cleaning effect weight, reflecting the cleaning efficiency under different temperature and humidity conditions. The tuning module combines the residual gas concentration data detected by the monitoring unit after cleaning. If the actual cleaning effect does not meet expectations, the environmental effect coefficient of the corresponding temperature and humidity range in the environmental compensation matrix is ​​adjusted; The effectiveness matrix generates the system health weight, which is used to evaluate the filling efficiency and equipment health. When the correlation coefficient between the predicted health weight and the actual pollution amount is lower than the system rated value, the tuning module re-divides the pressure operating range, adjusts the weight distribution of the steady-state pressure value and the dynamic fluctuation component, and reorganizes the frequency bands for spectral entropy calculation.

10. A multi-component high-precision gas filling system according to claim 9, characterized in that ： A multi-component high-precision gas filling system also includes a gas filling cabinet, which includes: A first exhaust pipe (1), a controller (2), multiple gas monitoring screens (3), a gas charging regulator (4), a pressure gauge (5), a first vent (6), a second vent (7), a PID controller (8), a gas inlet heat dissipation port (9), and a third vent (14) are arranged at the bottom of the control cabinet; The humidity detector (10) is arranged on one side of the gas filling cabinet; The detection end of the negative pressure switch (11) is connected to the filling chamber inside the cabinet through an air path; the control end is connected to the controller (2) and the PID controller (8) through a bus, and is used to provide real-time feedback on the negative pressure state of the chamber; The detection end of the flame detector (12) faces the filling chamber inside the cabinet, and the warning end is directly connected to the sprinkler (13) and the controller (2) through the control bus. Once the flame characteristics are recognized, the sprinkler (13) is immediately triggered and the gas filling regulator (4) is linked to shut down.

Images