High-safety full-automatic filling method and system applied to fluorine-nitrogen mixed gas filling

Abstract

The invention discloses a high-safety full-automatic filling method and system applied to fluorine and nitrogen mixed gas filling, and relates to the field of gas filling, and the method comprises the following steps: filling fluorine and nitrogen mixed gas into a target steel cylinder, collecting the filling quality in real time, and stopping when the target filling amount is reached. And active enhanced heat dissipation is applied to the steel cylinder in the first half period of each preset modulation period, heat dissipation is stopped in the second half period, and multiple complete periods are continuously and alternately executed. Meanwhile, the pressure in the bottle is continuously collected, the average pressure value of each period is calculated with the modulation period as an integral window, linear trend fitting is conducted on the average pressure value sequence of multiple continuous periods, and the pressure decline rate is obtained. If the decline rate exceeds a preset leakage judgment threshold value, it is judged that leakage exists in the steel cylinder; and if not, determining that the sealing is qualified, and completing filling. By implementing the method, the accuracy of detecting the leakage state of the steel cylinder can be improved.

Description

Technical Field

[0001] This application relates to the field of gas filling, and in particular to a high-safety fully automatic filling method and system for filling fluorine-nitrogen mixed gases. Background Technology

[0002] Fluorine (F2) is an irreplaceable key process gas in semiconductor manufacturing, photovoltaic industry, and nuclear fuel processing. Due to its extremely strong oxidizing and corrosive properties, the sealing test during its filling, storage, and transportation is a core aspect of ensuring safe production.

[0003] In related technologies, the industrial testing of the sealing performance of filled steel cylinders mainly adopts the static pressure holding method. That is, after filling, the cylinder valve is closed, the initial pressure value is recorded, and after a period of pressure holding, the pressure value is read again. If the pressure drop exceeds a preset threshold, it is determined to be a leak.

[0004] However, during the high-pressure filling process, the gas compression continuously injects heat into the cylinder, and the cylinder temperature is significantly higher than the ambient temperature at the end of filling. In the subsequent pressure-holding detection phase, the cylinder continues to dissipate heat, causing its temperature to drop. According to the gas law, under sealed, constant-volume conditions, this temperature drop directly leads to a continuous decrease in pressure inside the cylinder. Since heat dissipation from the cylinder follows Newton's law of cooling, the rate of pressure reduction caused by the thermal effect is a slow variable that continuously decreases as the temperature difference narrows. Within a short time window, the contribution of the thermal effect between windows is approximately constant, but it accumulates significantly within the longer observation window required by the pressure-holding method. This thermal effect pressure drop and the leakage pressure drop are superimposed on the same pressure reading, and their magnitudes are on the same order of magnitude. This causes well-sealed cylinders to be mistakenly judged as leaking due to the thermal effect pressure drop exceeding the threshold, or, to avoid false alarms, the threshold is raised, and cylinders that are actually leaking pass the detection because the leakage pressure drop is masked by the thermal effect background. Extending the waiting time until the temperature reaches complete equilibrium to reduce the thermal effect is also not feasible due to the safety risks of continuous fluorine gas escape and the constraints of production line efficiency. Summary of the Invention

[0005] This application provides a high-safety fully automatic filling method and system for filling fluorine-nitrogen mixed gases, which improves the accuracy of cylinder leakage detection.

[0006] Firstly, this application provides a high-safety fully automatic filling method for filling fluorine-nitrogen mixed gas, applied to an automatic filling system. The method includes: filling a target cylinder with fluorine-nitrogen mixed gas; collecting real-time data on the filling quality of the target cylinder; stopping filling when the collected filling quality data reaches the target filling amount; applying active enhanced heat dissipation to the target cylinder during the first half of each preset modulation cycle; stopping active heat dissipation during the second half of each preset modulation cycle; continuously alternating multiple complete preset modulation cycles; continuously collecting the internal pressure of the target cylinder; using the preset modulation cycle as an integration window; calculating the cycle-average pressure value corresponding to each modulation cycle based on the internal pressure data collected during each complete modulation cycle; performing linear trend fitting on the pressure value sequence formed by the cycle-average pressure values ​​corresponding to multiple consecutive preset modulation cycles to obtain a pressure deceleration rate; determining that the target cylinder leaks when the pressure deceleration rate exceeds a preset leakage judgment threshold; and determining that the target cylinder is sealed correctly and the filling of the fluorine-nitrogen mixed gas is completed when the pressure deceleration rate does not exceed the preset leakage judgment threshold.

[0007] In the above embodiments, active enhanced heat dissipation is applied to the target cylinder during the first half of each preset modulation cycle, and active heat dissipation is stopped during the second half of the cycle. The cycle-averaged pressure value is calculated using the preset modulation cycle as the integration window, and a linear trend fitting is performed on the pressure value sequence composed of multiple consecutive cycle-averaged pressure values ​​to obtain the pressure reduction rate. This decouples the thermal effect pressure drop caused by the continuous heat dissipation of the cylinder after filling from the leakage pressure drop. At the end of filling, the cylinder temperature rises due to the work done by gas compression, and the heat dissipation of the cylinder during the pressure holding stage causes the internal pressure to continuously decrease. This thermal effect pressure drop is superimposed on the leakage pressure drop in the same pressure reading. The cycle-averaged calculation using the modulation cycle as the integration window makes the net contribution of the thermal effect of active enhanced heat dissipation in the first half of the cycle and natural heat dissipation in the second half of the cycle tend to be consistent in each integration window; the slope obtained by linear fitting on the sequence of consecutive cycle-averaged pressure values ​​only reflects the pressure reduction caused by leakage, the thermal effect background is reduced, and the accuracy of leakage status determination is improved.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the step of applying active enhanced heat dissipation to the target cylinder during the first half of each preset modulation cycle specifically includes: at the beginning of the first half of the current modulation cycle, acquiring the surface temperature of the target cylinder to obtain the current cycle reference surface temperature; continuously applying cooling airflow to the target cylinder during the first half of the cycle, and acquiring the real-time cylinder surface temperature in real time; adjusting the flow rate of the cooling airflow in a closed-loop feedback manner based on the difference between the real-time cylinder surface temperature and the current cycle reference surface temperature to ensure that the active enhanced heat dissipation amplitude remains consistent in each modulation cycle; and stopping the application of cooling airflow at the end of the first half of the cycle to complete the active enhanced heat dissipation of the current modulation cycle.

[0009] In the above embodiment, the reference surface temperature is collected at the beginning of each first half cycle, and the cooling airflow is adjusted in a closed loop during the first half cycle. The active enhanced heat dissipation cooling amplitude remains consistent in each modulation cycle, and the contribution of the thermal effect to the average pressure value of each cycle is uniform and stable during the cycle, avoiding deviations in the linear trend fitting results of the pressure value sequence due to differences in the heat dissipation amplitude of each cycle.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the step of calculating the periodic average pressure value corresponding to each modulation cycle based on the bottle pressure data collected within each complete modulation cycle, using a preset modulation cycle as the integration window, specifically includes: dividing the bottle pressure data collected within each complete modulation cycle into a first-half-cycle pressure dataset and a second-half-cycle pressure dataset according to time sequence; removing transient pressure fluctuation data and retaining steady-state pressure data from the first-half-cycle pressure dataset and the second-half-cycle pressure dataset respectively, to obtain the first-half-cycle steady-state pressure data and the second-half-cycle steady-state pressure data respectively; and merging the first-half-cycle steady-state pressure data and the second-half-cycle steady-state pressure data with equal weight to obtain the periodic average pressure value corresponding to the current modulation cycle.

[0011] In the above embodiments, by removing transient fluctuation data from the pressure datasets of the first and second halves of the cycle and then merging them with equal weight, the interference of transient fluctuations near the start and stop times of cooling airflow on the mean is reduced. At the same time, the contribution weight of the thermal state to the cycle average pressure is the same in the first and second halves of the cycle, avoiding the introduction of systematic bias due to the imbalance of data volume and improving the calculation accuracy of the cycle average pressure value.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the step of performing linear trend fitting on a pressure value sequence consisting of periodic average pressure values ​​corresponding to multiple consecutive preset modulation periods to obtain a pressure deceleration rate specifically includes: performing an initial least squares linear fitting on the pressure value sequence to obtain an initial fitted line; calculating the residual of each periodic average pressure value in the pressure value sequence relative to the initial fitted line; assigning fitting weights to each periodic average pressure value according to the absolute value of the residual corresponding to each periodic average pressure value, performing a weighted least squares linear fitting on the pressure value sequence with the fitting weights to obtain a corrected fitted line; and using the absolute value of the slope of the corrected fitted line as the pressure deceleration rate.

[0013] In the above embodiments, by assigning weights to each data point based on the absolute value of the initial fitting residual, and then performing weighted least squares correction fitting, the influence weight of occasional outliers on the pressure deceleration rate estimation results is reduced, effectively reducing the interference of abnormal data points introduced by environmental fluctuations and sensor noise on the slope estimation, and improving the robustness of the pressure deceleration rate estimation results.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, after determining that the target cylinder has a leak when the pressure deceleration rate exceeds a preset leakage judgment threshold, the method further includes: shutting off the filling shut-off valve connecting the target cylinder and the filling pipeline, so that the target cylinder enters a gas path isolation state; in the gas path isolation state, keeping the filling shut-off valve closed, and continuing to perform multiple consecutive and complete preset modulation cycles on the target cylinder; continuously collecting the cylinder pressure, and calculating the average pressure value of the isolation cycle corresponding to each modulation cycle in the gas path isolation state using the preset modulation cycle as the integration window; performing linear trend fitting on the isolation pressure value sequence composed of the average pressure values ​​of the isolation cycles corresponding to multiple consecutive preset modulation cycles to obtain the isolation pressure deceleration rate; when the isolation pressure deceleration rate exceeds the preset leakage judgment threshold, determining that the leakage source is the cylinder body of the target cylinder, and outputting a cylinder body leakage alarm; when the isolation pressure deceleration rate does not exceed the preset leakage judgment threshold, determining that the leakage source is the filling interface between the filling pipeline and the target cylinder, and outputting a filling interface leakage alarm.

[0015] In the above embodiment, the target cylinder is placed into a gas path isolation state by shutting off the filling shut-off valve. Multiple modulation cycles are then executed, and the isolation pressure deceleration rate is calculated and compared with a threshold to determine the source of leakage. If the pressure deceleration rate still exceeds the threshold after isolation, it is determined to be a cylinder leak; if it drops below the threshold, it is determined to be a filling interface leak. The two types of leak sources can be distinguished and corresponding alarms can be output without disassembling the filling equipment.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after determining that the target cylinder has a leak when the pressure deceleration rate exceeds a preset leakage determination threshold, the method further includes: within multiple consecutive preset modulation cycles, taking the average of the cylinder pressure data collected in the first half cycle and the cylinder pressure data collected in the second half cycle of each preset modulation cycle, and arranging them in chronological order to obtain a pressure average sequence for the first half cycle and a pressure average sequence for the second half cycle; performing linear trend fitting on the pressure average sequence for the first half cycle and the pressure average sequence for the second half cycle respectively to obtain the pressure deceleration rate for the first half cycle. The pressure deceleration rate in the first half of the cycle and the pressure deceleration rate in the second half of the cycle are both considered. When both the pressure deceleration rate in the first half of the cycle and the pressure deceleration rate in the second half of the cycle exceed the preset leakage judgment threshold, it is confirmed that there is a real leak in the target cylinder. The pressure deceleration rates in the first half of the cycle and the pressure deceleration rates in the second half of the cycle are weighted and fused to obtain the fused pressure deceleration rate, which is then output as the final leakage rate evaluation value of the target cylinder. When only one of the pressure deceleration rates in the first half of the cycle and the pressure deceleration rate in the second half of the cycle exceeds the preset leakage judgment threshold, the preset modulation period is extended, and the sealing test of the target cylinder is re-executed.

[0017] In the above embodiments, by constructing equal pressure sequences for the first half-cycle and the second half-cycle respectively, linear trend fitting is performed on each sequence to obtain a dual-channel pressure deceleration rate, which is then used to cross-validate the leakage determination. If both channels exceed the threshold, a true leakage is confirmed; if only one channel exceeds the threshold, the modulation period is extended and the detection is repeated, reducing the probability of false alarms caused by residual thermal interference in single-channel determination.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, the step of weighted fusion of the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle to obtain the fusion pressure deceleration rate specifically includes: taking the arithmetic mean of the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle to obtain the initial fusion pressure deceleration rate; calculating the absolute value of the difference between the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle to obtain the thermal asymmetry index; determining the thermal bias correction amount based on the thermal asymmetry index, wherein the thermal bias correction amount is positively correlated with the thermal asymmetry index; and subtracting the thermal bias correction amount from the initial fusion pressure deceleration rate to obtain the fusion pressure deceleration rate.

[0019] In the above embodiment, the thermal asymmetry index is obtained by calculating the absolute value of the difference between the pressure deceleration rates of the two half-cycles, and the thermal bias correction amount is determined accordingly. The fusion pressure deceleration rate is obtained by subtracting this correction amount from the arithmetic mean. The thermal asymmetry index quantifies the bias magnitude introduced by the thermal state asymmetry between the two half-cycles. The thermal bias component in the corrected fusion pressure deceleration rate is partially reduced, making the output leakage rate assessment value closer to the actual gas leakage rate.

[0020] In a second aspect, embodiments of this application provide an automatic filling system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the automatic filling system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on an automatic filling system, cause the automatic filling system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on an automatic filling system, cause the automatic filling system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the automatic filling system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. Because the method of calculating the periodic average pressure value with a preset modulation period as the integration window and obtaining the pressure reduction rate by linear trend fitting of the series of consecutive periodic average pressure values, the net contribution of the thermal effect in each integration window tends to be consistent. The fitting slope only reflects the pressure reduction caused by leakage. This effectively solves the problem in the prior art that the thermal effect pressure drop and the leakage pressure drop are superimposed on the same reading and cannot be distinguished. Thus, the accurate determination of the sealing status of the steel cylinder after filling is realized.

[0026] 2. Because a closed-loop feedback adjustment method is adopted to regulate the cooling airflow based on the difference between the real-time surface temperature and the reference temperature in the first half of the cycle, the active heat dissipation enhancement amplitude remains consistent in each modulation cycle. The contribution of the thermal effect to the average pressure value in each cycle is uniform and stable during the cycle. This effectively solves the problem that the residual thermal effect in the pressure value sequence fluctuates due to inconsistent heat dissipation amplitude, which interferes with the accuracy of linear trend fitting. As a result, the stability of the pressure deceleration rate estimation result is improved.

[0027] 3. Because the method of merging the steady-state pressure data of the first and second halves of the cycle with equal weight after removing transient fluctuation data is adopted to calculate the periodic average pressure value, the transient noise interference is reduced. The contribution weight of the thermal state of the first and second halves of the cycle to the periodic average pressure is the same, which effectively solves the problem of systematic deviation in the periodic average pressure value caused by transient fluctuations and unbalanced data volume. This results in more accurate calculation of the periodic average pressure value and more reliable linear trend fitting data input. Attached Figure Description

[0028] Figure 1 This is a schematic flowchart of a high-safety fully automatic filling method for filling fluorine-nitrogen mixed gas in this application embodiment;

[0029] Figure 2 This is another schematic diagram of the process applied to the high-safety fully automatic filling method for fluorine-nitrogen mixed gas filling in the embodiments of this application;

[0030] Figure 3 This is a schematic diagram of the physical device structure of an automatic filling system in the embodiments of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0034] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0035] Fluorine-nitrogen mixtures (NF3 / N2) are indispensable specialty gases in high-end electronics industries such as semiconductor integrated circuit manufacturing, photovoltaic solar cell production, and flat panel display device fabrication. They are mainly used in critical processes such as CVD chamber cleaning, etching process atmosphere preparation, and equipment maintenance. Because fluorine-nitrogen mixtures are both highly toxic and corrosive—the permissible exposure concentration (TLV-TWA) of NF3 is only 10 ppm, and the gas itself has a strong oxidizing and corrosive effect on common metal materials—its filling and production process has always been one of the highest safety risks and most technically demanding processes in the specialty gas industry.

[0036] In typical industrial applications, gas production companies need to fill fluorine-nitrogen mixed gases from storage tanks into steel cylinders (including single cylinders and containerized units) of different specifications according to customer order requirements, ensuring specified purity and target pressure, before delivering them to downstream gas-consuming units such as semiconductor wafer fabs and photovoltaic module factories. The entire filling process involves multiple sequential steps, including pipeline connection, airtightness verification, pipeline replacement, vacuum pretreatment, pressurization filling, and post-filling pipeline inerting. Errors in any step can lead to serious consequences such as toxic gas leakage, overpressure filling, or substandard product purity.

[0037] In the application scenario described in this application, the filling workshop is equipped with multiple independent filling stations, each with a piping system connected to the filling panel. After the operator connects the cylinder or container to be filled to the corresponding station, they leave the operating area. The entire subsequent filling process is then taken over and executed by the high-safety fully automated filling system proposed in this application. The system automatically completes a closed-loop process sequentially, including leak detection and pressure holding, purging and replacement, vacuuming, cyclic replacement, pressurized filling, and pre-disassembly piping purging. No operator intervention is required throughout the process, and operators are physically isolated from hazardous filling areas. Regarding filling accuracy control, the PLC system reads the pressure feedback value from the high-precision pressure sensor in real time and compares it with the operator's preset target filling pressure within milliseconds. The system automatically closes the main material valve and cylinder valve and stops the compressor the instant the pressure reaches the target value, eliminating the risk of overfilling. In terms of safety protection, the signals of toxic gas detectors installed in key areas such as the filling panel, compressor and buffer tank are connected to the GDS system in real time. Once a leak is detected, the safety instrument system controlled by PLC will trigger a series of automatic response actions within milliseconds, such as closing the emergency shut-off valve, starting nitrogen purging and emergency ventilation interlock, suppressing the risk of the accident from the outset.

[0038] In actual filling production, this application scenario also places extremely high demands on product purity—downstream semiconductor wafer manufacturing customers require the purity of the filled products to be no less than 6N grade (99.9999%). This necessitates that the filling system use 316LEP (electropolished) stainless steel pipelines with VCR metal-sealed connections throughout the entire process. All high-purity pipelines must undergo automated track welding in a Class 100 cleanroom and pass helium leak testing, particle size testing, and water and oxygen content testing before being put into use. This ensures that no contaminants such as metal particles, moisture, or oxygen are introduced during the filling process due to pipeline materials or connection processes, guaranteeing stable and compliant product purity. The filling system and method proposed in this application were developed specifically to address the simultaneous high safety, high purity, and high efficiency requirements of the aforementioned application scenario. Through the organic integration of a fully automated control process, a multi-layered active safety protection architecture, and ultra-high purity assurance processes, a complete engineering implementation plan is provided for the filling production of fluorine-nitrogen mixed gases.

[0039] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a high-safety fully automatic filling method for filling fluorine-nitrogen mixed gases in this application embodiment.

[0040] S101. Fill the target cylinder with a fluorine-nitrogen mixture and collect the filling quality data in real time. Stop filling when the collected filling quality data reaches the target filling amount.

[0041] The target cylinder refers to an industrial high-pressure gas cylinder to be filled with a fluorine-nitrogen mixed gas, which typically has a specific pressure resistance rating and a corrosion-resistant lining; the fluorine-nitrogen mixed gas refers to a mixed process gas prepared with fluorine (F2) as the active component and nitrogen (N2) as the dilution carrier in a certain volume or mass ratio; the filling mass refers to the mass of gas filled into the target cylinder during the filling process, which is obtained by continuously measuring the difference between the total mass of the cylinder and the initial mass; the target filling quantity refers to the upper limit value of the gas mass to be filled in this filling, which is preset and written into the control parameters and is used as the basis for determining whether to stop filling; the filling mass data refers to the sequence of differences between the instantaneous mass reading of the cylinder and the initial mass continuously output by the weighing sensor according to the sampling period during the filling process.

[0042] Specifically, before filling begins, the automatic filling system records the initial mass of the target cylinder and loads the target filling quantity into the control logic. During filling, the gas source control valve opens, and the fluorine-nitrogen mixture continuously fills the target cylinder along the filling pipeline. Weighing sensors installed at the filling station continuously collect the real-time total mass of the target cylinder at a fixed sampling frequency. Subtracting the initial mass value recorded before filling from the total mass value collected each time yields the current filling mass data, reflecting the actual cumulative mass of the filled gas. At each sampling moment, the control system compares the filling mass data with the target filling quantity. When the filling mass data reaches or exceeds the target filling quantity, the control system sends a closing command to the gas source control valve, terminating gas filling and ending the filling phase.

[0043] In some embodiments, real-time acquisition of filling quality and filling termination control can be achieved in several ways: Optionally, a direct lower-level control method can be adopted: First, the weighing sensor converts the cylinder mass signal into an analog electrical signal, which is then sampled by an analog-to-digital converter and outputs a digital mass value to a programmable logic controller (PLC); the PLC calculates the current filling quality data by subtracting the current digital mass value from the initial mass value; the PLC compares the filling quality data with a preset target filling amount, and when the filling quality data reaches the target filling amount, the PLC directly outputs a digital control signal to drive the filling valve actuator to close, thus terminating the filling process; Optionally, a coordinated upper-level and lower-level control method can be adopted: First, the weighing sensor periodically uploads the collected mass data to the industrial control computer via a fieldbus (such as Modbus or PROFIBUS protocol); the industrial control computer runs the filling management software, calculates the filling quality data in real time and displays it; when the filling quality data reaches the target filling amount, the industrial control computer sends a stop filling command to the lower-level computer via a communication interface; after receiving the command, the lower-level computer drives the filling shut-off valve to close, thus achieving termination control of the filling stage. It is understandable that other quality acquisition and termination control methods can also be used to achieve automatic stopping of the filling process, such as filling quality confirmation methods based on multi-sensor redundancy verification, which are not limited here.

[0044] S102. Apply active enhanced heat dissipation to the target cylinder during the first half of each preset modulation cycle, and stop active heat dissipation to the target cylinder during the second half of each preset modulation cycle, and continuously alternately execute multiple complete preset modulation cycles.

[0045] The preset modulation cycle refers to the time period pre-configured during the sealing test stage after filling, which specifies the time rhythm of alternating active cooling enhancement and cessation of active cooling enhancement. The first half of the cycle refers to the time interval in the early stage of each preset modulation cycle, used to perform active cooling enhancement operation. The second half of the cycle refers to the time interval in the late stage of each preset modulation cycle, where external cooling intervention is stopped and the target cylinder is in a state of natural heat dissipation. Active cooling enhancement refers to the process of actively accelerating the dissipation of heat from the cylinder to the outside by continuously applying cooling airflow and other external means to the outer surface of the target cylinder. Natural heat dissipation refers to the process of the target cylinder dissipating heat through natural convection and thermal radiation between itself and the environment without external cooling intervention.

[0046] Specifically, after the filling stage, the automatic filling system enters the sealing test stage, alternately performing active enhanced cooling and deactivation of active cooling on the target cylinder according to a preset modulation cycle. In the first half of each preset modulation cycle, the control system activates the cooling device, applying cooling airflow to the outer surface of the target cylinder to forcibly accelerate heat dissipation from the cylinder surface, causing a significant drop in the cylinder's temperature during the first half of the cycle. The internal gas temperature and pressure then decrease accordingly. In the second half of each preset modulation cycle, the control system shuts off the cooling device and stops applying cooling airflow. The target cylinder relies solely on natural convection and radiation to exchange heat with the outside environment, resulting in a slower heat dissipation rate and a more stable temperature. After one complete modulation cycle, the control system immediately starts timing for the next modulation cycle, activating the cooling device again in the first half of the cycle. This alternating process continues until the prescribed number of complete modulation cycles is completed.

[0047] In some embodiments, the periodic alternation of active cooling enhancement and cessation of active cooling can be achieved in various ways: Optionally, a solenoid valve timed switch control method is adopted: First, a cooling airflow nozzle is installed at the filling station, and a solenoid valve is installed on the nozzle inlet pipe. The control terminal of the solenoid valve is connected to a programmable logic controller (PLC). The controller configures timing logic according to the duration of the first and second halves of a preset modulation period. At the beginning of each first half of the period, it automatically outputs an open signal to the solenoid valve, and the cooling airflow continuously blows from the nozzle to the surface of the target cylinder at a set flow rate. At the end of the first half of the period, it outputs a close signal to the solenoid valve, the cooling airflow is interrupted, and the target cylinder enters... The system operates in a half-cycle natural cooling state. After the second half-cycle ends, the timing logic automatically enters the next modulation cycle and repeats the above process. Optionally, a temperature feedback-triggered cycle switching method can be used: First, a temperature sensor is installed on the surface of the target cylinder. At the beginning of the first half-cycle, the cooling airflow is activated and the cylinder surface temperature is continuously monitored. When the cylinder surface temperature drops to the preset target cooling amount, the cooling task of the first half-cycle is deemed complete, triggering the cooling device to shut down and entering the second half-cycle. During the second half-cycle, the cylinder temperature is continuously monitored. When the temperature rises back to the set proportion of the temperature at the end of the first half-cycle, the second half-cycle is deemed to have ended, and the first half-cycle of the next modulation cycle begins. It is understood that other timing or feedback control methods can also be used to achieve the alternating switching between active cooling and de-cooling during the modulation cycle, such as an adaptive cooling power adjustment method based on heat flow sensor feedback, which is not limited here.

[0048] In some embodiments, the step of applying active enhanced heat dissipation to the target cylinder during the first half of each preset modulation cycle specifically includes: at the beginning of the first half of the current modulation cycle, acquiring the cylinder surface temperature of the target cylinder to obtain the current cycle reference surface temperature; continuously applying cooling airflow to the target cylinder during the first half of the cycle and acquiring the real-time cylinder surface temperature in real time; adjusting the flow rate of the cooling airflow in a closed-loop feedback manner based on the difference between the real-time cylinder surface temperature and the current cycle reference surface temperature to ensure that the active enhanced heat dissipation amplitude remains consistent in each modulation cycle; and stopping the application of cooling airflow at the end of the first half of the cycle to complete the active enhanced heat dissipation of the current modulation cycle.

[0049] The current cycle reference surface temperature is the temperature value obtained by a single sampling of the bottle surface at the beginning of the first half of a modulation cycle. It serves as the reference starting point for cooling control in this cycle. For example, if 28.5°C is sampled at the beginning of the first half of the third modulation cycle, then T_base_3 = 28.5°C. The cooling airflow is compressed cold air continuously sprayed onto the outer wall of the cylinder, driven by a flow control valve. Forced cooling is achieved by removing heat from the bottle surface, and its flow rate is adjusted in real time by the controller. The real-time bottle surface temperature is the time-series value T_real(t) of the bottle's outer wall temperature, continuously sampled at a fixed sampling frequency during the first half of the cycle, reflecting the current cooling effect. Closed-loop feedback regulation uses the difference between T_real(t) and the current cycle's set target temperature as a deviation signal to drive the flow control valve to adjust the cooling airflow in real time, causing the actual cooling effect to converge towards the target value. For example, the flow rate is increased when T_real(t) is higher than the target temperature and decreased when it is lower. The active enhancement of heat dissipation amplitude is the maximum temperature drop ΔT actually achieved on the bottle surface in the first half of the cycle, starting from the current cycle reference surface temperature. The control objective is to keep this value consistent in each modulation cycle.

[0050] The complete execution process of active enhanced heat dissipation consists of three stages. In the first stage, at the beginning of the first half of the cycle, a single temperature reading is collected by a contact temperature sensor (such as a platinum resistance RTD or a type K thermocouple) installed at a fixed measuring point on the bottle surface. This reading is recorded as the current cycle's reference surface temperature T_base_i. At the same time, the target cooling temperature for this cycle is calculated based on the preset target temperature drop ΔT_target: T_sp_i = T_base_i - ΔT_target, where ΔT_target is a fixed value, such as 5°C, that is solidified and written into the control parameters through calibration experiments. This ensures that even if the initial bottle temperature is different in different modulation cycles, the temperature drop of the bottle caused by active heat dissipation is the same fixed value each time. In the second stage, during the first half of the cycle, the controller acquires T_real(t) in real time at a fixed frequency (e.g., 10 Hz), calculates the deviation signal e(t) = T_real(t) - T_sp_i, and calculates the flow control command based on the PI control law: u(t) = Kp × e(t) + Ki × ∫e(t)dt, where Kp is the proportional gain and Ki is the integral gain, which are tuned through response experiments of applying a step cooling load to the test bench. The controller converts u(t) into a 0~100% opening command and outputs it to the proportional flow control valve, driving the valve to adjust the cooling airflow Q(t), so that T_real(t) converges to T_sp_i and remains stable. For specific numerical examples: if T_base_i = 28.5°C and ΔT_target = 5°C, then T_sp_i = 23.5°C. In the initial stage of the first half-cycle, T_real(t) drops rapidly from 28.5°C. When T_real(t) = 25°C, the deviation e = 1.5°C, and the controller outputs a larger flow command to accelerate cooling. When T_real(t) converges to 23.5°C, the deviation approaches zero, and the controller maintains the flow rate at the smaller value required to maintain a stable temperature drop. In the third stage, at the end of the first half-cycle, the controller closes the flow control valve to zero opening, the cooling airflow stops, and the cylinder enters the natural thermal recovery stage of the second half-cycle. Since each modulation cycle calculates the target temperature with its own T_base_i as a reference, the heat dissipation amplitude ΔT_target remains consistent across cycles, reducing the impact of initial temperature drift during the cycle on the active heat dissipation intensity and ensuring the comparability of the thermal modulation effects of each modulation cycle.

[0051] S103. The internal pressure of the target cylinder is continuously collected. Using a preset modulation period as the integration window, the average pressure value of each modulation period is calculated based on the internal pressure data collected in each complete modulation period.

[0052] Here, the internal pressure refers to the pressure of the gas inside the target cylinder, which is continuously collected by a pressure sensor installed on the filling pipeline or cylinder valve port. Its value reflects the thermodynamic state of the gas inside the target cylinder. The integral window refers to the time span selected when performing statistical calculations on time-series data. In this step, a preset modulation period is used as the time window for statistical processing of the internal pressure data. The periodic average pressure value refers to the representative pressure statistic calculated after extracting steady-state data and merging equally weighted data from the internal pressure data collected within a complete modulation period. It is used to describe the pressure level of the target cylinder within that modulation period. The first half-cycle pressure dataset refers to the set of all internal pressure sampling points collected during the first half-cycle period. The second half-cycle pressure dataset refers to the set of all internal pressure sampling points collected during the second half-cycle period. The steady-state pressure data refers to the pressure sampling points that remain in a stable state after removing transient fluctuation sampling points generated during the start-stop transition of cooling airflow.

[0053] Specifically, after the sealing test begins, the pressure sensor continuously collects the internal pressure of the target cylinder and outputs a time-series pressure data stream at a fixed sampling frequency. Each complete preset modulation cycle is used as an integration window, and all internal pressure sampling points collected within this window are divided into a first-half cycle pressure dataset and a second-half cycle pressure dataset. Since the cooling airflow starts at the beginning of the first half cycle and stops at the beginning of the second half cycle, the disturbances experienced by the cylinder near the start and stop times cause temporary non-steady-state fluctuations in internal pressure. This data segment cannot represent the stable pressure level within that period. Therefore, transient fluctuation sampling points corresponding to the start-stop transition period are identified and removed from both the first-half and second-half cycle pressure datasets, while the steady-state pressure sampling points within their respective stable periods are retained, resulting in first-half cycle steady-state pressure data and second-half cycle steady-state pressure data, respectively. The first-half cycle steady-state pressure data and the second-half cycle steady-state pressure data are then merged with equal weights. That is, after assigning equal weights to both sets of steady-state data, the arithmetic mean of all sampling points in the merged dataset is taken to obtain the periodic average pressure value corresponding to the current modulation cycle.

[0054] In some embodiments, the removal of transient fluctuation data and the calculation of the average pressure value of each modulation cycle can be achieved in various ways: Optionally, a fixed transition time truncation method can be adopted: First, based on the airflow start-stop response characteristics of the cooling device, pre-calibration is performed to determine the steady-state establishment time for the first half-cycle and the second half-cycle, for example, a fixed number of seconds are required after starting or stopping the cooling airflow before entering steady state; for the pressure dataset of the first half-cycle of each modulation cycle, the sampling points within the transition time corresponding to the initial segment are deleted, and the sampling points of the subsequent steady-state segment are retained; similarly, the sampling points within the transition time corresponding to the initial segment are deleted for the pressure dataset of the second half-cycle. The system first selects and retains the steady-state sampling points. Then, it merges the two sets of steady-state sampling points with equal weight and takes their arithmetic mean to obtain the periodic average pressure value. Optionally, a dynamic variance threshold steady-state identification method can be used: first, the local pressure variance of the pressure datasets for the first and second halves of the cycle is calculated using a sliding time window. Sampling points with local variance exceeding a preset steady-state variance threshold are marked as transient fluctuation data and removed, while sampling points with local variance below the threshold are retained as steady-state pressure data. The steady-state pressure data samples for the first and second halves of the cycle are then weighted and merged. The arithmetic mean of the merged samples is taken to obtain the periodic average pressure value for the current modulation cycle. It is understood that other steady-state data identification and mean calculation methods can also be used to calculate the periodic average pressure value, such as steady-state interval extraction methods based on autocorrelation coefficients; this is not limited here.

[0055] In some embodiments, the step of calculating the periodic average pressure value corresponding to each modulation cycle based on the bottle pressure data collected within each complete modulation cycle, using a preset modulation cycle as the integration window, specifically includes: dividing the bottle pressure data collected within each complete modulation cycle into a first-half-cycle pressure dataset and a second-half-cycle pressure dataset according to time sequence; removing transient pressure fluctuation data and retaining steady-state pressure data from the first-half-cycle pressure dataset and the second-half-cycle pressure dataset respectively, thereby obtaining the first-half-cycle steady-state pressure data and the second-half-cycle steady-state pressure data respectively; and merging the first-half-cycle steady-state pressure data and the second-half-cycle steady-state pressure data with equal weight to obtain the periodic average pressure value corresponding to the current modulation cycle.

[0056] An integral window refers to the time range for statistical calculation, defined as the duration of a complete modulation cycle. All pressure sampling points collected within this range participate in the calculation of the average pressure value for the current cycle, without mixing with data from adjacent cycles. The first-half cycle pressure dataset and the second-half cycle pressure dataset are obtained by dividing all pressure sampling points within a complete modulation cycle into two time-series sets. The first half cycle corresponds to the active cooling period, and the second half cycle corresponds to the natural recovery period, each forming an independent pressure time-series data set. Transient pressure fluctuation data refers to pressure sampling points at the beginning of each half cycle when the gas pressure has not yet reached a stable state due to rapid temperature changes. During this stage, the pressure change rate is large and cannot represent the steady-state gas pressure. For example, the data segment showing a sharp drop in pressure in the first few seconds of the first half cycle at the beginning of cooling. The first-half cycle steady-state pressure data and the second-half cycle steady-state pressure data are pressure sampling points retained after removing transient fluctuation data from their respective datasets. During this stage, the temperature tends to stabilize, and the pressure change rate drops to a lower level. Equal-weighted merging combines the steady-state pressure data from the first half of the cycle with the steady-state pressure data from the second half of the cycle into a single dataset. Each sampling point contributes the same weight to participate in the mean calculation, ultimately yielding the cycle-average pressure value corresponding to the current modulation cycle.

[0057] The calculation process for the periodic average pressure value consists of four steps. The first step is to divide the dataset according to the time sequence: Let the duration of the current modulation period be T and the sampling frequency be f_s. Then, a total of N_total = T × f_s pressure sampling points are collected in one complete period. Taking the duration of the first half period T / 2 as the boundary, the first N_total / 2 sampling points are assigned to the first half period pressure dataset D_f, and the last N_total / 2 sampling points are assigned to the second half period pressure dataset D_b. The second step is to remove transient fluctuation data within each half-cycle: Perform the following operations independently on D_f and D_b: calculate the pressure change rate between adjacent sampling points using the sliding difference method, r(t) = (P(t) - P(t-1)) / (1 / f_s). Starting from the first sampling point of each half-cycle, mark the continuous data segment where |r(t)| continuously exceeds the preset transient judgment threshold r_thresh as a transient segment and remove it. When |r(t)| does not exceed r_thresh for M consecutive sampling points, the system is considered to have entered a steady state. From the (M+1)th point that meets the condition, the subsequent data is retained as steady-state pressure data. r_thresh and M are determined by experimental calibration of the pressure response of an empty bottle under the same modulation parameters, for example, r_thresh = 0.05 kPa / s and M = 5. This operation applies to both D_f and D_b, resulting in the first half-cycle steady-state pressure dataset S_f = {P_f,1, P_f,2, …, P_f,nf} and the second half-cycle steady-state pressure dataset S_b = {P_b,1,P_b,2, …, P_b,nb}, where nf and nb are the number of sampling points in the two steady-state datasets, and they do not need to be equal. The third step is equal-weight merging: all sampling points from S_f and S_b are combined into a single dataset S = S_f ∪ S_b, with a total of nf + nb points, each with a weight of 1. The fourth step is calculating the periodic average pressure value: P_avg_i = (ΣS_f + ΣS_b) / (nf + nb) = (P_f,1 + … + P_f,nf + P_b,1 + … + P_b,nb) / (nf + nb), which is used as the periodic average pressure value for the i-th modulation cycle. For example, if the modulation period T = 60 s and f_s = 1 Hz, then each half-cycle has 30 original data points; after transient rejection, the first half-cycle retains 24 steady-state points nf, with an average of 502.3 kPa, and the second half-cycle retains 26 points nb, with an average of 503.1 kPa. Therefore, P_avg_i = (502.3×24 + 503.1×26) / (24 + 26) = 502.73 kPa.

[0058] S104. Perform linear trend fitting on the pressure value sequence consisting of the periodic average pressure values ​​corresponding to multiple consecutive preset modulation periods to obtain the pressure deceleration rate.

[0059] Among them, the pressure value sequence refers to a time-series data sequence composed of the average pressure values ​​of each of the multiple consecutive preset modulation cycles arranged in the order of modulation cycle numbers. Each element in the sequence corresponds to a representative pressure statistic of a modulation cycle. Linear trend fitting refers to mathematically fitting the pressure value sequence using a linear function model to solve the optimal linear equation describing the trend of the sequence with the modulation cycle number. The initial fitting line refers to the linear equation obtained after performing an initial least squares linear fitting with equal weights for each data point. The residual refers to the difference between the observed value of the average pressure value of each cycle in the pressure value sequence and the predicted value of the initial fitting line at the corresponding horizontal coordinate position. The fitting weight refers to the weight value assigned to each data point according to the absolute value of the residual of each data point for weighted least squares fitting. The corrected fitting line refers to the corrected linear equation obtained after performing a weighted least squares linear fitting on the pressure value sequence with the fitting weights of each data point. The pressure deceleration rate refers to the absolute value of the slope of the corrected fitting line, which characterizes the rate at which the pressure inside the target gas cylinder decreases as the modulation cycle progresses.

[0060] Specifically, after the periodic average pressure values ​​of multiple consecutive modulation cycles have accumulated, they are arranged according to the modulation cycle number to form a pressure value sequence. A linear fitting model is established with the modulation cycle number as the independent variable and the periodic average pressure value as the dependent variable. The first step is to perform equal-weighted least-squares linear fitting on the pressure value sequence: by minimizing the sum of squares of the vertical distances from each data point to the fitted line, the slope and intercept are solved to obtain the initial fitted line. The second step is to calculate the residuals of each periodic average pressure value in the pressure value sequence relative to the predicted values ​​corresponding to the initial fitted line. The absolute value of each residual is taken. Data points with larger absolute residual values ​​are considered outliers that deviate significantly from the normal linear trend, and their contribution to the fitting result should be reduced; data points with smaller absolute residual values ​​are considered normal data points with higher confidence and should be given a larger contribution weight. The third step involves calculating fitting weights for each data point based on the magnitude of the absolute residuals using a predetermined inverse proportional function (e.g., using the absolute residual plus the reciprocal of a small positive number as the weight). Weighted least-squares linear fitting is then performed on the pressure value sequence using these fitting weights. By minimizing the sum of squared weighted longitudinal distances between the data points, the slope and intercept of the corrected fitting line are obtained. The absolute value of the slope of the corrected fitting line is then taken as the pressure deceleration rate output in this step.

[0061] In some embodiments, linear trend fitting of the pressure value sequence and calculation of the pressure deceleration rate can be achieved in various ways: Optionally, a two-step iterative weighted least squares method is adopted: First, an initial least squares linear fitting is performed on the pressure value sequence with equal weights for each data point, and the slope and intercept of the initial fitted line are calculated; then, the absolute value of the residual of the average pressure value of each period relative to the initial fitted line is calculated, and the fitting weight of each data point is calculated according to the reciprocal of the absolute value of the residual (with a small positive number added to prevent division by zero); a second weighted least squares linear fitting is performed on the pressure value sequence with this fitting weight to solve for the corrected fitted line. The slope of the pressure value sequence is used, and its absolute value is taken as the pressure deceleration rate output. Optionally, a residual absolute value threshold truncation weighting method can be used: First, an equal-weighted initial least squares linear fit is performed on the pressure value sequence to obtain an initial fitted line; the absolute value of the residual of each period's average pressure value relative to the initial fitted line is calculated, and a residual absolute value truncation threshold is set. The weights of data points whose residual absolute values ​​exceed the threshold are reset to zero, and the weights of data points whose residual absolute values ​​do not exceed the truncation threshold are reset to one; the remaining effective data points are then subjected to least squares linear fit using the truncated weights to solve for the slope of the corrected fitted line, and its absolute value is taken as the pressure deceleration rate. It is understood that other robust linear regression methods can also be used to achieve trend fitting of the pressure value sequence and estimation of the pressure deceleration rate, such as robust linear regression methods based on the Huber loss function; this is not limited here.

[0062] In some embodiments, the step of performing linear trend fitting on a pressure value sequence consisting of periodic average pressure values ​​corresponding to multiple consecutive preset modulation periods to obtain a pressure deceleration rate specifically includes: performing an initial least squares linear fitting on the pressure value sequence to obtain an initial fitted line; calculating the residual of each periodic average pressure value in the pressure value sequence relative to the initial fitted line; assigning fitting weights to each periodic average pressure value based on the absolute value of the residual corresponding to each periodic average pressure value; performing a weighted least squares linear fitting on the pressure value sequence with the fitting weights to obtain a corrected fitted line; and using the absolute value of the slope of the corrected fitted line as the pressure deceleration rate.

[0063] The pressure value sequence is a sequence {P_1, P_2, …, P_N} consisting of the periodic average pressure values ​​of N consecutive modulation cycles arranged chronologically by cycle numbers i = 1, 2, …, N, for example, {505.2, 504.8, 504.3, 503.9, 503.5} kPa. Initial least squares linear fitting involves performing a standard linear regression on all N data points with equal weights, using modulation cycle number i as the independent variable and P_i as the dependent variable, to find the initial slope k0 and initial intercept b0 that minimize the sum of squared residuals. The initial fitted line is the linear equation y = k0×i + b0, determined by k0 and b0, representing an initial estimate of the overall trend of the pressure value sequence. The residual is the difference between the average pressure value P_i of the i-th period and the predicted value at point i on the initial fitted line, e_i = P_i - (k0×i + b0), reflecting the degree to which that point deviates from the overall trend. For example, when P_3 = 504.3 and the predicted value is 504.5, the residual is -0.2 kPa. The fitting weight w_i is a positive real number determined inversely proportional to the absolute value of the residual |e_i| of each data point. The larger |e_i| is, the smaller w_i is, and the weaker the contribution of that data point to the weighted fitting, thus reducing the impact of outlier points on the final slope. Weighted least squares linear fitting is the process of performing linear regression with w_i as the weight of each data point to find the corrected slope k and corrected intercept b that minimize the sum of squared weighted residuals. The corrected fitted line is the linear equation y = k×i + b determined by k and b, which is the final estimate of the trend of the pressure value sequence and is less affected by outliers. The pressure deceleration rate is the absolute value of the slope k of the corrected fitted line, |k|, in kPa / cycle, representing the average decrease in internal pressure of the cylinder per modulation cycle.

[0064] The calculation of the pressure deceleration rate consists of five steps, using a pressure value sequence {P_1, P_2, …, P_N} over N periods as input. The first step is to perform an initial equal-weighted least-squares linear fit: with i as the independent variable and P_i as the dependent variable, a linear model P_i = k0×i + b0 is established. k0 and b0 are solved by minimizing the residual sum of squares Σ(P_i - k0×i - b0)², with analytical solutions k0 = [N×Σ(i×P_i) - Σi×ΣP_i] / [N×Σi² - (Σi)²], b0 = (ΣP_i - k0×Σi) / N, where all summations are accumulations of i from 1 to N. The second step is to calculate the residuals for each data point: For each i, calculate e_i = P_i - (k0×i + b0), resulting in N residual values. The absolute value of the residual |e_i| reflects the degree to which the point deviates from the initial trend line. Points with larger absolute values ​​are outliers affected by external interference or short-term fluctuations during pressure acquisition. The third step is to calculate the fitting weights for each data point based on the absolute values ​​of the residuals: w_i = 1 / (|e_i| + ε), where ε is a small positive number to prevent division by zero, and its value is on the order of 0.01% of the pressure sensor's range. For example, when the range is 0~1000 kPa, ε = 0.001 kPa. The larger |e_i| is, the smaller w_i is, indicating that the deviation of the point from the overall trend is greater, and the weaker its contribution to the slope in the subsequent weighted fitting, thus effectively suppressing the interference of outliers on the final rate estimation. The fourth step is to perform weighted least squares linear fitting with weights w_i: minimize the weighted sum of squared residuals Σ w_i×(P_i - k×i - b)², let W = Σw_i, Wx = Σ(w_i×i), Wy = Σ(w_i×P_i), Wxx = Σ(w_i×i²), Wxy = Σ(w_i×i×P_i), then the correction slope k = (W×Wxy - Wx×Wy) / (W×Wxx - Wx²), and the correction intercept b = (Wy - k×Wx) / W. The corrected fitting line is determined by k and b. The fifth step is to take the pressure deceleration rate R = |k| and output it in kPa / cycle. For a specific numerical example: Suppose N = 5, and the pressure value sequence is {505.2, 504.8, 504.1, 503.9, 503.2} kPa. The initial equal-weighted fitting yields k0 ≈ -0.50 kPa / cycle. Calculate the residuals at each point. Assuming P_3 = 504.1 deviates significantly from the fitted value, its |e_3| is larger, and therefore w_3 is smaller. After weighted fitting, the corrected slope k ≈ -0.48 kPa / cycle is obtained. The large deviation of P_3 on the slope is effectively suppressed, and the final output pressure deceleration rate R = |−0.48| = 0.48 kPa / cycle.

[0065] S105. When the pressure deceleration rate exceeds the preset leakage judgment threshold, the target cylinder is judged to have a leak.

[0066] Among them, the pressure deceleration rate refers to the absolute value of the slope of the corrected fitting line obtained by linear trend fitting of the periodic average pressure value sequence of multiple consecutive modulation cycles, which characterizes the rate at which the pressure inside the target cylinder continuously decreases as the modulation cycle progresses; the preset leakage judgment threshold refers to the pressure deceleration rate boundary value that is pre-calibrated and written into the control parameters according to the fluorine filling safety standards and the allowable pressure fluctuation range under normal sealing conditions of the target cylinder, used to distinguish between normal sealing conditions and abnormal conditions with leakage; leakage refers to the phenomenon that the fluorine-nitrogen mixture inside the target cylinder escapes to the outside through cylinder defects, valve sealing failure, or filling interface gaps; judgment refers to the conclusive assessment of the current sealing status of the target cylinder by the control system based on the comparison result of the pressure deceleration rate and the preset leakage judgment threshold.

[0067] Specifically, after obtaining the pressure deceleration rate calculation result, the control system compares this value with a preset leakage judgment threshold stored in the control parameters. The preset leakage judgment threshold is determined in advance based on the target cylinder's specifications, operating pressure range, and fluorine filling safety regulations, and is calibrated before the system is put into use. When the pressure deceleration rate is greater than the preset leakage judgment threshold, it indicates that the rate of pressure decrease inside the target cylinder as the modulation cycle progresses has exceeded the allowable range caused by residual heat effects under normal sealing conditions. The excess is judged to be caused by gas leakage. Based on this, the control system determines that the target cylinder has a leak and triggers subsequent alarms and handling procedures, prohibiting the target cylinder from being transferred to the normal circulation process.

[0068] In some embodiments, the comparison between the pressure deceleration rate and the preset leakage judgment threshold, and the output of the leakage judgment conclusion, can be achieved in several ways: Optionally, a single-time threshold hard judgment method is adopted: First, the pre-calibrated leakage judgment threshold is stored in the control system parameter table; in the sealing test process, after the control system calculates the pressure deceleration rate, it directly compares it with the threshold in the parameter table for a single value; if the pressure deceleration rate is greater than the threshold, a leakage judgment conclusion is immediately generated, the current pressure deceleration rate value, the time of exceeding the threshold, and the cylinder number are recorded, and a leakage alarm signal is sent to the alarm module to trigger the subsequent processing process; Optionally, a continuous multiple threshold cumulative confirmation method is adopted: First, in the sealing test process, the pressure deceleration rate of the pressure value subsequence composed of the latest several modulation cycles is continuously calculated in a rolling manner, and each calculation result is compared with the threshold; a continuous threshold counter is set, and when the continuous calculation results exceed the threshold, the cumulative count reaches the preset continuous threshold number threshold before the final leakage judgment conclusion is output. A single threshold exceedance only increments the counter without outputting the final judgment, thereby avoiding misjudgment caused by occasional interference. It is understandable that other comparison logics can be used to determine the leakage status, such as the probability over-threshold determination method based on sliding window statistical testing, which is not limited here.

[0069] S106. When the pressure deceleration rate does not exceed the preset leakage judgment threshold, the target cylinder is judged to be sealed and the filling of the fluorine-nitrogen mixed gas is completed.

[0070] Among them, the pressure deceleration rate refers to the absolute value of the slope of the corrected fitting line obtained by linearly fitting the sequence of periodic average pressure values ​​of multiple consecutive modulation cycles; the preset leakage judgment threshold refers to the pre-calibrated boundary value of the pressure deceleration rate used to distinguish between normal sealing and leakage states; sealing qualified means that after the target cylinder is filled, the sealing test process determines that the pressure deceleration rate of the fluorine-nitrogen mixed gas inside it is within the allowable range corresponding to the normal sealing state, and it has the conditions for safe storage, transportation or use; completing the filling means that after the target cylinder is filled with fluorine-nitrogen mixed gas and the sealing test is qualified, the entire filling and testing process is completed, including filling record archiving, filling station reset and target cylinder transfer to subsequent processes.

[0071] Specifically, the control system compares the calculated pressure deceleration rate with a preset leakage threshold. When the pressure deceleration rate does not exceed the preset leakage threshold, it indicates that the rate of pressure decrease in the target cylinder as the modulation cycle progresses does not exceed the allowable range corresponding to residual thermal effects under normal sealing conditions, thus eliminating the possibility of continuous gas leakage. Based on this, the control system determines that the target cylinder is sealed correctly and outputs a sealing qualification test conclusion to the system's host computer or operating terminal. Simultaneously, it writes key information such as the filling quality, filling time, cylinder pressure data for each modulation cycle, pressure deceleration rate calculation results, and sealing qualification conclusion into the system database, completing the process data archiving for this filling operation. After the sealing test process is completed, the filling interface between the filling pipeline and the target cylinder is disconnected, the filling station returns to standby mode, and the target cylinder enters the subsequent labeling, qualified product warehousing, or transfer process, completing the filling of the fluorine-nitrogen mixed gas.

[0072] In some embodiments, the filling process after the seal qualification is determined can be completed in multiple ways: Optionally, an automatic unlocking and data archiving linkage method can be adopted: First, while outputting the seal qualification conclusion, the control system sends an automatic unlocking command to the mechanical locking mechanism of the filling station, allowing the operator to pull the filling connector out of the target cylinder valve; the system simultaneously writes the data of the entire filling process (including the filling quality curve, pressure data of each modulation cycle, pressure deceleration rate value, and seal qualification conclusion) into the database and automatically generates an electronic record sheet for this filling; the operator verifies the information on the record sheet. After verification, the target cylinder is transferred to the qualified product area, and the filling station is reset. Optionally, a label printing and batch tracking method can be used: First, after the control system outputs a seal qualification judgment, it sends a printing instruction to the label printer to print a barcode or QR code label containing information such as the filling date, gas composition ratio, filling quality, seal inspection qualification mark, and batch number. The operator affixes the label to the designated location on the target cylinder, and the control system simultaneously updates the cylinder's filling status to qualified in the batch management module and records its batch affixation, completing the entire process of information traceability from filling to warehousing. It is understood that other post-filling processing methods can also be used, such as interfaceing with the warehouse management system to achieve automatic updates of cylinder inventory information; this is not limited here.

[0073] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating the high-safety fully automatic filling method for filling fluorine-nitrogen mixed gases in this application.

[0074] S201. When the pressure deceleration rate exceeds the preset leakage judgment threshold, the target cylinder is judged to have a leak.

[0075] Among them, the pressure deceleration rate refers to the absolute value of the slope of the corrected fitting line obtained after performing a weighted least squares linear fitting on the sequence of periodic average pressure values ​​of multiple consecutive modulation cycles, reflecting the rate at which the pressure inside the cylinder decreases as the modulation cycle progresses; the preset leakage judgment threshold refers to the boundary value of the pressure drop range allowed under normal sealing conditions of the target cylinder, which is pre-calibrated and written into the control parameters according to the fluorine filling safety regulations; leakage refers to the phenomenon of gas escaping to the outside through the sealing failure point.

[0076] The calculated pressure deceleration rate is compared with a preset leakage detection threshold. When the pressure deceleration rate exceeds the threshold, the target cylinder is determined to be leaking, triggering subsequent isolation and tracing procedures. For example, if the preset leakage detection threshold is set to 0.5 kPa / cycle and the calculated pressure deceleration rate is 1.2 kPa / cycle, the detection condition is met. The system records the current pressure deceleration rate value and the corresponding modulation cycle number, locks the cylinder's filling status, and prevents it from entering the qualified product process.

[0077] S202. Close the filling shut-off valve connecting the target cylinder and the filling pipeline to put the target cylinder into a gas path isolation state.

[0078] Among them, the filling shut-off valve refers to the shut-off valve installed between the filling pipeline and the target cylinder valve port to open or close the gas passage, and the actuator is driven by an electrical signal to open or close; the filling pipeline refers to the high-pressure gas transmission pipeline connecting the gas source and the filling port of the target cylinder; the gas path isolation state refers to the working state in which the target cylinder and the filling pipeline are completely disconnected in the gas path after the filling shut-off valve is closed, and the internal gases of the two are no longer connected.

[0079] After determining that a leak exists in the target cylinder, the control system sends a shut-off command to the actuator of the filling shut-off valve. The actuator drives the valve core to close, cutting off the gas passage between the filling pipeline and the target cylinder, and the target cylinder enters a gas path isolation state. After the filling shut-off valve closes, the differential pressure sensor installed on both sides of the shut-off valve reads the gas path pressure difference. When this pressure difference is greater than the set confirmation value, the gas path isolation state is confirmed to have been established. The establishment of the gas path isolation state is a prerequisite for subsequent leak tracing. By eliminating the influence of gas connectivity on the filling pipeline side, the subsequent pressure detection results reflect only the airtightness of the target cylinder itself.

[0080] S203. In the gas path isolation state, keep the filling shut-off valve closed and continue to perform multiple preset modulation cycles continuously and completely on the target cylinder.

[0081] Among them, the gas path isolation state refers to the working state in which the target cylinder and the filling pipeline are completely disconnected in the gas path after the filling shut-off valve is closed; the filling shut-off valve is closed, which means that the valve core of the shut-off valve is always in the fully closed position and the gas cannot flow between the pipeline and the cylinder; the preset modulation cycle refers to the time period in which active enhanced heat dissipation is applied in the first half of the cycle and active heat dissipation is stopped in the second half of the cycle, and its duration and parameter settings are completely consistent with the modulation cycle in S102.

[0082] After the gas path isolation state is established, the filling shut-off valve remains closed. The target cylinder continues to undergo alternating operations of active enhanced cooling and cessation of active cooling with the same modulation cycle parameters as before isolation, continuously running multiple preset modulation cycles. Cooling airflow is applied to the target cylinder during the first half of each modulation cycle, and the cooling airflow is stopped during the second half, maintaining the same thermal excitation mode as before isolation. Maintaining the same thermal modulation method ensures that the average pressure value sequence acquired during the isolation cycle under gas path isolation conditions possesses the same thermal effect reduction characteristics as before isolation, thus making the calculated results of the subsequent isolation pressure reduction rate directly comparable to the pressure reduction rate before isolation.

[0083] S204. The internal pressure of the target gas cylinder is continuously collected, and the average pressure value of the isolation period corresponding to each modulation period under the gas path isolation state is calculated with the preset modulation period as the integration window.

[0084] Among them, the internal pressure of the cylinder refers to the pressure of the gas inside the target cylinder, which is continuously output by a pressure sensor installed inside the filling shut-off valve (i.e., the cylinder side) at a fixed sampling frequency; the integration window refers to the time span taken when performing statistical processing on the time-series pressure data, and here a complete preset modulation period is used as the integration window; the average pressure value of the isolation period refers to the representative pressure average value calculated by extracting steady-state data and equally weighting the internal pressure data collected within a complete modulation period under gas path isolation conditions.

[0085] In gas path isolation mode, the pressure sensor continuously outputs bottle pressure sampling data. Using each complete preset modulation cycle as an integration window, all sampling points within the window are divided into a first-half cycle pressure dataset and a second-half cycle pressure dataset according to time sequence. Transient fluctuation sampling points generated during the cooling airflow start-stop transition phase are removed from both datasets, retaining the steady-state sampling points to obtain the first-half cycle steady-state pressure data and the second-half cycle steady-state pressure data. The two sets of steady-state pressure data are then weighted and averaged to obtain the isolation cycle average pressure value for that modulation cycle. For example, if the modulation cycle is 60 seconds long, with 30 seconds before and after, and transient data for 5 seconds before and after each start-stop moment is removed, the steady-state sampling points within each 25-second interval are weighted and averaged to obtain the isolation cycle average pressure value for that cycle.

[0086] S205. Perform linear trend fitting on the isolation pressure value sequence composed of the average isolation pressure values ​​corresponding to multiple consecutive preset modulation periods to obtain the isolation pressure deceleration rate.

[0087] Among them, the isolation pressure value sequence refers to the time series data sequence composed of the average pressure values ​​of the isolation cycles of multiple consecutive modulation cycles under the gas path isolation state, arranged in the order of cycle number; linear trend fitting refers to the least square regression of the above sequence with a linear function to solve the linear equation describing the trend of the sequence with the cycle number; isolation pressure deceleration rate refers to the absolute value of the slope of the fitted line after performing weighted least square linear fitting on the isolation pressure value sequence.

[0088] The average pressure values ​​of multiple consecutive modulation cycles under gas path isolation conditions are arranged by cycle number to form an isolation pressure value sequence. A linear trend fitting is performed with the modulation cycle number as the independent variable and the average pressure value of the isolation cycle as the dependent variable. First, an equal-weighted least squares linear fit is performed on the isolation pressure value sequence to obtain an initial fitted line. The slope and intercept are obtained by solving a system of equations that minimize the sum of squared longitudinal distances between each data point. The absolute value of the residual of each isolation cycle's average pressure value relative to the predicted value corresponding to the initial fitted line is calculated. The fitting weight of each data point is calculated by adding the reciprocal of a small positive number to the absolute value of the residual; the larger the absolute value of the residual, the smaller the weight. A weighted least squares linear fit is then performed on the isolation pressure value sequence using the above weights to obtain a corrected fitted line. The absolute value of the slope of the corrected fitted line is taken as the isolation pressure deceleration rate. This calculation process is completely consistent with the calculation method of the pressure deceleration rate in S104 to ensure the comparability of the pressure deceleration rates before and after isolation.

[0089] S206. When the isolation pressure deceleration rate exceeds the preset leakage judgment threshold, the leakage source is determined to be the cylinder body of the target cylinder, and a cylinder leakage alarm is output.

[0090] Among them, the isolation pressure deceleration rate refers to the absolute value of the slope of the corrected fitting line after performing a weighted least squares linear fitting on the isolation pressure value sequence under gas path isolation conditions; the target cylinder body refers to the metal shell of the cylinder and its inherent components (including the cylinder body wall, the cylinder shoulder weld and the cylinder's own valve sealing surface), which is different from the detachable filling interface between the filling pipeline end and the cylinder valve port; the cylinder body leakage alarm refers to the alarm signal carrying leakage source information output by the control system to the operating terminal after determining that the leakage source is the cylinder body.

[0091] The calculated isolation pressure deceleration rate under gas path isolation conditions is compared with a preset leakage detection threshold. If the isolation pressure deceleration rate exceeds this threshold, the leak is determined to originate from the target cylinder. The determination is based on the following criteria: after the filling shut-off valve is closed, the gas path between the filling pipeline and the target cylinder is completely disconnected, and the gas inside the cylinder cannot escape through the filling port. If, under these conditions, the pressure inside the cylinder continues to decrease at a rate exceeding the threshold, the leak point must be located within the cylinder itself, including defects in the cylinder wall material or failure of the cylinder's valve sealing surface. Once the determination is established, the control system outputs a cylinder leakage alarm to the operating terminal. The alarm information includes the cylinder number, the isolation pressure deceleration rate, and the alarm time. The cylinder is then locked and prohibited from transfer until manual inspection and handling.

[0092] S207. When the isolation pressure deceleration rate does not exceed the preset leakage judgment threshold, the leakage source is determined to be the filling interface between the filling pipeline and the target cylinder, and a filling interface leakage alarm is output.

[0093] Among them, the filling interface refers to the detachable connection between the filling connector at the end of the filling pipeline and the valve port of the target cylinder, including the sealing interface such as the connector sealing ring and the threaded connection sealing surface; the filling interface leakage alarm refers to the alarm signal carrying the leakage source information output by the control system to the operation terminal after determining that the leakage source is the filling interface.

[0094] The calculated isolation pressure deceleration rate under gas path isolation conditions is compared with a preset leakage judgment threshold. If the isolation pressure deceleration rate does not exceed the threshold, the leakage source is determined to be the filling interface between the filling pipeline and the target cylinder. The judgment criteria are as follows: After gas path isolation, the filling shut-off valve closes the gas path where the filling interface is located. If the pressure deceleration rate inside the cylinder recovers to below the threshold after isolation, it indicates that the target cylinder itself is well sealed. The continuous pressure drop detected before isolation is caused by poor sealing of the filling interface, resulting in gas leakage through the interface gap. After the filling shut-off valve is closed, this leakage path is blocked, and the abnormal pressure drop inside the cylinder disappears. The control system outputs a filling interface leakage alarm to the operating terminal. The alarm information includes the filling station number, the pressure deceleration rate value before isolation, and the alarm time, prompting the inspection and replacement of the filling joint and sealing ring. The target cylinder can be re-tested for sealing after the filling interface is repaired and inspected.

[0095] S208. Within multiple consecutive preset modulation cycles, the average values ​​of the bottle pressure data collected in the first half of each preset modulation cycle and the bottle pressure data collected in the second half of each preset modulation cycle are respectively taken and arranged in chronological order to obtain the average pressure sequence of the first half cycle and the average pressure sequence of the second half cycle.

[0096] The first half-cycle refers to the time period during which active cooling is applied in each preset modulation cycle; the second half-cycle refers to the time period during which active cooling is stopped in each preset modulation cycle; the average pressure value of the first half-cycle segment refers to the representative pressure value obtained by taking the arithmetic mean of all pressure sampling points inside the bottle within the first half-cycle of a certain modulation cycle; the average pressure value of the second half-cycle segment refers to the representative pressure value obtained by taking the arithmetic mean of all pressure sampling points inside the bottle within the second half-cycle of a certain modulation cycle; the average pressure sequence of the first half-cycle segment refers to the sequence formed by arranging the average pressure values ​​of the first half-cycle segment of each of several consecutive modulation cycles in chronological order according to the cycle number; the average pressure sequence of the second half-cycle segment refers to the sequence formed by arranging the average pressure values ​​of the second half-cycle segment of each of several consecutive modulation cycles in chronological order according to the cycle number. The difference between this and the average pressure value of the isolated cycle in S204 is that S204 combines the steady-state data of the first and second half-cycles with equal weight and takes a unified average, while S208 keeps the first and second half-cycles independent and takes the arithmetic mean of all sampling points of each half-cycle, forming two independent sequences.

[0097] Within multiple consecutive preset modulation cycles, the pressure data inside the bottle for each complete preset modulation cycle is processed independently for the two half-cycles before and after. The specific calculation process is as follows: For the i-th modulation cycle, there are nf sampling points in the first half-cycle, and the pressure values ​​inside the bottle at each point are denoted as Pf_i_1, Pf_i_2, …, Pf_i_nf. The arithmetic mean is taken to obtain the average pressure value for the first half-cycle segment PF_i = (Pf_i_1 + Pf_i_2 + … + Pf_i_nf) / nf; there are nb sampling points in the second half-cycle, and the pressure values ​​inside the bottle at each point are denoted as Pb_i_1, Pb_i_2, …, Pb_i_nb. The arithmetic mean is taken to obtain the average pressure value for the second half-cycle segment PB_i = (Pb_i_1 + Pb_i_2 + … + Pb_i_nb) / nb. Repeat the above calculation for N consecutive modulation cycles to obtain the average voltage sequence {PF_1, PF_2, …, PF_N} for the first half of the cycle and the average voltage sequence {PB_1, PB_2, …, PB_N} for the second half of the cycle. Both sequences are aligned with the modulation cycle number i as the time index. For example, when the modulation cycle length is 60 seconds and the sampling frequency is 1 Hz, each half-cycle contains 30 sampling points. The average voltage value for the first half of the cycle is obtained by averaging these 30 points, and the average voltage value for the second half of the cycle is obtained by averaging the other 30 points.

[0098] S209. Perform linear trend fitting on the average pressure sequence of the first half-cycle and the average pressure sequence of the second half-cycle respectively to obtain the pressure deceleration rate of the first half-cycle and the pressure deceleration rate of the second half-cycle.

[0099] The definitions of the first-half-cycle average pressure sequence and the second-half-cycle average pressure sequence are given in S208. Linear trend fitting refers to performing least-squares regression on the two sequences using a linear function to solve for the linear equations describing the trend of each sequence with the modulation cycle number. The first-half-cycle pressure deceleration rate RF is the absolute value of the slope of the corrected fitted line obtained after performing a weighted least-squares linear fit on the first-half-cycle average pressure sequence; the second-half-cycle pressure deceleration rate RB is the absolute value of the slope of the corrected fitted line obtained after performing a weighted least-squares linear fit on the second-half-cycle average pressure sequence. The two rates are calculated separately based on their respective independent sequences and do not share the same fitting process.

[0100] Using modulation period number i as the independent variable, and the average voltage values ​​of the first half-cycle average voltage sequence and the second half-cycle average voltage sequence as the dependent variable, a complete weighted least squares linear fitting process is performed independently on each of the two sequences. Taking the first half-cycle average pressure sequence as an example: First, perform equal-weighted least squares linear fitting to solve for the initial slope kF0 and intercept bF0, minimizing the sum of squared residuals Σ(PF_i - kF0×i - bF0)² for each cycle; calculate the absolute value of the residuals at each data point eF_i = |PF_i - (kF0×i + bF0)|, and calculate the weight of each point according to wF_i = 1 / (eF_i + ε), where ε is a small positive number to prevent division by zero; perform weighted least squares linear fitting on the first half-cycle average pressure sequence with weight wF_i to solve for the corrected slope kF, minimizing the weighted sum of squared residuals Σ wF_i×(PF_i - kF×i - bF)², and take RF = |kF| as the pressure deceleration rate for the first half-cycle. Perform the same process on the second half-cycle average pressure sequence to obtain the pressure deceleration rate RB = |kB| for the second half-cycle. The above calculation process is consistent with the calculation method of the isolation pressure deceleration rate in S205, so as to ensure the comparability of the values ​​among the three.

[0101] S210. When both the pressure deceleration rate of the first half-cycle and the pressure deceleration rate of the second half-cycle exceed the preset leakage judgment threshold, it is confirmed that there is a real leak in the target cylinder. The pressure deceleration rate of the first half-cycle and the pressure deceleration rate of the second half-cycle are weighted and fused to obtain the fused pressure deceleration rate, and the fused pressure deceleration rate is output as the final leakage rate evaluation value of the target cylinder.

[0102] The definitions of the pressure deceleration rate RF and the pressure deceleration rate RB in the first half-cycle and the second half-cycle are given in S209. Weighted fusion refers to the process of correcting the initial fusion result by introducing a thermal bias correction based on the arithmetic mean, thereby reducing the impact of thermal modulation asymmetry between the first and second half-cycles on the rate estimation and obtaining a fusion value that is closer to the actual leakage rate. The fused pressure deceleration rate Rfused refers to the fused value of the pressure deceleration rates between the first and second half-cycles after thermal bias correction, which serves as the final evaluation output of the target cylinder leakage rate. The thermal asymmetry index ΔR refers to the absolute value of the difference between RF and RB, i.e., ΔR = |RF - RB|, which reflects the magnitude of the difference in pressure deceleration rate caused by thermal modulation effect in the two half-cycles. The thermal bias correction C refers to the correction value of the initial fused pressure deceleration rate determined according to the positive correlation mapping relationship of ΔR. The larger ΔR is, the larger C is.

[0103] When both RF and RB exceed the preset leakage detection threshold, a real leak is determined in the target cylinder, and the fusion pressure deceleration rate is calculated according to the following steps: First, calculate the initial fusion pressure deceleration rate: Rinit = (RF + RB) / 2, which is the arithmetic mean of the pressure deceleration rates for the first and second half-cycles. Second, calculate the thermal asymmetry index: ΔR = |RF - RB|. Active cooling in the first half-cycle lowers the gas temperature compared to the second half-cycle, resulting in a systematic difference in the absolute steady-state pressure levels between the two half-cycles. This difference creates a thermal bias at the rate calculation level. A larger ΔR indicates a greater impact of the thermal bias on Rinit, and Rinit is more likely to overestimate the actual leakage rate. The third step is to determine the thermal bias correction C based on ΔR, using a linear mapping method: C = α × ΔR, where α is a preset thermal bias correction coefficient, with a value range of 0 < α < 0.5, determined through calibration experiments—repeated measurements are performed on a standard cylinder with a known actual leakage rate under the same modulation parameters, and the α value that minimizes the root mean square deviation between Rfused and the actual leakage rate is selected and written into the control parameters. The fourth step is to calculate the fusion pressure deceleration rate:

[0104] Rfused = Rinit - C = (RF + RB) / 2 - α × |RF - RB|

[0105] For example, if RF = 1.8 kPa / cycle, RB = 1.2 kPa / cycle, α = 0.2, then Rinit = 1.5 kPa / cycle, ΔR = 0.6 kPa / cycle, C = 0.12 kPa / cycle, and Rfused = 1.5 - 0.12 = 1.38 kPa / cycle. This value is used as the final leakage rate assessment value of the target cylinder, and is written into the detection record along with the cylinder number and detection time.

[0106] In some embodiments, the step of weighted fusion of the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle to obtain the fusion pressure deceleration rate specifically includes: taking the arithmetic mean of the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle to obtain the initial fusion pressure deceleration rate; calculating the absolute value of the difference between the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle to obtain the thermal asymmetry index; determining the thermal bias correction amount based on the thermal asymmetry index, wherein the thermal bias correction amount is positively correlated with the thermal asymmetry index; and subtracting the thermal bias correction amount from the initial fusion pressure deceleration rate to obtain the fusion pressure deceleration rate.

[0107] The pressure deceleration rate RF in the first half-cycle and the pressure deceleration rate RB in the second half-cycle are the absolute values ​​of the corrected slopes obtained after performing weighted least-squares linear fitting on the average pressure sequences of the first and second half-cycles, respectively. The former corresponds to the period when active cooling is applied in each modulation cycle, and the latter corresponds to the period when active cooling is stopped. The unit is kPa / cycle. For example, RF = 1.8 kPa / cycle means that the average pressure in the first half-cycle decreases by an average of 1.8 kPa after each modulation cycle. The preset leakage judgment threshold is the pressure deceleration rate benchmark value written into the control parameters through calibration experiments. If the pressure deceleration rate exceeds this value, it is determined that the pressure drop is caused by actual leakage rather than measurement noise. The initial fused pressure deceleration rate Rinit is the arithmetic mean of RF and RB. For example, when RF = 1.8 and RB = 1.2, Rinit = 1.5 kPa / cycle. The thermal asymmetry index ΔR is the absolute value of the difference between RF and RB, reflecting the magnitude of the rate difference caused by the thermal modulation effect in the first and second half-cycles. For example, ΔR = |1.8 - 1.2| = 0.6 kPa / cycle. The thermal bias correction C is a correction value determined based on a positive correlation with ΔR; the larger ΔR is, the larger C is. It is used to quantify and reduce the interference of thermal modulation asymmetry on the initial fusion rate. The fusion pressure deceleration rate Rfused is the final rate value obtained by subtracting C from Rinit, and serves as the final evaluation output for the target cylinder leakage rate.

[0108] When both RF and RB exceed the preset leakage detection threshold, a real leak is determined in the target cylinder, and the following four-step calculation process is executed to obtain the fusion pressure deceleration rate. Step 1: Calculate the initial fusion pressure deceleration rate: Rinit = (RF + RB) / 2. Take the arithmetic mean of the rates of the first and second half-cycles, assigning equal initial weights to both to obtain a basic fusion rate value. Step 2: Calculate the thermal asymmetry index: ΔR = |RF - RB|. This index reflects the absolute magnitude of the difference between the rate estimates of the two half-cycles—the first half-cycle continuously applies active heat dissipation, resulting in a gas steady-state temperature lower than the steady-state temperature after heat dissipation stops in the second half-cycle. The gas temperature in the two half-cycles has a systematic difference, leading to the introduction of thermal biases in opposite directions in their respective rate estimates. The larger ΔR is, the greater the absolute influence of the thermal bias, and the greater the degree to which Rinit overestimates the true leakage rate. Therefore, it is necessary to subtract a correction amount positively correlated with ΔR from Rinit to restore the true leakage rate. The third step is to calculate the thermal bias correction C based on ΔR using a linear mapping method: C = α × ΔR, where α is a preset thermal bias correction coefficient with a value range of 0 < α < 0.5. α is determined through calibration experiments. Specifically, RF and RB are repeatedly collected from standard cylinders with known actual leakage rates under the same modulation period parameters. For multiple sets of standard samples with known leakage rates, Rfused = (RF + RB) / 2 - α × |RF - RB| is calculated. The α value that minimizes the root mean square difference between Rfused and the known actual leakage rate of all samples is selected and fixed into the control parameters for subsequent detection. The fourth step is to subtract C from Rinit to obtain the fusion pressure deceleration rate: Rfused = Rinit - C = (RF + RB) / 2 - α × |RF - RB|. Rfused is used as the final leakage rate evaluation value of the target cylinder and output as the cylinder number and detection time, and written into the detection record. For example, if RF = 1.8 kPa / cycle, RB = 1.2 kPa / cycle, and α = 0.2, then Rinit = (1.8 + 1.2) / 2 = 1.5 kPa / cycle, ΔR = |1.8 - 1.2| = 0.6 kPa / cycle, C = 0.2 × 0.6 = 0.12 kPa / cycle, and Rfused = 1.5 - 0.12 = 1.38 kPa / cycle. We will use 1.38 kPa / cycle as the final leakage rate assessment value for this cylinder.

[0109] S211. If either the current half-cycle pressure deceleration rate or the second half-cycle pressure deceleration rate exceeds the preset leakage judgment threshold, the target cylinder will be re-tested for sealing after the preset modulation period is extended.

[0110] "Only one of them exceeds the preset leakage judgment threshold" means that exactly one of RF and RB is greater than the preset leakage judgment threshold while the other does not exceed the threshold, which is different from the two cases where both exceed the threshold (S210) and neither exceeds the threshold (no leakage conclusion); "Extending the preset modulation cycle" means increasing the total duration of each preset modulation cycle from the current setting value T to the new duration T', with the duration of the first and second half cycles each extended proportionally from T / 2 to T' / 2; "Re-execute the sealing test" means re-performing the entire process from continuous pressure acquisition to calculating RF and RB and making a judgment on the target cylinder with the extended modulation cycle parameters.

[0111] When only one of RF and RB exceeds the preset leakage threshold, it indicates that the amplitude of pressure fluctuation interference caused by thermal modulation is on the same order of magnitude as the amplitude of the actual leakage signal under the current modulation cycle duration. This results in a significant inconsistency in the apparent deceleration rate between the first and second half-cycles, making it impossible to make a reliable judgment of actual leakage under the current parameters. The physical mechanism is as follows: When the modulation cycle duration is short, the duration of the temperature transition phase in each half-cycle accounts for a large proportion of the total half-cycle duration. The short-term fluctuations in gas pressure caused by rapid temperature changes have a relatively prominent interference amplitude on the average pressure value of the half-cycle segment, causing the apparent deceleration rate between the first and second half-cycles to differ beyond the range explainable by the actual leakage contribution. After extending the modulation cycle, the proportion of the duration of the temperature stabilization segment to the duration of the temperature transition segment in each half-cycle increases. The contribution of thermal modulation interference is sufficiently diluted in the mean of the longer integration window, making PF_i and PB_i more accurately reflect the steady-state gas pressure in each half-cycle. Consequently, when an actual leakage exists, RF and RB tend to be consistent and both exceed the threshold, while when there is no actual leakage, both return to below the threshold. The specific implementation of the extension operation is as follows: Multiply the current modulation cycle duration T by a preset extension coefficient β (e.g., β = 2) to obtain a new modulation cycle duration T' = β × T. Extend the duration of each half-cycle to T' / 2, and re-execute the complete detection process from S208 to S210 with the new parameters. For example, if the current modulation cycle duration T = 60 seconds, after setting β = 2, the new modulation cycle duration T' = 120 seconds, with each half-cycle being 60 seconds. Re-perform the sealing test on the target cylinder with these parameters until a clear conclusion is reached that both RF and RB exceed the threshold or neither exceeds the threshold.

[0112] The automatic filling system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the physical device structure of an automatic filling system in the embodiments of this application.

[0113] It should be noted that, Figure 3The structure of the automated filling system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0114] like Figure 3 As shown, the automatic filling system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on a program stored in Read-Only Memory (ROM) 302 or a program loaded from storage section 308 into Random Access Memory (RAM) 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0115] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0116] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0117] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0118] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0119] Specifically, the automatic filling system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the high-safety fully automatic filling method for filling fluorine-nitrogen mixed gases provided in the above embodiment.

[0120] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the automatic filling system described in the above embodiments; or it may exist independently and not incorporated into the automatic filling system. The storage medium carries one or more computer programs that, when executed by a processor of the automatic filling system, cause the automatic filling system to implement the high-safety fully automatic filling method for filling fluorine-nitrogen mixed gases provided in the above embodiments.

[0121] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0122] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0123] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A highly safe, fully automated filling method for filling fluorine-nitrogen mixed gases, characterized in that, Applied to an automatic filling system, the method includes: Fluorine-nitrogen mixed gas is introduced into the target cylinder, and the filling quality of the target cylinder is collected in real time. When the collected filling quality data reaches the target filling amount, the filling is stopped. Active enhanced heat dissipation is applied to the target cylinder during the first half of each preset modulation cycle, and active heat dissipation is stopped during the second half of each preset modulation cycle. Multiple complete preset modulation cycles are executed alternately in a continuous manner. The internal pressure of the target steel cylinder is continuously collected. Using the preset modulation period as the integration window, the average pressure value of each modulation period is calculated based on the internal pressure data collected in each complete modulation period. A linear trend fit is performed on the pressure value sequence consisting of the periodic average pressure values ​​corresponding to multiple consecutive preset modulation periods to obtain the pressure deceleration rate. When the pressure deceleration rate exceeds a preset leakage detection threshold, it is determined that the target cylinder is leaking; When the pressure deceleration rate does not exceed the preset leakage judgment threshold, the target cylinder is determined to be sealed and the filling of the fluorine-nitrogen mixed gas is completed.

2. The method according to claim 1, characterized in that, The step of applying active enhanced heat dissipation to the target cylinder during the first half of each preset modulation cycle specifically includes: At the beginning of the first half of the current modulation cycle, the surface temperature of the target cylinder is collected to obtain the reference surface temperature of the current cycle. During the first half of the cycle, a cooling airflow is continuously applied to the target cylinder, and the real-time surface temperature of the cylinder is collected. Based on the difference between the real-time bottle surface temperature and the current cycle reference surface temperature, the flow rate of the cooling airflow is adjusted in a closed-loop feedback manner to ensure that the active heat dissipation enhancement amplitude remains consistent in each modulation cycle. At the end of the first half of the cycle, the cooling airflow is stopped, and the active enhanced heat dissipation of the current modulation cycle is completed.

3. The method according to claim 1, characterized in that, The step of calculating the average pressure value of each modulation cycle based on the bottle pressure data collected within each complete modulation cycle, using the preset modulation cycle as the integration window, specifically includes: The pressure data inside the bottle collected in each complete modulation cycle is divided into a first half-cycle pressure dataset and a second half-cycle pressure dataset according to time sequence. From the first half-cycle pressure dataset and the second half-cycle pressure dataset, respectively, transient pressure fluctuation data is removed and steady-state pressure data is retained to obtain the first half-cycle steady-state pressure data and the second half-cycle steady-state pressure data. The steady-state pressure data of the first half-cycle and the steady-state pressure data of the second half-cycle are combined with equal weight to obtain the periodic average pressure value corresponding to the current modulation cycle.

4. The method according to claim 1, characterized in that, The step of performing linear trend fitting on the pressure value sequence composed of the periodic average pressure values ​​corresponding to multiple consecutive preset modulation periods to obtain the pressure deceleration rate specifically includes: An initial least-squares linear fit is performed on the pressure value sequence to obtain an initial fitted line; Calculate the residual of the average pressure value of each period in the pressure value sequence relative to the initial fitted straight line; Based on the absolute value of the residual corresponding to the average pressure value of each cycle, a fitting weight is assigned to the average pressure value of each cycle, and the pressure value sequence is subjected to a weighted least squares linear fit using the fitting weight to obtain a corrected fitting line. The absolute value of the slope of the corrected fitted line is used as the pressure deceleration rate.

5. The method according to claim 1, characterized in that, After the step of determining that the target cylinder is leaking when the pressure deceleration rate exceeds a preset leakage detection threshold, the method further includes: Shut down the filling shut-off valve connecting the target cylinder to the filling pipeline, so that the target cylinder enters a gas path isolation state; Under the gas path isolation state, the filling shut-off valve remains closed, and multiple consecutive and complete preset modulation cycles are continued to be performed on the target cylinder; The internal pressure of the target gas cylinder is continuously collected, and the average pressure value of the isolation period corresponding to each modulation period under the gas path isolation state is calculated with the preset modulation period as the integration window. A linear trend fit is performed on the isolation pressure value sequence, which is composed of the average isolation pressure values ​​corresponding to multiple consecutive preset modulation periods, to obtain the isolation pressure deceleration rate. When the isolation pressure deceleration rate exceeds the preset leakage judgment threshold, the leakage source is determined to be the cylinder body of the target cylinder, and a cylinder leakage alarm is output. When the isolation pressure deceleration rate does not exceed the preset leakage judgment threshold, the leakage source is determined to be the filling interface between the filling pipeline and the target cylinder, and a filling interface leakage alarm is output.

6. The method according to claim 1, characterized in that, After the step of determining that the target cylinder is leaking when the pressure deceleration rate exceeds a preset leakage detection threshold, the method further includes: Within multiple consecutive preset modulation cycles, the average values ​​of the bottle pressure data collected in the first half of each preset modulation cycle and the bottle pressure data collected in the second half of each preset modulation cycle are respectively taken and arranged in chronological order to obtain the average pressure sequence of the first half cycle and the average pressure sequence of the second half cycle. Linear trend fitting is performed on the average pressure sequence of the first half-cycle segment and the average pressure sequence of the second half-cycle segment respectively to obtain the pressure deceleration rate of the first half-cycle and the pressure deceleration rate of the second half-cycle. When both the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle exceed the preset leakage judgment threshold, it is confirmed that the target cylinder has a real leak. The pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle are weighted and fused to obtain the fused pressure deceleration rate, and the fused pressure deceleration rate is output as the final leakage rate evaluation value of the target cylinder. If either the pressure deceleration rate in the first half-cycle or the pressure deceleration rate in the second half-cycle exceeds the preset leakage threshold, the preset modulation period is extended, and the sealing test on the target cylinder is re-performed.

7. The method according to claim 6, characterized in that, The step of weightedly fusing the pressure deceleration rate of the first half-cycle and the pressure deceleration rate of the second half-cycle to obtain the fused pressure deceleration rate specifically includes: The initial fusion pressure deceleration rate is obtained by arithmetically averaging the pressure deceleration rate of the first half-cycle and the pressure deceleration rate of the second half-cycle. The absolute value of the difference between the pressure deceleration rate in the first half-cycle and the pressure deceleration rate in the second half-cycle is calculated to obtain the thermal asymmetry index. A thermal bias correction amount is determined based on the thermal asymmetry index, and the thermal bias correction amount is positively correlated with the thermal asymmetry index; The fusion pressure deceleration rate is obtained by subtracting the thermal bias correction from the initial fusion pressure deceleration rate.

8. An automatic filling system, characterized in that, The automatic filling system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the automatic filling system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the automatic filling system, the automatic filling system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the automatic filling system, the automatic filling system performs the method as described in any one of claims 1-7.

Images