Liquid micro-leakage precision detection method based on liquid level pressure difference

Abstract

The application is a liquid micro-leakage precision detection method based on liquid level pressure difference, and relates to the technical field of container leakage detection, comprising: removing outliers and smoothing filtering on the effective pressure difference time sequence to obtain an actual pressure difference sequence and a residual sequence, calculating an arithmetic mean and a standard deviation, and converting the standard deviation to obtain a pre-processed actual pressure difference sequence and a standardized theoretical pressure difference sequence.In the application, a thermodynamic quasi-steady state judgment mechanism can accurately identify the stable stage of the thermodynamic state of the container, avoid the influence of transient changes after pressurization on the detection result, and help shorten the detection period; through dynamic gas volume calculation and theoretical pressure difference decay sequence generation, the thermal expansion and cold contraction effect of the container and the liquid, the gas-liquid two-phase equilibrium, and the atmospheric pressure fluctuation and other factors are comprehensively considered, and the pressure difference change law under the condition of no leakage can be accurately simulated.

Description

Technical Field

[0001] This invention relates to the field of container leak detection technology, and in particular to a precise detection method for micro-leaks of liquids based on liquid level pressure difference. Background Technology

[0002] The sealing performance of liquid containers is crucial for safe production, product quality control, and environmental protection in various industries such as chemical, petroleum, food, and pharmaceutical. Micro-leakage detection, as a core technology for assessing container sealing reliability, plays a key role throughout the entire industrial production process. Currently, the industry has developed several mature detection technologies, including the bubble method, pressure decay method, and helium mass spectrometry leak detection method, each suitable for different testing scenarios and accuracy requirements. Among these, the pressure decay method has gained widespread application in industrial testing scenarios with large-scale production and conventional accuracy requirements due to its convenient equipment deployment, simple operation process, and controllable testing costs.

[0003] In practical applications, there is still room for optimization in detection methods based on the principle of pressure decay. Natural fluctuations in ambient temperature and atmospheric pressure can affect the internal pressure of the container through thermal expansion and contraction and changes in the gas-liquid two-phase equilibrium. Pressure changes caused by these environmental factors can easily be confused with pressure decay caused by actual leakage, which may affect the accuracy of the detection results.

[0004] Furthermore, the thermodynamic transient process after the container is pressurized, the pressure spikes generated by liquid sloshing, and the high-frequency noise caused by on-site mechanical vibration can also affect the quality of the differential pressure signal, which to some extent limits the detection method's ability to identify minute leaks. Summary of the Invention

[0005] The purpose of this invention is to propose a precise detection method for micro-leakage of liquid based on liquid level pressure difference in order to solve the above-mentioned problems.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: Precise detection methods for micro-leaks in liquids based on liquid level and pressure difference include: Acquire the original liquid level and pressure difference signal and various environmental parameter signals of the container under test. After determining that the internal thermodynamic state of the container has entered the quasi-steady state period, determine the effective detection start time and record the effective pressure difference time series vector and environmental parameter vector. The dynamic gas volume is calculated based on environmental parameter vectors, and the theoretical total absolute pressure is derived by combining the ideal gas law and the Antoine equation. The theoretical pressure difference attenuation sequence is generated by subtracting the external atmospheric pressure signal. Outlier removal and smoothing filtering are performed on the effective pressure difference time series vector to obtain the actual pressure difference sequence and residual sequence. The arithmetic mean and standard deviation are calculated, and standard score conversion is performed to obtain the preprocessed actual pressure difference sequence and standardized theoretical pressure difference sequence. Construct a local distance matrix, calculate the cumulative distance matrix using the state transition equation under a banded constraint window, find the regularized path by backtracking and calculate the normalized dynamic time regularized distance; The path deviation area of ​​the regularized path is calculated, and an adaptive threshold is generated by combining the variance of the residual sequence and the variance of the ambient temperature signal. The equivalent pressure drop value is calculated using the normalized dynamic time regularization distance and standard deviation and converted into the equivalent volume leakage rate. When the judgment condition is met, the micro-leakage detection result is output.

[0007] Preferably, the process of acquiring the original liquid level pressure difference signal and various environmental parameter signals of the container under test, and determining that the internal thermodynamic state of the container has entered the quasi-steady state period, is as follows: A differential pressure sensor and a multi-point temperature sensor array are configured in the container under test to acquire gas phase temperature signal, liquid phase temperature signal, ambient temperature signal and external atmospheric pressure signal as various environmental parameter signals; A hardware timer is used to broadcast a synchronization conversion pulse to the analog-to-digital conversion channel, which simultaneously freezes the analog signal and performs quantization conversion to eliminate time phase deviation during multi-channel acquisition. Within a time sliding window, the difference between the maximum and minimum values ​​of the gas phase temperature signal and the liquid phase temperature signal is extracted respectively, and divided by the time length of the time sliding window to obtain the rate of change of the gas phase temperature signal and the rate of change of the liquid phase temperature signal. When the rate of change of both the gas phase temperature signal and the rate of change of the liquid phase temperature signal are lower than the steady-state determination threshold for multiple consecutive sampling periods, it is determined that the thermodynamic state inside the container has passed the transient change period and entered the quasi-steady-state period.

[0008] Preferably, the calculation of dynamic gas volume based on environmental parameter vectors is performed using the following method: ; ; in, Sampling time, This refers to the total dynamic volume of the container. Initial calibration temperature The internal total volume calibration value of the lower container. Let be the coefficient of volumetric expansion of the container material. The ambient temperature signal is part of the environmental parameter vector. For the dynamic volume of the liquid, Initial calibration temperature The initial volume calibration value of the liquid below. is the coefficient of volumetric expansion of the liquid. The liquid phase temperature signal is part of the environmental parameter vector. Before calculating the dynamic gas volume, an initial calibration temperature is set, and the calibration values ​​of the total internal volume of the container and the initial volume of the liquid are obtained at the initial calibration temperature. By introducing the volume expansion coefficient of the container material and the volume expansion coefficient of the liquid, the thermal expansion and contraction effect caused by the fluctuation of the ambient temperature signal and the liquid phase temperature signal is quantified. The dynamic gas volume is obtained by subtracting the dynamic volume of the liquid from the total dynamic volume of the container.

[0009] Preferably, the process of generating the theoretical pressure drop decay sequence includes the following formula: ; in, This represents the theoretical pressure difference value in the theoretical pressure difference decay sequence. For the amount of gaseous substance, Let be the ideal gas constant. The gas phase temperature signal is part of the environmental parameter vector. The saturated vapor pressure was calculated using the Antoine equation combined with the liquid phase temperature signal. This refers to the external atmospheric pressure signal in the environmental parameter vector; Before using the formula to derive the theoretical total absolute pressure, the amount constant of the gas substance sealed inside the container is deduced by effectively detecting the actual absolute pressure, dynamic gas volume and gas phase temperature signal at the initial moment. In the derivation process, the saturated vapor pressure generated by liquid evaporation is calculated point by point to compensate for the vapor partial pressure under the gas-liquid two-phase equilibrium state.

[0010] Preferably, the logic for outlier removal from the effective pressure difference time series vector is as follows: Set a fixed-width sliding data window. As the sliding data window slides point by point on the effective pressure difference time series vector, sort the data points contained in the sliding data window and find the median. Calculate the absolute deviation between each data point in the sliding data window and the median, and then find the median of the absolute deviations. When the absolute deviation between the data point at the center of the sliding data window and the median is greater than the product of the judgment threshold coefficient and the median of the absolute deviation value, the data point at the center is determined to be an abnormal spike data caused by liquid sloshing, and the abnormal spike data is forcibly replaced with the median to preserve the step change trend of the differential pressure signal.

[0011] Preferably, the logic for the smoothing filter and standard score conversion is as follows: A low-order polynomial is constructed within a sliding window. The polynomial coefficients are found by solving the least squares normal equations. The smoothed value at the center of the window is calculated using the fitted polynomial to replace the noisy data, thus obtaining the actual pressure difference sequence after filtering out high-frequency mechanical vibration noise. The residual sequence is obtained by subtracting the actual pressure difference sequence from the effective pressure difference time series vector. Calculate the arithmetic mean and standard deviation of the actual pressure difference sequence, and subtract the arithmetic mean from each data point in the actual pressure difference sequence and the theoretical pressure difference decay sequence, and divide by the standard deviation. Map the result to a dimensionless mathematical space with a mean of zero and a variance of one, in order to eliminate the static reference deviation caused by different initial pressurization states.

[0012] Preferably, the process of calculating the cumulative distance matrix using the state transition equation under the banded constraint window includes the following state transition equation: ; ; Among them, the matching point The constraints of the strip constraint window must be satisfied. , This is the width parameter of the strip constraint window; For elements in the cumulative distance matrix, These are elements in the local distance matrix. , and These are the cumulative distances of three adjacent preceding states; Before recursively solving the state transition equation, the boundary conditions are initialized. The cumulative distance of the starting position is set as the local spatial distance of the starting position, and the path that does not conform to the physical logic is assigned a maximum value. The three pre-states in the state transition equation represent the synchronous matching on the time axis, the actual pressure difference response lag, and the actual pressure difference response lead, respectively.

[0013] Preferably, the calculation of the path deviation area of ​​the regularized path specifically includes: Define the main diagonal in the matrix space to represent absolute synchronization of the time axis; Extract the actual sequence time index and the theoretical sequence time index corresponding to any node on the regularized path; Calculate the absolute value of the difference between the actual sequence time index and the theoretical sequence time index and divide it by the square root of two to obtain the orthogonal geometric distance from the node to the main diagonal of the matrix space; Integrate and sum the orthogonal geometric distances from all nodes on the regular path to the main diagonal, and divide the sum by the total length of the regular path to obtain the path deviation area that quantifies the degree of continuous morphological distortion and bending caused by microleakage.

[0014] Preferably, the process of generating an adaptive threshold by combining the variance of the residual sequence and the variance of the ambient temperature signal adopts the following calculation method: ; ; in, An adaptive threshold for normalized dynamic time warping distance. An adaptive threshold for the path deviation area. Let V be the variance of the residual sequence. Let V be the variance of the ambient temperature signal in the ambient parameter vector. and This is the noise penalty factor. and This is the temperature fluctuation penalty coefficient. and The basic tolerance constant; The variance of the residual sequence obtained in the morphological preprocessing step of the differential pressure time series signal is calculated to characterize the combined level of high-frequency electrical noise and mechanical vibration noise of the sensor in the detection environment, thereby realizing dynamic adjustment of the threshold increase when the environment is harsh and the threshold decrease when the environment is stable.

[0015] Preferably, the content converted into equivalent volumetric leakage rate includes the following formula: ; in, Equivalent volume leakage rate, This is the equivalent pressure drop value. The average volume of the dynamic gas volume. The effective detection period is the time span. For standard reference temperature, The average gas phase temperature is the gas phase temperature signal in the environmental parameter vector. Standard reference atmospheric pressure; Before using the formula transformation, the square root of the normalized dynamic time-normalized distance is multiplied by the standard deviation of the actual pressure difference sequence to restore it to the equivalent pressure drop value with pressure physical dimensions, so as to characterize the pressure drop caused purely by the loss of gas mass inside the container after excluding the interference of ambient temperature and atmospheric pressure.

[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention uses a thermodynamic quasi-steady-state determination mechanism to accurately identify the stable stage of the thermodynamic state inside the container, avoiding the influence of transient changes after pressurization on the detection results and helping to shorten the detection cycle. By calculating the dynamic gas volume and generating the theoretical pressure difference decay sequence, it comprehensively considers factors such as the thermal expansion and contraction effect of the container and liquid, the gas-liquid two-phase equilibrium, and atmospheric pressure fluctuations, and can more accurately simulate the pressure difference change law under leak-free conditions.

[0017] 2. By comparing and analyzing the actual pressure difference sequence with the theoretical sequence, this invention can effectively distinguish between pressure changes caused by environmental factors and pressure decay caused by actual leakage, significantly reducing the occurrence of false positive detection results, improving the accuracy and reliability of the detection results, and providing a more reliable basis for evaluating the sealing performance of containers in industrial production. Attached Figure Description

[0018] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0019] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.

[0020] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.

[0021] Example 1

[0022] Its specific implementation method is combined with the appendix Figure 1 Please provide a detailed explanation.

[0023] Appendix Figure 1 The flowchart of the precise detection method for liquid micro-leakage based on liquid level and pressure difference provided in the embodiments of the present invention shows the complete steps from synchronous acquisition of pressure difference and environmental data to distortion path quantification and micro-leakage determination.

[0024] In this embodiment, it includes: Step 1: Synchronous acquisition of differential pressure and environmental data: Acquire the original liquid level differential pressure signal and various environmental parameter signals of the container under test. After determining that the internal thermodynamic state of the container has entered the quasi-steady state period, determine the effective detection start time and record the effective differential pressure time series vector and environmental parameter vector. The core objective of this step is to extract, with high fidelity, the fundamental multidimensional time-series data from the complex physical environment for subsequent morphological matching and theoretical sequence generation.

[0025] In practical industrial applications of micro-leak detection, the pressure changes inside the container under test are extremely weak. These weak signals are often masked by natural fluctuations in ambient temperature, tidal changes in atmospheric pressure, mechanical vibrations in industrial settings, and the thermodynamic instability of the gas-liquid two-phase system inside the container.

[0026] Establishing a high-precision, strictly synchronized, and state-adaptive multi-source data acquisition architecture is a physical prerequisite for ensuring the effectiveness of all subsequent digital signal processing algorithms. This step involves not only the acquisition of a single differential pressure signal, but also the comprehensive perception and digital characterization of various environmental parameters affecting the thermodynamic state inside the container.

[0027] A multi-source sensor hardware sensing network needs to be constructed. A high-precision differential pressure sensor is configured at the designated detection interface of the container under test.

[0028] The positive pressure side of the sensor is tightly connected to the gas space inside the container through a pressure-conducting pipeline, while the negative pressure side is directly connected to the external atmospheric environment. It is used to obtain the relative pressure difference signal between the gas space pressure inside the container and the atmospheric pressure of the external environment in real time.

[0029] To capture extremely slight pressure decay (typically in the Pascal or even microPascal range), this micro differential pressure sensor requires a microelectromechanical system (MEMS) chip based on the silicon piezoresistive effect or capacitive sensing principle, giving it extremely high resolution and extremely low zero-point drift characteristics.

[0030] To accurately reconstruct the complex thermodynamic state inside the container, a multi-point temperature sensor array must be configured.

[0031] Since there are usually two phases, gas and liquid, inside the container, and the specific heat capacity and thermal conductivity of the gas and liquid are significantly different, the temperature response rates of the gas and liquid phases are completely different when the external ambient temperature changes, resulting in a significant temperature gradient.

[0032] For example, gases have lower thermal inertia and respond more quickly to temperature changes; while liquids have higher heat capacity and their temperature changes are relatively slower.

[0033] The temperature sensor array specifically includes: a gas phase temperature sensor positioned at the center of the gas space inside the container, used to characterize the average thermodynamic temperature of the gas space inside the container; a liquid phase temperature sensor positioned deep within the lower part of the liquid inside the container, used to characterize the average thermodynamic temperature of the liquid inside the container; and an ambient temperature sensor positioned on the outer surface of the container or in the external space adjacent to the container, used to characterize the real-time environmental thermal shock conditions outside the container. These temperature sensors typically use high-precision platinum resistance thermometers (such as PT100 or PT1000) to ensure high linearity and measurement accuracy over a wide temperature range.

[0034] Considering that external atmospheric pressure is not constant, changes in weather systems (such as the alternation of high and low pressure cyclones) or differences in altitude can cause significant fluctuations in atmospheric pressure, which directly affect the measurement benchmark of the differential pressure sensor.

[0035] A high-precision absolute pressure sensor also needs to be installed at the testing site to obtain the absolute numerical signal of the external atmospheric pressure. This absolute pressure sensor has a vacuum reference chamber encapsulated inside, which can be unaffected by external air pressure fluctuations and output the true absolute ambient air pressure value.

[0036] All the weak analog signals output by the aforementioned sensors (such as millivolt-level voltage signals or 4-20 mA current signals) must be connected to a high-performance data acquisition terminal with multi-channel synchronous sampling capabilities.

[0037] Before a signal enters the analog-to-digital converter (ADC), it must pass through a precision conditioning circuit in the analog front end. This conditioning circuit includes an instrumentation amplifier with a high common-mode rejection ratio and an anti-aliasing low-pass filter. The instrumentation amplifier amplifies weak sensor signals to the preset input range of the ADC, while its differential input characteristics strongly suppress power frequency electromagnetic interference and common-mode noise commonly found in industrial environments. Anti-aliasing low-pass filters, based on the Nyquist sampling theorem, filter out high-frequency noise components above half the sampling frequency, preventing high-frequency noise from folding in the spectrum during digitization and thus mixing into the low-frequency effective signal band, ensuring the spectral purity of the original signal.

[0038] Setting a unified global sampling clock is crucial in the process of digitizing data acquisition.

[0039] Let the global sampling period be The sampling start time is Then, a continuous sequence of sampling times can be represented as .

[0040] At each sampling moment, the data acquisition terminal needs to synchronously read the signals from each sensor; That is, the original data set collected at any valid sampling time includes: The original liquid level pressure difference signal represents the pressure difference between the inside and outside of the container at that moment; the gas phase temperature signal represents the real-time temperature of the gas space inside the container; the liquid phase temperature signal represents the real-time temperature of the liquid inside the container; the ambient temperature signal represents the real-time ambient temperature outside the container; and the external atmospheric pressure signal represents the real-time absolute ambient pressure.

[0041] To eliminate the slight time phase deviation that may exist during multi-channel analog-to-digital conversion, this method employs a strict hardware-triggered synchronization mechanism.

[0042] Traditional software polling methods can cause data misalignment at the millisecond or even tens of millisecond level in physical time due to processor interrupt response delays and the uncertainty of instruction execution time.

[0043] In subsequent dynamic time warping algorithms, this time misalignment is mistakenly identified as a morphological distortion, leading to serious computational errors.

[0044] The hardware timer inside the data acquisition terminal is in each When the cycle arrives, a nanosecond-level synchronous conversion pulse is broadcast to the sample-and-hold circuits of all analog-to-digital conversion channels, forcing all channels to freeze the analog signal at the same instant, and then quantization conversion is performed sequentially.

[0045] In the initial stage of continuous data collection, it is necessary to consider that the internal system of the container under test is in a state of strong thermodynamic non-equilibrium immediately after the filling, pressurization or sealing operation is completed.

[0046] For example, the pressurization process causes the gas temperature to rise sharply according to the first law of thermodynamics, and then the temperature gradually decreases as heat is dissipated to the environment; gases dissolved in liquids may precipitate due to pressure changes, or the gas may redissolve into the liquid. This gas-liquid mass transfer process is accompanied by complex volume and pressure changes.

[0047] If data is directly extracted for micro-leakage detection during such periods of drastic nonlinear fluctuations, the pressure difference fluctuations caused by thermodynamic changes will far exceed the pressure difference decreases caused by micro-leakage, rendering the detection results meaningless. This step introduces a steady-state adaptive detection mechanism based on a sliding data window.

[0048] A fixed-length time-sliding circular buffer is allocated in the memory of the data acquisition terminal. As the sampling process progresses, the latest sampled data is continuously moved into the buffer, while the oldest data is synchronously moved out.

[0049] At each sampling moment, the processor extracts the gas phase temperature signal and liquid phase temperature signal within the current sliding window, and calculates the rate of change of these two signals over the window time span.

[0050] The rate of change is calculated by taking the difference between the maximum and minimum temperature values ​​within the window and dividing it by the window duration. A pre-calibrated steady-state threshold is set; when both the gas phase temperature change rate and the liquid phase temperature change rate are continuous... When the sampling period is lower than the steady-state determination threshold, the system logically determines that the gas-liquid two-phase thermodynamic state inside the container has passed through the period of drastic transient change and entered a relatively gentle quasi-steady-state evolution period.

[0051] The parameters here It is a positive integer, and its specific value depends on the volume of the container, the thermal conductivity of the material, and the specific heat capacity of the liquid.

[0052] Once the system determines that the state has entered a quasi-steady state, it will officially mark the current moment as the effective detection start moment.

[0053] Starting from this initial moment, the system continuously collects and records data for a fixed length. The effective data sequence is used to form a multi-dimensional time-series data matrix for subsequent core algorithm processing. This matrix completely contains all synchronously sampled data from the start time to the end time of detection. To facilitate subsequent digital signal processing and matrix operations, the acquired discrete time-series data is organized into a one-dimensional vector form.

[0054] The specific definition is as follows: Effective pressure difference time series vector It contains A continuous set of raw differential pressure data points; vapor phase temperature vector Liquid phase temperature vector Ambient temperature vector ; and atmospheric pressure vector .

[0055] These vectors are aligned in memory and together form the complete data foundation characterizing the physical state evolution of a container over a specific time period.

[0056] In terms of the underlying mechanism for data storage and transmission, in order to prevent data loss or memory overflow during high-speed continuous sampling, the system adopts a double buffering technology at the underlying hardware level.

[0057] When the first memory buffer is full When a data point is filled, the hardware controller will automatically trigger a direct memory access (DMA) request to asynchronously move the filled data block to the main processor's computing memory space.

[0058] During DMA transfer, the data output stream of the analog-to-digital converter seamlessly switches to a second memory buffer to continue writing new sampled data. This mechanism ensures the absolute continuity of the timing signal on the time axis, providing a high-quality data source with no breaks and high fidelity for subsequent morphological preprocessing and theoretical sequence generation.

[0059] Through the rigorous synchronous acquisition architecture and steady-state determination logic described above, this step successfully extracted reliable multidimensional state parameters from the complex physical world, laying a solid data foundation for implementing complex digital signal processing algorithms.

[0060] Step 2, Dynamic generation of theoretical pressure difference decay sequence: Calculate dynamic gas volume based on environmental parameter vector, derive theoretical total absolute pressure by combining ideal gas law and Antoine equation, and subtract external atmospheric pressure signal to generate theoretical pressure difference decay sequence; The core task of this step is to dynamically construct a theoretical pressure difference decay sequence in the virtual space of a digital processor, based on the high-fidelity multidimensional environmental parameter time-series signal obtained in step one, using the basic physical laws of thermodynamics and fluid mechanics, assuming that the container is in an absolutely sealed and leak-free state.

[0061] In actual industrial settings, ambient temperature and atmospheric pressure are constantly fluctuating dynamically. These fluctuations act on the container under test through heat conduction and mechanical transmission, causing corresponding nonlinear changes in the pressure inside the container.

[0062] Using a fixed, static theoretical curve as a reference will fail to accurately reflect the impact of environmental disturbances, leading to significant spurious errors in subsequent comparisons. A dynamic theoretical sequence must be calculated and generated point-by-point based on real-time collected environmental data.

[0063] This theoretical sequence will serve as a benchmark template for subsequent dynamic time warping algorithms, and the accuracy of its generation directly determines the reliability of the final micro-leakage detection.

[0064] A thermodynamic equation of state model for the gas space inside the container to be tested needs to be established. Under the ideal assumption that the container is absolutely sealed and there is no loss of gas mass, the amount of gas inside the container remains constant.

[0065] According to the physical principles of the ideal gas law, at any effective sampling moment, there is a strict proportional relationship between the absolute pressure of the gas inside the container, the volume occupied by the gas, and the absolute temperature of the gas phase space.

[0066] That is, the product of the absolute pressure of a gas and the volume of the gas is equal to the product of the amount of gas, the ideal gas constant, and the absolute temperature of the gas phase.

[0067] In order to solve for the theoretical absolute pressure of the gas at each moment in digital space, the gas volume at the corresponding moment must be calculated precisely.

[0068] In a closed container containing liquid, the volume occupied by gas is not constant; it is equal to the total internal volume of the container minus the volume occupied by liquid.

[0069] In actual working conditions, both the solid outer shell material of the container and the liquid inside are affected by temperature changes, resulting in thermal expansion and contraction. This thermal expansion and contraction causes the total internal volume of the container and the volume of the liquid to become dynamic variables that change over time.

[0070] To accurately quantify this physical process, the volumetric expansion coefficient of the container material is introduced. and the volume expansion coefficient of liquid Let the initial calibration temperature be... Below, the total internal volume of the container is calibrated as follows: The initial volume calibration value of the liquid is .

[0071] At any sampling time (in From arrive (Integer index), considering the heating or cooling effect of ambient temperature on the container shell, establish a compensation calculation formula for the container's dynamic total volume: ; This formula indicates the real-time total volume of the container. It is based on the initial volume In addition to the ambient temperature Deviation from rated temperature The volume expansion or contraction is determined by the resulting volumetric expansion or contraction. Here, it is assumed that the expansion of the container material is isotropic, so the volumetric expansion coefficient can be approximated as three times the linear expansion coefficient.

[0072] Similarly, considering the influence of liquid phase temperature on liquid volume, a compensation calculation formula for liquid dynamic volume is established: ; This formula indicates the real-time volume of the liquid. It is based on the initial volume In addition to the liquid phase temperature Deviation from rated temperature The resulting volume change constitutes the volumetric expansion coefficient of the liquid. The coefficient of volumetric expansion of a liquid is typically much greater than that of a solid container; therefore, even small fluctuations in the liquid phase temperature can cause significant changes in the liquid volume, thereby substantially compressing or releasing the gas space. Based on these two dynamic volume variables, the true dynamic volume occupied by the gas inside the container at each moment can be calculated, i.e., using the real-time total volume. Subtract real-time liquid volume .

[0073] After obtaining the dynamic gas volume, it is also necessary to determine the amount of gaseous substance in the ideal gas law.

[0074] Since the amount of gaseous substance is constant under the no-leakage assumption, the system can effectively detect the start time (i.e., the index). The initial physical state of the container is calibrated once at the initial moment. At the initial moment, the initial raw pressure difference signal and the initial external atmospheric pressure signal are obtained through the sensor, and the actual absolute pressure inside the container at the initial moment is obtained by adding the two together.

[0075] Using the actual absolute pressure at the initial moment, the dynamic gas volume at the initial moment, and the gas phase temperature at the initial moment, the amount constant of the gas inside the container is derived by inversely applying the ideal gas law.

[0076] This calibration process firmly locks the initial pressurization state of the container into the mathematical model.

[0077] Substituting the calculated amount of substance constant back into the equation of state, we can derive the value at any given time. The theoretical absolute pressure of the dry gas inside is solely caused by temperature fluctuations and volume changes. However, the above calculations only consider the partial pressure contribution of the dry gas.

[0078] In a real-world closed container containing liquid, the gas phase space is typically in a state of gas-liquid two-phase equilibrium, meaning that the gas space is filled with saturated vapor produced by the evaporation of the liquid. According to physicochemical principles, the magnitude of the saturated vapor pressure is closely related only to the surface temperature of the liquid, and is independent of the volume of the gas phase space.

[0079] To improve the accuracy of the theoretical model, a dynamic compensation term for the saturated vapor pressure must be introduced; The system uses the physical logic of the Antoine equation to calculate the saturated vapor pressure. The Antoine equation is a mathematical model that describes the empirical relationship between the saturated vapor pressure of a pure liquid and its temperature.

[0080] The system processor calculates the saturated vapor pressure generated by liquid evaporation at each sampling moment based on pre-stored Antoine constants associated with specific liquid types and real-time acquired liquid phase temperature signals. Adding the previously calculated absolute pressure of the dry gas to this saturated vapor pressure yields the theoretical total absolute pressure inside the container under leak-free conditions.

[0081] Since the differential pressure sensor installed on site measures the difference between the pressure inside the container and the external atmospheric pressure, a reference conversion must be performed in order to generate a theoretical differential pressure sequence that is directly comparable to the actual collected differential pressure signal.

[0082] The system needs to subtract the synchronously acquired external atmospheric pressure signal from the calculated theoretical total absolute pressure.

[0083] In summary, at any given moment, all the physical processes described above Theoretical pressure difference The core derivation formula is: ; in, To determine the amount of gaseous substance obtained, Let be the ideal gas constant. For real-time gas phase temperature, and The dynamic volume calculated above, The saturated vapor pressure is calculated in real time. This represents the real-time external atmospheric pressure.

[0084] The system executes efficient loop calculation logic in a digital processor.

[0085] Let time index from Increase one by one to The processor reads the elements of the environmental parameter vectors collected in step one by one, substitutes them into the above physical model and core formula, and calculates the corresponding theoretical pressure difference value.

[0086] Finally, all the calculated theoretical pressure difference values ​​are arranged in strict chronological order in memory, generating a file of length [length missing]. Theoretical pressure drop decay sequence vector .

[0087] The theoretical sequence vector It profoundly implies the pressure difference evolution pattern that a container should exhibit if it remains absolutely sealed under the current specific and complex environmental parameter fluctuations.

[0088] It not only compensates for the nonlinear thermal expansion and contraction effect caused by ambient temperature through a dynamic volume formula, but also eliminates the measurement reference drift interference caused by external air pressure fluctuations by introducing a real-time atmospheric pressure signal.

[0089] Through this dynamic generation mechanism driven by multidimensional real-time data, this method successfully decouples complex environmental interference factors from the subsequent leakage judgment logic at the algorithm level, making the generated theoretical sequence a precise "digital ruler".

[0090] This provides a highly reliable benchmark for extracting extremely weak leakage features through morphological comparison in subsequent steps, fully demonstrating the deep integration of data processing and computational extrapolation in complex physical scenarios.

[0091] Step 3, Morphological preprocessing of differential pressure time series signal: Outlier removal and smoothing filtering are performed on the effective differential pressure time series vector to obtain the actual differential pressure sequence and residual sequence. The arithmetic mean and standard deviation are calculated, and standard score conversion is performed to obtain the preprocessed actual differential pressure sequence and standardized theoretical differential pressure sequence. The core task of this step is to process the raw liquid level and pressure difference time-series vector acquired in step one. Perform in-depth morphological cleaning and feature alignment.

[0092] In actual industrial sites or complex testing environments, the raw signals acquired by differential pressure sensors inevitably contain various physical interference components.

[0093] These interferences mainly come from three aspects: First, high-frequency vibration noise caused by on-site mechanical equipment (such as pump and valve operation, motor rotation), which manifests as dense spikes in the time domain; Second, low-frequency sloshing caused by accidental vibration of the liquid inside the container due to external factors will superimpose transient abnormal spikes on the pressure difference signal. Third, the zero-point low-frequency drift is determined by the physical characteristics of the sensor's own electronic components and the slow drift of the ambient temperature.

[0094] If the raw, unprocessed signal is used directly in subsequent morphological matching algorithms, the noise and outliers will severely distort the true geometry of the signal, causing huge computational biases in the matching algorithm.

[0095] It is necessary to filter out various interferences and achieve dimensional unification with the theoretical sequence by using a series of specific digital signal processing logics, while preserving the morphological characteristics of the slow voltage drop caused by micro-leakage.

[0096] For transient abnormal spikes caused by liquid sloshing or accidental external impacts, the system processor employs a nonlinear filtering algorithm based on absolute median difference logic for outlier detection and removal. Traditional mean filtering algorithms, when dealing with such large-amplitude transient spikes, distribute the energy of the spike evenly across surrounding data points, thus obscuring the true pressure difference step characteristics.

[0097] To overcome this deficiency, the system sets a fixed-width sliding data window in memory. This window moves along the original pressure differential time-series vector... The processor slides up point by point, extracts all data points contained within the window, sorts them, and finds the median of the data points within the window, denoted as . .

[0098] The processor calculates the median for each data point within the calculation window. The absolute deviation values ​​between the two values ​​are calculated, and these absolute deviation values ​​are sorted again to find the median of the absolute deviation.

[0099] Based on statistical principles, the system sets a threshold coefficient for judgment; for the data point at the center position of the current sliding window, the processor compares it with the median. Whether the absolute deviation is greater than the product of the above-mentioned threshold coefficient and the median of the absolute deviation.

[0100] If the value is greater than the product, the central data point is logically determined to be an abnormal spike caused by liquid sloshing. In this case, the processor forcibly replaces the value of the abnormal point with the median of the current window. ; If the value is not greater than the product, the point is considered to be a normal physical fluctuation, and its original value is retained.

[0101] By traversing the entire original sequence, the system obtains the intermediate sequence after removing outlier spikes. This median-based processing logic is able to preserve the true physical trend of the differential pressure signal's step changes and slow decay after eliminating abrupt noise.

[0102] To address high-frequency mechanical vibration noise, the system processor employs a smoothing filtering algorithm based on the principle of local polynomial least squares fitting for the intermediate sequence. Processing is required. In micro-leak detection, the local polynomial characteristics of the differential pressure signal (such as changes in the slope of the curve and convexity-concavity transitions) contain extremely important information about the leakage evolution.

[0103] Traditional moving average filtering, when used to smooth high-frequency noise, severely weakens signal peaks and blurs curve boundary features.

[0104] To address this issue, the system sets up a sliding window and constructs a low-order polynomial within it to approximate the data points. The processor solves a system of least-squares normal equations to find a set of polynomial coefficients that minimizes the sum of squared errors between the polynomial curve and the actual data points within the window. After solving, the fitted polynomial is used to calculate a smoothed value at the center of the window, which replaces the original noisy data.

[0105] As the window slides, the processor continuously repeats the polynomial fitting and center point calculation process described above, ultimately obtaining the smoothed pressure difference sequence. .

[0106] After this processing, high-frequency spikes in the signal were effectively eliminated, and the differential pressure curve exhibited a smooth trajectory that conformed to the laws of fluid mechanics.

[0107] Morphological feature alignment and dimensional uniformity are essential. In actual testing operations, due to differences in the initial pressurization state during each sealing operation, or slight height differences in the installation position of the differential pressure sensor, the actual collected and smoothed differential pressure sequence will vary. Compared with the theoretical pressure difference sequence generated in step two There is usually a static baseline deviation in absolute values. Since the core invention of this scheme lies in comparing the "degree of morphological distortion" of the two curves rather than the difference in absolute values, the influence of this static baseline deviation must be eliminated through mathematical transformation.

[0108] The system processor uses standard-splitting conversion logic to process the actual pressure differential sequence and the theoretical pressure differential sequence independently. For the smoothed actual pressure differential sequence... The processor first iterates through the entire sequence and calculates the arithmetic mean of all data points; then it calculates the sum of squares of the differences between each data point and the arithmetic mean, and finally finds the standard deviation of the sequence, denoted as . After obtaining the arithmetic mean and standard deviation, the processor subtracts the arithmetic mean from each data point in the sequence and divides the difference by the standard deviation. After this linear transformation, the preprocessed actual pressure difference time series vector is obtained. Similarly, the processor processes the theoretical pressure drop decay sequence generated in step two. Perform the exact same standard score transformation operation, calculate its own arithmetic mean and standard deviation, and convert each theoretical data point into a standard score to obtain the standardized theoretical pressure drop time series vector. .

[0109] After the above standardization process, the actual sequence and theoretical sequence All are mapped to a dimensionless mathematical space with a mean of zero and a variance of one.

[0110] At this point, the absolute height difference between the two curves on the vertical axis is eliminated, and what remains is purely their fluctuation pattern and relative decay trend on the time axis.

[0111] This preprocessing step eliminates the interference of initial static pressure differences on subsequent matching algorithms, allowing the subsequent dynamic time warping algorithm to focus on capturing morphological variations caused by micro-leakage.

[0112] Step 4, Dynamic Time Warping Sequence Matching Calculation: Construct a local distance matrix, calculate the cumulative distance matrix using the state transition equation under a banded constraint window, find the warped path by backtracking and calculate the normalized dynamic time warping distance; The core task of this step is to quantify the preprocessed actual pressure difference sequence. With standardized theoretical pressure difference sequence Morphological differences between them.

[0113] In complex container thermodynamic systems, changes in ambient temperature are transmitted to the container shell, then conducted from the shell to the internal gas, and finally cause a pressure difference response. This physical process exhibits significant thermal inertia.

[0114] This means that the actual pressure difference curve may be locally stretched, compressed, or delayed relative to the theoretical curve on the time axis. If the difference between the two curves at the same moment is calculated point by point using the traditional Euclidean distance, this nonlinear distortion on the time axis will cause the peaks or troughs that were originally similar in shape to be misaligned, resulting in huge spurious errors and serious misjudgments of micro-leakage.

[0115] To overcome the algorithmic bottleneck caused by this physical phenomenon, this step introduces the logic of dynamic time warping algorithm, which finds the shortest cumulative distance path between two sequences by performing elastic deformation on the time axis.

[0116] For ease of logical description, the preprocessed actual pressure difference sequence is denoted as the query sequence. The standardized theoretical pressure difference sequence is denoted as the reference sequence. Both sequences have a length of [missing information]. .

[0117] The system processor constructs a dimension in memory. Local distance matrix Each element in this matrix Represents the query sequence The first in Data points With reference sequence The first in Data points The local spatial distance between them.

[0118] To highlight subtle morphological differences, the processor uses squared Euclidean distance as a metric for local distance, which is calculated as the square of the difference between the values ​​of two data points. The formula is expressed as follows: The matrix The sequence was recorded in its entirety. and sequence The degree of morphological difference between any two data points.

[0119] To prevent the algorithm from exhibiting excessive time distortions that defy physical laws when searching for matching paths, global path constraints must be introduced. In real thermodynamic systems, the time delay caused by temperature conduction is finite, and infinitely long lags are impossible.

[0120] The system uses a strip constraint window to limit the search range of the matching path. The width parameter of the constraint window is set to... ( (This is a positive integer, and its value is pre-calibrated based on the thermal time constant of the container under test and the sampling frequency).

[0121] When calculating cumulative distance, the system requires matching points. The following core constraints must be met: If the absolute value of the difference in time indices Larger than window width If the system deems such a time-span matching impossible in physics, the processor will directly assign an extremely large penalty value (such as infinity) to the local distance of that location.

[0122] This constraint not only significantly reduces the time complexity of the algorithm in the processor, but more importantly, it eliminates ill-conditioned matching paths and ensures the physical rationality of morphological alignment.

[0123] Based on the local distance matrix and global constraints, the system processor constructs the cumulative distance matrix. .matrix The size is also Its elements Indicates starting from the origin Reaching the current matching point The processor calculates the distance of the path with the smallest cumulative distance among all possible paths. It uses dynamic programming to recursively solve for the matrix. .

[0124] First, initialize the boundary conditions, setting the starting point... Set the value to And assign maximum values ​​to the paths in the first row and first column that do not conform to physical logic.

[0125] The processor calculates the matrix one by one using the core state transition equations, following either row-major or column-major order. The elements in.

[0126] For satisfying the constraints point Its cumulative distance The calculation formula is: ; This state transition equation profoundly reflects the physical meaning of elastic regularity of the time axis: reaching the current matching point The shortest path must come from one of its three adjacent preceding states.

[0127] in, This represents synchronization on the time axis, meaning that there is no relative time delay between the actual signal and the theoretical signal; Represents the query sequence Lagging behind the reference sequence in time This corresponds to the physical phenomenon that the actual pressure difference response is relatively slow; Represents the query sequence It is ahead of the reference sequence in time. .

[0128] By selecting the preceding state with the minimum cumulative distance at each step, the algorithm can achieve this within the allowed time window. Internally, the time axis is adaptively stretched or compressed to maximize the alignment of the morphological features of the two curves.

[0129] When the dynamic programming algorithm has traversed the entire... After the matrix, the element in the bottom right corner of the matrix This represents a sequence. and sequence The shortest cumulative distance globally after elastic time warping.

[0130] To obtain the specific morphological distortion trajectory, the processor needs to start from the bottom right corner of the matrix. Initially, the shortest cumulative distance path is found through backtracking operations, and denoted as the normalized path. Regularized path It is a sequence of matrix coordinate points, and the length of the path is denoted as . .

[0131] The logic of the backtracking process is as follows: Starting from the current node, in the matrix The process involves comparing the values ​​of the three preceding nodes, selecting the node with the smallest value as the next node in the path, and repeating this process until the path returns to the starting point. .

[0132] To eliminate sequence length and regular path length The processor will use the global shortest cumulative distance to determine the final distance metric. Divide by the total length of the regular path Calculate the normalized dynamic time warp distance .

[0133] The normalized distance As a highly condensed scalar feature, it precisely quantifies the degree of pure morphological distortion of the actual differential pressure sequence relative to the theoretical leak-free sequence after excluding time delay interference.

[0134] If the container is sealed, the shape of the actual sequence will be highly consistent with the theoretical sequence, and the regularized path will closely follow the diagonal of the matrix. The value will approach zero; Conversely, if micro-leakage exists, the actual differential pressure curve will exhibit an additional attenuation trend, leading to irreversible morphological distortion. The value will increase.

[0135] Through this sequence matching calculation based on dynamic programming, this step successfully transforms complex time-series signal differences into a reliable quantitative indicator.

[0136] Step 5, Distortion path quantification and micro-leakage determination: Calculate the path deviation area of ​​the regularized path, generate an adaptive threshold by combining the variance of the residual sequence and the variance of the ambient temperature signal, calculate the equivalent pressure drop value using the normalized dynamic time regularization distance and standard deviation and convert it into the equivalent volume leakage rate, and output the micro-leakage detection result when the determination condition is met. The core task of this step is to rigorously map the purely mathematical dynamic time-warped distance and warped path calculated in step four back to the actual leakage state in the physical world. In complex industrial detection scenarios, relying solely on a scalar distance value to determine micro-leakage has limitations. Residual environmental noise, inherent sensor measurement uncertainties, and the minute creep of the container material under pressure all lead to fundamental morphological differences between the actual pressure difference sequence and the theoretical sequence.

[0137] If a fixed threshold is used, false alarms are likely to occur in harsh environments or false alarms may occur in stable environments.

[0138] This step involves deeply analyzing the geometric topological features of regular paths to construct a dynamically adaptive judgment threshold model, and deriving the equivalent physical leakage rate by combining the gas state equation. Finally, it outputs high-confidence micro-leakage detection results through multi-dimensional comprehensive logic.

[0139] The system processor performs in-depth geometric topological feature extraction on the regular path obtained in step four and calculates the path deviation area.

[0140] In the matrix space of the dynamic time warping algorithm, if the actual pressure difference sequence and the theoretical pressure difference sequence are highly consistent in shape and have no time delay, then the warping path will closely fit the main diagonal of the matrix space.

[0141] The main diagonal represents absolute synchronization on the time axis. When a micro-leak occurs in the container, the actual pressure differential curve will show a continuous and irreversible additional decay.

[0142] To morphologically match this decay, the dynamic time warping algorithm is forced to align data points from later moments in the theoretical sequence with those from earlier moments in the actual sequence. This causes the warping path to systematically bend to one side in the matrix space. The degree of this bending directly implies the time-cumulative effect of leakage evolution.

[0143] To quantify this curvature, the processor defines the main diagonal of the matrix space in memory. For any node on the regularized path, the processor extracts its corresponding actual sequence time index and theoretical sequence time index, and calculates the orthogonal geometric distance from that node to the main diagonal.

[0144] The specific calculation logic is as follows: Calculate the absolute value of the difference between the two time indices and divide it by the square root of two. The processor integrates and sums the orthogonal distances from all nodes on the path to the main diagonal, and divides this sum by the total length of the normalized path to obtain the normalized path deviation area feature. This feature can extremely sensitively capture persistent morphological distortions caused by micro-leakage. Compared to simple cumulative distance, path deviation area is insensitive to local path jitter caused by random noise, but has a high amplification effect on global path offset caused by leakage.

[0145] A dynamic and adaptive decision threshold generation model is constructed. To overcome the shortcomings of fixed thresholds, the system processor needs to calculate a specific decision threshold in real time based on the actual environmental noise level and thermodynamic fluctuations within the current detection cycle.

[0146] The processor extracts the residual sequence generated in step three during the smoothing filtering process. This residual sequence is obtained by subtracting the smoothed pressure difference sequence from the original pressure difference sequence, and it represents the combined level of high-frequency electrical noise and mechanical vibration noise from the sensor in the current detection environment. The processor calculates the variance of this residual sequence, denoted as . .

[0147] The processor extracts the ambient temperature vector collected in step one, calculates its variance over the effective detection period, and denotes it as... This variance represents the degree of instability of the thermodynamic state of the external environment.

[0148] Based on the two dynamic parameters mentioned above, the system processor uses a preset linear combination model to calculate the adaptive threshold of the dynamic time warp distance. Adaptive threshold for path deviation area ; The calculation formula is as follows: ; In the above formula, and This is the noise penalty coefficient, used to quantify the interference weight of high-frequency vibration on shape matching; and This is the temperature fluctuation penalty coefficient, used to quantify the interference weight of thermodynamic instability on the differential pressure reference. and This is the inherent basic tolerance constant of the system, representing the normal measurement error allowed by the system under ideal quiet conditions.

[0149] These coefficients were all calibrated through multiple benchmark tests on leak-free standard containers. This dynamic threshold mechanism automatically raises the threshold to prevent false alarms when the environment is harsh (e.g., high mechanical vibration or drastic temperature fluctuations); and automatically lowers the threshold to improve sensitivity to even the smallest leaks when the environment is stable.

[0150] After obtaining the morphological features and dynamic thresholds, in order to meet the intuitive requirements of industrial sites for leakage levels, the dimensionless morphological differences must be inversely mapped to the equivalent volumetric leakage rate in the physical world.

[0151] The processor uses the inverse operation logic of the standard score conversion in step three to restore the normalized dynamic time warp distance to an equivalent pressure drop value with pressure physical dimensions. .

[0152] The specific operation is as follows: take the square root of the normalized dynamic time-normalized distance and multiply it by the standard deviation of the actual pressure difference sequence defined in step three. .

[0153] The equivalent pressure drop value This represents the pressure drop caused solely by the loss of gas mass inside the container, after excluding interference from ambient temperature and atmospheric pressure.

[0154] Based on the differential form of the ideal gas law, the processor converts the equivalent pressure drop into the equivalent volumetric leakage rate under standard conditions. .

[0155] Let the effective detection period span be... During the detection, the processor calculates the average volume of gas inside the container. (Obtained from the arithmetic mean of the dynamic gas volume sequence in step two), and the average gas phase temperature. .

[0156] Based on gas leakage dynamics, the equivalent volume leakage rate The core physics mapping formula is: ; in, This is the international standard reference temperature (usually taken as 293.15K). The international standard reference atmospheric pressure (usually taken as 101325 Pa) is used.

[0157] This formula rigorously converts the pressure deviation obtained from morphological matching into a physical leakage rate index that conforms to industrial standards, eliminating the influence of different geographical altitudes and seasonal temperatures on leakage rate calibration.

[0158] The system processor executes multi-dimensional comprehensive judgment logic and outputs the final detection result. The processor compares the extracted features with the dynamic threshold using Boolean logic.

[0159] The system sets three independent conditions for micro-leakage detection: Condition A is that the normalized dynamic time warp distance is greater than its adaptive threshold. Condition B is that the path deviation area is greater than its adaptive threshold. Condition C is the equivalent volumetric leakage rate. It exceeds the preset industrial qualification standard limit.

[0160] The system employs a multi-concurrency verification mechanism: The system only formally determines that the container under test has a microleakage state when conditions A, B, and C are all met simultaneously. This multi-condition intersection judgment logic effectively eliminates the risk of misjudgment caused by a single feature quantity exceeding its limit due to accidental interference. If a microleakage is determined to exist, the system further calculates the confidence score of the detection result.

[0161] The confidence score is calculated based on the relative magnitude by which the feature exceeds the threshold. The processor calculates the proportion of the dynamic time warp distance exceeding the threshold and the proportion of the path deviation area exceeding the threshold, respectively. The two are added together and then fed into a nonlinear saturation function (such as an inverse exponential function) to map it to a score between 0 and 100.

[0162] The higher the score, the more significant the leakage characteristics and the stronger the reliability of the detection results.

[0163] After completing all calculations and judgments, the system processor outputs a structured detection report data packet to an external host computer or control terminal via an industrial communication interface (such as RS485 bus or industrial Ethernet). This data packet includes: the final leakage status flag (e.g., 0 for a successful seal, 1 for a leak), and the quantified equivalent volumetric leakage rate. Confidence scores and dynamic threshold records for later retrospective analysis.

[0164] Thus, this method completes a rigorous digital data processing workflow, from synchronous acquisition of multi-dimensional data, dynamic reconstruction of theoretical models, morphological cleaning of time-series signals, and dynamic time warping elastic matching, all the way to the final quantization of topological features and mapping of physical leakage rates.

[0165] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0166] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0167] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0168] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0169] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A precise detection method for micro-leakage of liquid based on liquid level pressure difference, characterized in that, include: Acquire the original liquid level and pressure difference signal and various environmental parameter signals of the container under test. After determining that the internal thermodynamic state of the container has entered the quasi-steady state period, determine the effective detection start time and record the effective pressure difference time series vector and environmental parameter vector. The dynamic gas volume is calculated based on environmental parameter vectors, and the theoretical total absolute pressure is derived by combining the ideal gas law and the Antoine equation. The theoretical pressure difference attenuation sequence is generated by subtracting the external atmospheric pressure signal. Outlier removal and smoothing filtering are performed on the effective pressure difference time series vector to obtain the actual pressure difference sequence and residual sequence. The arithmetic mean and standard deviation are calculated, and standard score conversion is performed to obtain the preprocessed actual pressure difference sequence and standardized theoretical pressure difference sequence. Construct a local distance matrix, calculate the cumulative distance matrix using the state transition equation under a banded constraint window, find the regularized path by backtracking and calculate the normalized dynamic time regularized distance; The path deviation area of ​​the regularized path is calculated, and an adaptive threshold is generated by combining the variance of the residual sequence and the variance of the ambient temperature signal. The equivalent pressure drop value is calculated using the normalized dynamic time regularization distance and standard deviation and converted into the equivalent volume leakage rate. When the judgment condition is met, the micro-leakage detection result is output.

2. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, The process of acquiring the original liquid level and pressure difference signal and various environmental parameter signals of the container under test, and determining whether the internal thermodynamic state of the container has entered the quasi-steady state period is as follows: A differential pressure sensor and a multi-point temperature sensor array are configured in the container under test to acquire gas phase temperature signal, liquid phase temperature signal, ambient temperature signal and external atmospheric pressure signal as various environmental parameter signals; A hardware timer is used to broadcast a synchronization conversion pulse to the analog-to-digital conversion channel, which simultaneously freezes the analog signal and performs quantization conversion to eliminate time phase deviation during multi-channel acquisition. Within a time sliding window, the difference between the maximum and minimum values ​​of the gas phase temperature signal and the liquid phase temperature signal is extracted respectively, and divided by the time length of the time sliding window to obtain the rate of change of the gas phase temperature signal and the rate of change of the liquid phase temperature signal. When the rate of change of both the gas phase temperature signal and the rate of change of the liquid phase temperature signal are lower than the steady-state determination threshold for multiple consecutive sampling periods, it is determined that the thermodynamic state inside the container has passed the transient change period and entered the quasi-steady-state period.

3. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, Dynamic gas volume is calculated based on environmental parameter vectors, using the following specific calculation method: ； ； in, Sampling time, This refers to the total dynamic volume of the container. Initial calibration temperature The internal total volume calibration value of the lower container. Let be the coefficient of volumetric expansion of the container material. The ambient temperature signal is part of the environmental parameter vector. For the dynamic volume of the liquid, Initial calibration temperature The initial volume calibration value of the liquid below. is the coefficient of volumetric expansion of the liquid. The liquid phase temperature signal in the environmental parameter vector; Before calculating the dynamic gas volume, an initial calibration temperature is set, and the calibration values ​​of the total internal volume of the container and the initial volume of the liquid are obtained at the initial calibration temperature. By introducing the volume expansion coefficient of the container material and the volume expansion coefficient of the liquid, the thermal expansion and contraction effect caused by the fluctuation of the ambient temperature signal and the liquid phase temperature signal is quantified. The dynamic gas volume is obtained by subtracting the dynamic volume of the liquid from the total dynamic volume of the container.

4. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 3, characterized in that, The process of generating the theoretical differential pressure decay sequence includes the following formula: ； in, This represents the theoretical pressure difference value in the theoretical pressure difference decay sequence. For the amount of gaseous substance, Let be the ideal gas constant. The gas phase temperature signal is part of the environmental parameter vector. The saturated vapor pressure was calculated using the Antoine equation combined with the liquid phase temperature signal. This refers to the external atmospheric pressure signal in the environmental parameter vector; Before using the formula to derive the theoretical total absolute pressure, the amount constant of the gas substance sealed inside the container is deduced by effectively detecting the actual absolute pressure, dynamic gas volume and gas phase temperature signal at the initial moment. In the derivation process, the saturated vapor pressure generated by liquid evaporation is calculated point by point to compensate for the vapor partial pressure under the gas-liquid two-phase equilibrium state.

5. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, The logic for outlier removal from the effective pressure difference time series vector is as follows: Set a fixed-width sliding data window. As the sliding data window slides point by point on the effective pressure difference time series vector, sort the data points contained in the sliding data window and find the median. Calculate the absolute deviation between each data point in the sliding data window and the median, and then find the median of the absolute deviations. When the absolute deviation between the data point at the center of the sliding data window and the median is greater than the product of the judgment threshold coefficient and the median of the absolute deviation value, the data point at the center is determined to be an abnormal spike data caused by liquid sloshing, and the abnormal spike data is forcibly replaced with the median to preserve the step change trend of the differential pressure signal.

6. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, The logic for smoothing filtering and standard score conversion is as follows: A low-order polynomial is constructed within a sliding window. The polynomial coefficients are found by solving the least squares normal equations. The smoothed value at the center of the window is calculated using the fitted polynomial to replace the noisy data, thus obtaining the actual pressure difference sequence after filtering out high-frequency mechanical vibration noise. Subtracting the actual pressure difference sequence from the effective pressure difference time series vector yields the residual sequence. Calculate the arithmetic mean and standard deviation of the actual pressure difference sequence, and subtract the arithmetic mean from each data point in the actual pressure difference sequence and the theoretical pressure difference decay sequence, and divide by the standard deviation. Map the result to a dimensionless mathematical space with a mean of zero and a variance of one, in order to eliminate the static reference deviation caused by different initial pressurization states.

7. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, The process of calculating the cumulative distance matrix using the state transition equation under a banded constraint window includes the following state transition equation: ； ； Among them, the matching point The constraints of the strip constraint window must be satisfied. , This is the width parameter of the strip constraint window; For elements in the cumulative distance matrix, These are elements in the local distance matrix. , and These are the cumulative distances of three adjacent preceding states; Before recursively solving the state transition equation, the boundary conditions are initialized. The cumulative distance of the starting position is set as the local spatial distance of the starting position, and the path that does not conform to the physical logic is assigned a maximum value. The three pre-states in the state transition equation represent the synchronous matching on the time axis, the actual pressure difference response lag, and the actual pressure difference response lead, respectively.

8. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, Calculating the area of ​​deviation from a regular path specifically includes: Define the main diagonal in the matrix space to represent absolute synchronization of the time axis; Extract the actual sequence time index and the theoretical sequence time index corresponding to any node on the regularized path; Calculate the absolute value of the difference between the actual sequence time index and the theoretical sequence time index and divide it by the square root of two to obtain the orthogonal geometric distance from the node to the main diagonal of the matrix space; Integrate and sum the orthogonal geometric distances from all nodes on the regular path to the main diagonal, and divide the sum by the total length of the regular path to obtain the path deviation area that quantifies the degree of continuous morphological distortion and bending caused by microleakage.

9. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, The process of generating an adaptive threshold by combining the variance of the residual sequence and the variance of the ambient temperature signal uses the following calculation method: ； ； in, An adaptive threshold for normalized dynamic time warping distance. An adaptive threshold for the path deviation area. Let V be the variance of the residual sequence. Let V be the variance of the ambient temperature signal in the ambient parameter vector. and This is the noise penalty factor. and This is the temperature fluctuation penalty coefficient. and The basic tolerance constant; The variance of the residual sequence obtained in the morphological preprocessing step of the differential pressure time series signal is calculated to characterize the combined level of high-frequency electrical noise and mechanical vibration noise of the sensor in the detection environment, thereby realizing dynamic adjustment of the threshold increase when the environment is harsh and the threshold decrease when the environment is stable.

10. The precise detection method for liquid micro-leakage based on liquid level pressure difference according to claim 1, characterized in that, The content converted into equivalent volume leakage rate includes the following formula: ； in, Equivalent volume leakage rate, This is the equivalent pressure drop value. The average volume of the dynamic gas volume. The effective detection period is the time span. For standard reference temperature, The average gas phase temperature is the gas phase temperature signal in the environmental parameter vector. Standard reference atmospheric pressure; Before using the formula transformation, the square root of the normalized dynamic time-normalized distance is multiplied by the standard deviation of the actual pressure difference sequence to restore it to the equivalent pressure drop value with pressure physical dimensions, so as to characterize the pressure drop caused purely by the loss of gas mass inside the container after excluding the interference of ambient temperature and atmospheric pressure.

Images