Construction and application of pressure and displacement dual-channel leakage rate calculation model

Abstract

The invention discloses construction and application of a pressure and displacement dual-channel leakage rate calculation model, and the method comprises the steps: dividing a leakage risk region of target equipment, and synchronously collecting pressure and displacement real-time data to obtain dual-source monitoring data; performing time-frequency combined processing on the double-source data to generate feature associated data; establishing a signal-leakage response mapping model according to the characteristic association data and the structure data, analyzing leakage medium diffusion characteristics and tracking a path, and generating propagation dynamic data; analyzing the leakage rate influence factors to generate rate influence characteristic data, defining an influence range, performing multi-physics field coupling calculation, and generating field domain coupling data; and constructing a two-channel model to predict the leakage rate, identifying a leakage stage based on prediction data, correcting monitoring parameters, and generating monitoring control feedback data. The model is applied to an industrial equipment leakage dynamic monitoring system, the leakage rate can be accurately calculated, dynamic monitoring is achieved, and the monitoring accuracy and real-time performance are improved.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment leakage monitoring technology, specifically to the construction and application of a dual-channel leakage rate calculation model based on pressure and displacement. Background Technology

[0002] In industrial production, leakage from various equipment has always been a significant factor affecting production safety, efficiency, and environmental protection. As industrial equipment becomes increasingly larger, more complex, and more intelligent, the safety accidents, economic losses, and environmental pollution that equipment leaks can cause are becoming increasingly prominent. Therefore, accurate monitoring and effective control of equipment leaks have become crucial.

[0003] Traditional leak detection methods mostly rely on monitoring a single parameter, such as pressure or displacement. This single-parameter approach has significant limitations. Firstly, a single parameter cannot fully reflect the complex characteristics of the leak process, easily leading to inaccurate monitoring results, false alarms, or missed alarms. For example, judging a leak solely based on pressure changes may be affected by pressure fluctuations caused by other factors during equipment operation, thus failing to accurately identify the actual leak. Secondly, traditional methods often lack precise calculations of the leak rate, failing to provide strong data support for timely leak handling and risk assessment. In practical applications, without accurately knowing the leak rate, it is difficult to determine the severity and development trend of the leak, thus affecting the timeliness and accuracy of decision-making.

[0004] Furthermore, existing leak monitoring models have shortcomings in handling multiphysics coupling problems. Equipment leakage processes typically involve the interaction of multiple physical fields, such as pressure fluctuations, displacement deformation, and media properties. Traditional models struggle to effectively integrate these factors, resulting in inaccurate analysis of leak propagation dynamics and an inability to comprehensively track leak paths and assess the extent of leak impact. Simultaneously, traditional models exhibit poor adaptability to complex equipment structures and operating conditions, making it difficult to flexibly adjust and optimize them according to the characteristics of different equipment and actual operating conditions, thus significantly limiting monitoring effectiveness.

[0005] With the continuous development of sensor and data processing technologies, the application of various types of sensors has provided new possibilities for leak monitoring. However, how to effectively integrate multi-source monitoring data and construct an accurate leak rate calculation model remains a problem to be solved. Furthermore, how to efficiently process and analyze the large amounts of real-time data generated during the operation of industrial equipment to achieve dynamic prediction of leak rates and real-time correction of monitoring parameters is also a significant challenge facing the field of leak monitoring.

[0006] Existing leak monitoring technologies have shortcomings in terms of accuracy, comprehensiveness, adaptability, and multi-physics coupling processing, failing to meet the high-precision requirements of modern industrial equipment for leak monitoring. Therefore, there is an urgent need for a new leak rate calculation model and method that can comprehensively consider multiple factors, accurately calculate the leak rate, and achieve dynamic monitoring and real-time feedback. Summary of the Invention

[0007] The purpose of this invention is to provide a construction and application of a leakage rate calculation model with dual channels of pressure and displacement, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for constructing a leakage rate calculation model with dual channels of pressure and displacement, the method comprising:

[0009] A1: Divide the target equipment into leakage risk areas and obtain equipment structure data; determine key monitoring locations based on the equipment structure data, and deploy a multi-type sensor acquisition network;

[0010] A2: Perform signal stability optimization processing on the multi-type sensor acquisition network, and simultaneously acquire real-time data of pressure and displacement signals to obtain dual-source monitoring data; perform time-frequency joint processing on the dual-source monitoring data based on pressure amplitude, displacement frequency and signal phase to generate feature correlation data;

[0011] A3: Establish a signal-leakage response mapping model based on feature correlation data and equipment structure data; analyze the diffusion characteristics of the leakage medium based on the signal-leakage response mapping model, and conduct spatially related leakage path tracing to generate dynamic data on leakage propagation;

[0012] A4: Based on the dynamic data of leakage propagation and the dual-source monitoring data, conduct correlation analysis of factors affecting the leakage rate to generate rate influence characteristic data; based on the rate influence characteristic data, define the scope of leakage influence to generate leakage action area data; based on the leakage action area data, perform multi-physics field coupling calculations based on pressure fluctuations, displacement deformation, and medium properties to generate field coupling data.

[0013] A5: Construct a dual-channel calculation model based on field coupling data, and use the dual-channel calculation model to dynamically predict the leakage rate and generate rate prediction data; based on the rate prediction data, perform real-time correction of monitoring parameters to obtain leakage monitoring and control feedback data.

[0014] Preferably, A1 includes the following steps:

[0015] A11: Perform multi-location structural detection on the target equipment to obtain material sample data of the equipment;

[0016] A12: Identify the component types, connection methods, and wall thicknesses from the equipment material sample data to obtain the equipment structure data;

[0017] A13: Based on the equipment structure data, perform mechanical property analysis on the equipment material sample data based on elastic modulus, yield strength and Poisson's ratio to obtain the component mechanical property data;

[0018] A14: Analyze the flow characteristics of the medium based on the mechanical performance data of the components to obtain the flow characteristic data of the medium. The flow characteristic analysis of the medium includes the observation of the flow velocity inside the equipment, the permeability test of the material sample data of the equipment, and the analysis of the medium distribution based on the flow velocity and permeability.

[0019] A15: Combine media flow characteristic data, component mechanical performance data, and equipment construction data into equipment structure data;

[0020] A16: Determine key monitoring locations based on equipment structure data and deploy a multi-type sensor acquisition network.

[0021] Preferably, A16 includes the following steps:

[0022] A161: Based on the equipment structure data, perform abrupt change detection on the characteristics of the component connection interface to obtain interface abrupt change point data. Specifically, abrupt change detection involves identifying the location of significant changes in component type and mechanical properties.

[0023] A162: Based on the equipment structure data, the stress state of the components is sensitively assessed to identify areas of large deformation and obtain data on deformation-sensitive areas. The deformation sensitivity assessment includes analysis of the resistance to deformation based on elastic modulus and yield strength, assessment of the degree of deformation based on wall thickness and Poisson's ratio, and assessment of the scouring characteristics of the permeable medium.

[0024] A163: Determine the spatial distribution of key monitoring points based on preset equipment operating conditions and monitoring requirements data, interface mutation point data, and deformation-sensitive area data to obtain monitoring location layout data;

[0025] A164: Select sensor types based on the data from the monitoring locations, and deploy a multi-type sensor acquisition network including pressure sensors, displacement sensors, strain sensors, and flow velocity sensors.

[0026] Preferably, A2 includes the following steps:

[0027] A21: Perform stability optimization processing based on signal filtering and noise isolation on the multi-type sensor acquisition network, and acquire real-time signals during equipment operation to obtain dual-source monitoring data;

[0028] A22: The pressure signal in the dual-source monitoring data is segmented based on time series to establish a pressure-time correlation curve;

[0029] A23: Calculate the frequency characteristics and amplitude distribution of pressure fluctuations based on the pressure-time correlation curve, and perform correlation analysis between pressure fluctuations and location to obtain pressure-location evolution characteristic data;

[0030] A24: Based on the dual-source monitoring data, establish the relationship curve between displacement frequency and position, and calculate the frequency change rate at different position segments to obtain frequency spatiotemporal evolution characteristic data;

[0031] A25: Real-time changes in signal phase are recorded based on dual-source monitoring data, and correlation characteristic analysis of signal phase and position is performed to obtain phase spatiotemporal evolution characteristic data.

[0032] A26: Based on pressure-position evolution characteristic data, frequency spatiotemporal evolution characteristic data, and phase spatiotemporal evolution characteristic data, the comprehensive impact of signal combinations on leakage rate is analyzed, and the critical values ​​and effective ranges of key signal combinations are identified, thereby generating time-frequency evolution characteristic data of signal combinations.

[0033] A27: Establish a dynamic evolution model of a two-dimensional signal space based on time-frequency evolution characteristic data, and extract feature changes to obtain feature correlation data.

[0034] Preferably, A21 includes the following steps:

[0035] The stability optimization process based on signal filtering and noise isolation is performed on the multi-type sensor acquisition network, and real-time signals during equipment operation are collected to obtain dual-source monitoring data. The signal filtering includes digital filtering algorithm processing of sensors, interference signal shielding of signal cables, and normalization calibration of sensor signals. The noise isolation includes setting up vibration isolation brackets, using flexible connection components, and optimizing the noise source avoidance of sensor installation positions.

[0036] Preferably, A27 includes the following steps:

[0037] A271: Construct a two-dimensional signal space with pressure amplitude, displacement frequency, and signal phase as coordinate axes based on time-frequency evolution characteristic data, thereby obtaining signal space coordinate data;

[0038] A272: Establish characteristic change trajectories based on signal spatial coordinate data, and construct characteristic motion trajectory curves through spatial mapping of time-series sampling points, thereby obtaining characteristic trajectory data;

[0039] A273: Perform location-based hierarchical processing on feature trajectory data to identify the characteristic change characteristics of different location segments, including change trends, change rates, and interrelationships, thereby obtaining hierarchical characteristic data;

[0040] A274: Construct a dynamic model of a two-dimensional signal space based on hierarchical characteristic data, and establish a continuous expression of feature changes through spatial interpolation and numerical fitting to obtain dynamic evolution model data;

[0041] A275: Extract features from dynamic evolution model data based on gradient characteristics, curvature characteristics, and frequency characteristics of feature changes to obtain model characteristic data;

[0042] A276: Based on model characteristic data, perform signal correlation analysis based on the coupling relationship and feedback mechanism between signals to obtain signal correlation data;

[0043] A277: Extract indicators that characterize the dynamic changes of features based on model characteristic data and signal correlation data, thereby generating feature correlation data.

[0044] Preferably, A3 includes the following steps:

[0045] A31: Based on the feature association data and equipment structure data, establish a signal-leakage characteristic correspondence table that includes the mapping relationship between signal changes and leakage response in different components, thereby obtaining signal response data;

[0046] A32: Perform data standardization and feature dimensionality reduction on the signal response data, and establish a mathematical model of the signal-leakage response to obtain the response model data;

[0047] A33: Based on response model data, machine learning training is performed to construct a nonlinear mapping relationship between signal and leakage response, thereby obtaining a signal-leakage response mapping model;

[0048] A34: Based on the signal-leakage response mapping model, the diffusion characteristics of the leaking medium, including pressure transmission, medium penetration and leakage diffusion, are analyzed to obtain diffusion characteristic data;

[0049] A35: Based on diffusion characteristic data, spatially related leakage path calculations are performed to obtain leakage path data. The leakage path calculation includes the influence analysis of the pressure decay law of the leakage medium and the deformation resistance of the components.

[0050] A36: Identify key characteristics and abrupt change points in the leakage propagation dynamic data to obtain leakage propagation dynamic data.

[0051] Preferably, A4 includes the following steps:

[0052] A41: Based on the dynamic data of leakage propagation and the dual-source monitoring data, the influence of different signal combinations on the leakage rate is analyzed to obtain the rate influence data;

[0053] A42: Perform a leakage rate performance evaluation on the rate impact data based on the rate change law, leakage duration characteristics, and deformation propagation characteristics to obtain performance evaluation data;

[0054] A43: Analyze the evolution characteristics of leakage rate over time from the performance evaluation data to obtain rate impact characteristic data;

[0055] A44: Numerical simulation of the leakage influence range is performed based on the rate influence characteristic data, and a multi-field coupled analysis model including pressure field, displacement field and flow field is established to obtain the influence range data;

[0056] A45: Perform boundary identification and spatial partitioning on the impact range data to obtain the leakage effect area data;

[0057] A46: Perform multi-physics field coupling calculations based on pressure fluctuations, displacement deformation, and medium properties according to the data of the leakage area, identify the field evolution law and critical state characteristics, and generate field coupling data.

[0058] Preferably, A5 includes the following steps:

[0059] A51: Extract features from field-coupled data and construct a training sample set for a dual-channel computation model to obtain training sample data, which includes input features and rate labels.

[0060] A52: Construct a dual-channel computational network structure based on a hybrid architecture of feedforward neural network and recurrent neural network according to the training sample data, and train the model to obtain the rate prediction model.

[0061] A53: Optimize the prediction accuracy of the rate prediction model using a cross-validation method to obtain the computational model data;

[0062] A54: Dynamically predict the leakage rate using computational model data to obtain rate prediction data, which includes instantaneous rate, cumulative amount, and spatial distribution.

[0063] A55: The rate prediction data is analyzed to identify the key stages and rate mutation characteristics in the leakage formation process, thereby obtaining rate evolution data. The key stages include the initial leakage stage, the diffusion development stage, the stable leakage stage, the rate decay stage, and the leakage termination stage. In the initial stage, the pressure sampling frequency is increased to 5Hz. In the diffusion stage, displacement sensor points are added. In the stable stage, the medium parameter update cycle is shortened to 5 minutes. In the decay stage, the pressure sensor zero point is calibrated every hour. In the termination stage, the backup sensor cross-validation is initiated.

[0064] A56: Based on the rate evolution data, the monitoring parameters at each stage are corrected in real time to obtain leakage monitoring and control feedback data.

[0065] Preferably, the present invention also includes the application of the pressure and displacement dual-channel leakage rate calculation model constructed by the above method in the dynamic monitoring system for leakage of industrial equipment.

[0066] Compared with the prior art, the beneficial effects of the present invention are:

[0067] By dividing the target equipment into leakage risk zones, acquiring equipment structural data, and determining key monitoring locations accordingly, a multi-type sensor network is deployed to achieve comprehensive and accurate monitoring of equipment leaks. This monitoring location deployment method based on equipment structure and mechanical performance analysis can accurately capture key areas where equipment leaks may occur, avoiding the blindness of traditional monitoring methods and greatly improving the effectiveness and targeting of monitoring.

[0068] In terms of signal processing, stability optimization and time-frequency joint processing are performed on pressure and displacement signals collected by various types of sensors to generate feature-related data. Through signal filtering, noise isolation, and other processing methods, the reliability and stability of the data are effectively improved; while time-frequency joint processing can deeply analyze signal characteristics from both time and frequency dimensions, uncovering leakage-related information that cannot be reflected by a single signal, providing richer and more accurate data support for subsequent leakage analysis.

[0069] By establishing a signal-leakage response mapping model and analyzing the diffusion characteristics of the leaking medium and tracing the leakage path, the propagation process of a leak can be comprehensively and dynamically understood. This model fully considers factors such as pressure transmission, medium penetration, and leakage diffusion. Combined with equipment structural data and characteristic correlation data, it accurately reflects the mapping relationship between signal changes and leakage response, thereby achieving precise analysis and path tracing of the leakage diffusion process and providing a strong basis for early detection and timely handling of leaks.

[0070] By analyzing the correlation of factors affecting the leakage rate, the scope of the leakage impact is defined, and multiphysics coupling calculations are performed to generate field coupling data, making the leakage assessment more comprehensive and accurate. This multiphysics coupling analysis method fully considers the interaction of multiple factors such as pressure fluctuations, displacement deformation, and medium characteristics, and can accurately assess the scope and degree of the leakage's impact on equipment and the surrounding environment, providing scientific data support for developing reasonable leakage handling plans and risk assessments.

[0071] A dual-channel computational model was constructed to dynamically predict leakage rates, while real-time correction of monitoring parameters was implemented, significantly improving the real-time performance and accuracy of leakage monitoring. Based on a hybrid architecture of feedforward neural networks and recurrent neural networks, real-time monitoring data can be effectively processed to achieve dynamic prediction of leakage rates and timely reflect the development trend of leaks. Real-time correction of monitoring parameters allows for adjustments to the monitoring strategy based on actual leakage conditions, further improving monitoring accuracy and reliability, and providing stronger protection for the safe operation of industrial equipment.

[0072] Furthermore, the application of this model in the dynamic monitoring system for leakage in industrial equipment enables comprehensive monitoring and management of equipment leakage throughout the entire process, improving the safety and reliability of industrial production, reducing safety accidents and economic losses caused by leakage, and also contributing to environmental protection, thus having significant economic and social benefits. Attached Figure Description

[0073] Figure 1 This is a schematic diagram illustrating the working principle of the method for constructing the leakage rate calculation model with dual channels of pressure and displacement as described in this invention.

[0074] Figure 2 Design drawings for a method of acquiring equipment structure data;

[0075] Figure 3 Design diagram for a multi-type sensor acquisition network deployment method;

[0076] Figure 4 Design diagram for dual-source monitoring data processing method;

[0077] Figure 5 This is a design diagram for a feature-related data generation method. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Please see Figures 1-5 The present invention relates to a method for constructing a leakage rate calculation model with dual channels of pressure and displacement, the specific implementation steps of which are as follows:

[0080] A1: Divide the target equipment into leakage risk areas and obtain equipment structure data; determine key monitoring locations based on the equipment structure data, and deploy a multi-type sensor acquisition network;

[0081] A2: Perform signal stability optimization processing on the multi-type sensor acquisition network, and simultaneously acquire real-time data of pressure and displacement signals to obtain dual-source monitoring data; perform time-frequency joint processing on the dual-source monitoring data based on pressure amplitude, displacement frequency and signal phase to generate feature correlation data;

[0082] A3: Establish a signal-leakage response mapping model based on feature correlation data and equipment structure data; analyze the diffusion characteristics of the leakage medium based on the signal-leakage response mapping model, and conduct spatially related leakage path tracing to generate dynamic data on leakage propagation;

[0083] A4: Based on the dynamic data of leakage propagation and the dual-source monitoring data, conduct correlation analysis of factors affecting the leakage rate to generate rate influence characteristic data; based on the rate influence characteristic data, define the scope of leakage influence to generate leakage action area data; based on the leakage action area data, perform multi-physics field coupling calculations based on pressure fluctuations, displacement deformation, and medium properties to generate field coupling data.

[0084] A5: Construct a dual-channel calculation model based on field coupling data, and use the dual-channel calculation model to dynamically predict the leakage rate and generate rate prediction data; based on the rate prediction data, perform real-time correction of monitoring parameters to obtain leakage monitoring and control feedback data.

[0085] Example 1:

[0086] In A1, when identifying key monitoring locations and deploying a multi-type sensor network, the following detailed process must be followed: Conduct multi-location structural probing of the target equipment. Use non-destructive testing techniques such as ultrasonic testing and X-ray inspection to comprehensively acquire material sample data, covering information such as material type and manufacturing process of each component. Perform in-depth analysis of the material sample data to identify component types, such as flanges, pipes, and valves, and determine the connection methods between components, such as welding, threaded connections, or flange connections. Simultaneously, measure the wall thickness of each component to obtain the equipment's structural data.

[0087] Mechanical performance analysis was performed on the material sample data of the equipment based on the equipment structure data. Based on mechanical parameters such as elastic modulus, yield strength, and Poisson's ratio, the stress state and deformation trend of each component under different working conditions were evaluated, and the mechanical performance data of the components were obtained. For example, for a pipe component (material 20# steel), taking the elastic modulus E=200GPa, yield strength σs=350MPa, and Poisson's ratio ν=0.3, when the internal pressure P=2MPa, the circumferential stress was calculated using the formula σθ=PD / (2t) (D=200mm is the pipe inner diameter, t=10mm is the wall thickness), resulting in σθ=2×200 / (2×10)=20MPa, which is much smaller than the yield strength of 350MPa, indicating that it is in the elastic deformation stage; the lateral deformation ΔD=D×ν×σθ / E=200×0.3×20 / (200×103)=0.006mm, which can be ignored.

[0088] Then, the flow characteristics of the medium are analyzed based on the mechanical performance data of the components. This process includes real-time observation of the flow velocity inside the equipment, which can be achieved using a Doppler flow meter (measurement range 0.1-10 m / s, accuracy ±0.5%). Three measuring points are evenly distributed across the pipe cross-section, and the average flow velocity is taken as v=2.5 m / s. The permeability of the equipment material sample data is tested using Darcy's law k=QμL / (AΔP), where Q=0.001m3 / s (permeability), μ=0.001Pa·s (medium viscosity), L=0.1m (sample length), A=0.01m2 (cross-sectional area), and ΔP=1000Pa (pressure difference). The calculated k=0.001×0.001×0.1 / (0.01×1000)=1×10-8m2. Based on the medium distribution analysis of flow velocity and permeability, CFD simulation shows that the medium forms a vortex zone at the bend (flow velocity drops to 0.5 m / s), and the flow velocity increases to 4.8 m / s at the valve throttling point.

[0089] By combining data on medium flow characteristics, component mechanical properties, and equipment construction, comprehensive equipment structural data is formed. This data comprehensively reflects the equipment's structural composition, mechanical properties, and medium flow characteristics, providing a sufficient basis for subsequent determination of key monitoring locations.

[0090] After obtaining the equipment structure data, key monitoring locations are identified and a multi-type sensor network is deployed. Step A161 is executed to perform abrupt change detection on the interface features of component connections based on the equipment structure data. Specifically, abrupt change detection identifies locations where there are significant changes in component type and mechanical properties (a threshold is set: a change in elastic modulus > 20% or a change in yield strength > 15% is considered a significant change). For example, at the connection between a DN200 flange (material Q235, E=210GPa) and a DN150 pipe (material 20# steel, E=200GPa), the rate of change in elastic modulus is (210-200) / 200 = 5% < 20%, but the flange thickness increases from 10mm to 20mm (a 100% change rate). This is identified as an interface abrupt change point. Finite element analysis is used to calculate the coordinates of the interface abrupt change point (X=1200mm, Y=500mm), forming the interface abrupt change point data.

[0091] A162 was conducted to perform a sensitivity assessment of the stress state of components based on equipment structural data. This assessment included several aspects: Deformation resistance analysis based on elastic modulus and yield strength, using a deformation resistance coefficient K=E×σs / 1000 (unit: GPa² / MPa). For example, K=200×400 / 1000=80 for valve housing (E=200GPa, σs=400MPa), and K=70 for pipeline (E=200GPa, σs=350MPa), indicating that the pipeline has relatively weak deformation resistance; and deformation degree assessment based on wall thickness and Poisson's ratio, using a deformation coefficient... The coefficient ε = t × ν (t is the wall thickness in mm, ν is Poisson's ratio). For thin-walled pipes (t = 6 mm, ν = 0.3), ε = 1.8, while for thick-walled pipes (t = 12 mm, ν = 0.3), ε = 3.6, indicating a higher degree of deformation in the former. Based on the evaluation of the media erosion characteristics, the erosion index I = k × v (k is the permeability in m², v is the flow velocity in m / s) is used. For valve sealing surfaces (k = 5 × 10⁻⁹ m², v = 3 m / s), I = 1.5 × 10⁻⁸, which is identified as a high-erosion-risk area. Through these quantitative assessments, areas with significant component deformation are identified, yielding data on deformation-sensitive areas.

[0092] Entering A163, based on preset equipment operating conditions and monitoring requirements data, interface abrupt change point data, and deformation-sensitive area data, the spatial distribution of key monitoring points is comprehensively determined. Preset equipment operating conditions include parameters such as pressure, temperature, and medium flow rate during normal operation, while monitoring requirements specify indicators such as monitoring accuracy and frequency. Interface abrupt change points and deformation-sensitive areas are designated as key monitoring areas. Combining equipment operating conditions and monitoring requirements, the location and number of monitoring points are rationally planned, resulting in monitoring location layout data.

[0093] Execute A164 to select sensor types based on the monitoring location data and deploy a multi-type sensor acquisition network. This network includes: MPX5010DP piezoresistive pressure sensor (range 0-10kPa, accuracy ±0.2%FS, response time <1ms); LK-G80 laser displacement sensor (measuring range 0-80mm, resolution 0.01μm, sampling frequency 50kHz); BX120-3AA resistance strain gauge (sensitivity coefficient 2.08±1%, operating temperature -30~80℃); and TDS-100H ultrasonic flow meter (measuring pipe diameter DN50-DN2000, accuracy ±0.5% of reading, measurable flow velocity 0.01-12m / s). During deployment, the pressure sensor is installed in the flange side hole via an M20×1.5 threaded interface (torque 25 N·m), the laser displacement sensor is fixed by a magnetic bracket (angle with the measured surface <5°), and the strain gauge is attached to the point of maximum circumferential stress in the pipeline using a semi-bridging method (curing temperature 60℃, heat preservation for 2 hours) to ensure that the sensor installation deviation is <0.5 mm and the grounding resistance is <4 Ω.

[0094] From acquiring and analyzing equipment structural data to determining key monitoring locations and deploying a multi-type sensor network, the system achieves precise delineation and effective monitoring of leakage risk areas in the target equipment, providing reliable data support for the subsequent construction of leakage rate calculation models. The entire process fully considers the equipment's structural characteristics, mechanical properties, and media flow characteristics, ensuring the scientific validity and effectiveness of the monitoring system. It can promptly and accurately detect potential leakage signs, thus guaranteeing the safe operation of industrial equipment.

[0095] Example 2:

[0096] When establishing feature-related data in A2, the following detailed process must be followed. Signal stability optimization processing is performed on the multi-type sensor acquisition network, specifically through signal filtering and noise isolation. In the signal filtering stage, digital filtering algorithms, such as finite impulse response (FIR) or infinite impulse response (IIR) filters, are used to process the sensors to eliminate high-frequency noise and power frequency interference. Signal cables are shielded using metal braided mesh or aluminum foil wrapping to reduce external electromagnetic interference. Sensor signals are normalized and calibrated to convert the output signals of different ranges and types of sensors into a standard voltage or current signal range, such as 0-5V or 4-20mA, ensuring data consistency and comparability. In the noise isolation stage, vibration isolation brackets are installed to reduce mechanical vibration interference, and flexible connection components are used to reduce the transmission of rigid vibrations during equipment operation. Simultaneously, the sensor installation position is optimized to avoid internal noise sources, such as pumps and compressors, which experience severe vibrations. This allows for the acquisition of real-time signals during equipment operation, resulting in dual-source monitoring data: real-time pressure and displacement signals.

[0097] The pressure signal from dual-source monitoring data undergoes time-series segmentation processing. Based on the periodicity of equipment operation or key points of change in operating conditions, the continuous pressure signal is divided into several time segments, each corresponding to an operating stage or stable condition of the equipment, thereby establishing a pressure-time correlation curve. This curve, with time on the horizontal axis and pressure amplitude on the vertical axis, visually reflects the pressure change trend over time.

[0098] The frequency characteristics and amplitude distribution of pressure fluctuations are calculated based on the pressure-time correlation curve. Signal processing methods such as Fast Fourier Transform (FFT) are used to analyze the energy distribution of the pressure signal in different frequency bands, obtaining the main frequency components of the pressure fluctuations. Simultaneously, the probability of pressure amplitude occurrence in different intervals is statistically analyzed, and amplitude distribution histograms are plotted to clarify the frequency characteristics and amplitude distribution of pressure fluctuations. Based on this, correlation analysis between pressure fluctuations and location is performed. Pressure signals from different monitoring locations are compared, and the correlation coefficients of pressure fluctuations at each location are calculated to determine the propagation law and coupling relationship of pressure fluctuations within the equipment space, obtaining pressure-location evolution characteristic data.

[0099] A curve relating displacement frequency to position was established based on dual-source monitoring data. With the monitoring position on the horizontal axis and the displacement signal frequency on the vertical axis, the dominant frequency or frequency distribution of the displacement signal at each position was plotted on a coordinate system, forming an intuitive curve. By analyzing this curve, the rate of frequency change at different positions was calculated—that is, the ratio of the difference in displacement frequency between adjacent positions to the position interval—thus obtaining spatiotemporal frequency evolution characteristic data, reflecting the changing trend of displacement frequency in the spatial and temporal dimensions of the equipment.

[0100] Real-time changes in signal phase are recorded based on dual-source monitoring data. A phase detection algorithm is used to calculate the phase of the pressure and displacement signals at each sampling moment, recording the phase changes over time. Then, a correlation feature analysis between signal phase and position is performed, comparing the differences and consistency of signal phases at different monitoring locations, calculating the relationship between phase difference and position interval, obtaining phase spatiotemporal evolution characteristic data, and revealing the propagation characteristics and phase coupling relationship of the signal phase in the equipment space.

[0101] After acquiring pressure-location evolution characteristic data, frequency spatiotemporal evolution characteristic data, and phase spatiotemporal evolution characteristic data, these data are comprehensively analyzed to interpret the combined impact of signal combinations on the leakage rate. Multivariate statistical analysis methods, such as principal component analysis (PCA) or partial least squares analysis (PLA), are used to identify signal combinations that significantly affect the leakage rate, determine the critical values ​​and effective intervals of key signal combinations (i.e., when the signal combination falls within a certain interval, the leakage rate may change significantly), and thus generate time-frequency evolution characteristic data of the signal combinations.

[0102] A dynamic evolution model of a two-dimensional signal space is established based on time-frequency evolution characteristic data, and feature changes are extracted to obtain feature correlation data. The specific implementation steps are as follows:

[0103] A271: A two-dimensional signal space is constructed based on time-frequency evolution characteristic data, with pressure amplitude, displacement frequency, and signal phase as coordinate axes. This two-dimensional signal space uses pressure amplitude and displacement frequency as the primary dimensions and signal phase as the auxiliary dimension. Signal data at each monitoring moment is mapped into this space to obtain signal space coordinate data, where each coordinate point represents the signal state at a specific moment.

[0104] A272: Establishing a feature change trajectory based on signal spatial coordinate data. Connecting the signal spatial coordinates at different times in chronological order forms a trajectory line reflecting the change of signal state over time. By spatial mapping of time-series sampling points, a feature motion trajectory curve is constructed. This curve intuitively displays the dynamic change process of signal features in two-dimensional space, thus obtaining feature trajectory data.

[0105] A273: Perform location-based hierarchical processing on feature trajectory data. Based on the structural partitions of the equipment or the distribution of monitoring locations, the feature trajectory is divided into different location segments. The trajectory data within each location segment is analyzed to identify the characteristic changes in different location segments, including trends, rates of change, and interrelationships, thus obtaining hierarchical characteristic data. For example, in the high-temperature, high-pressure area of ​​the equipment, the trend of signal characteristic changes may differ from that in the low-temperature, low-pressure area, and the rate of change may also be faster.

[0106] A274: A dynamic model of a two-dimensional signal space is constructed based on hierarchical characteristic data. Using spatial interpolation and numerical fitting methods, a continuous expression of feature changes is established in the two-dimensional signal space, transforming discrete feature trajectory data into a continuous mathematical model, thus obtaining dynamic evolution model data. This model can predict the changing trends of signal characteristics under different operating conditions.

[0107] A275: Feature extraction based on gradient, curvature, and frequency characteristics of dynamic evolution model data. Gradient characteristics reflect the speed of signal feature changes, curvature characteristics reflect the degree of bending of feature changes, and frequency characteristics reflect the periodicity of feature changes. By extracting these characteristics, model characteristic data is obtained, enabling a deeper understanding of the dynamic changes in signal characteristics.

[0108] A276: Signal correlation analysis based on model characteristic data, focusing on the coupling relationships and feedback mechanisms between signals. This involves analyzing the mutual influence and coupling relationships between pressure amplitude, displacement frequency, and signal phase, determining their feedback mechanisms, obtaining signal correlation data, and clarifying the intrinsic connections between signals.

[0109] A277: Extract indicators representing the dynamic changes of features based on model characteristic data and signal correlation data. These indicators can reflect signal stability, the degree of drastic change in features, etc. By extracting these indicators, feature correlation data is generated, providing key feature inputs for the subsequent establishment of a signal-leakage response mapping model.

[0110] Through the detailed implementation steps outlined above, from the acquisition and processing of dual-source monitoring data to the analysis and extraction of signal features, and then to the construction of a two-dimensional signal space dynamic evolution model and the generation of feature correlation data, time-frequency joint processing of pressure and displacement signals was achieved. This process deeply explored the inherent correlation characteristics between signals, laying a solid data foundation for accurately establishing the mapping relationship between signals and leakage responses. The entire process fully utilized signal processing and data analysis techniques, ensuring the accuracy and reliability of the feature correlation data and effectively reflecting the signal characteristic change patterns under equipment leakage conditions.

[0111] Example 3:

[0112] When establishing a signal-leakage response mapping model in A3, the following detailed process must be followed. First, establish a signal-leakage characteristic correspondence table based on feature correlation data and equipment structure data. Feature correlation data includes dynamic change indicators of signal characteristics such as pressure amplitude, displacement frequency, and signal phase. Equipment structure data covers information such as the material, connection method, mechanical properties, and media flow characteristics of each component. By analyzing the mapping relationship between signal changes and leakage responses of different components under leakage conditions, a one-to-one correspondence is formed between signal characteristics and response information such as leakage degree and leakage location, thus obtaining signal response data. For example, when a leak occurs at a flange connection, the pressure amplitude monitored by the nearby pressure sensor will exhibit a specific attenuation pattern, and the displacement frequency detected by the displacement sensor will also change accordingly. These signal changes correspond to the severity of the leak.

[0113] Data standardization and feature dimensionality reduction are performed on the signal response data. Data standardization transforms signal data of different dimensions and ranges into data of a uniform scale. Common methods include min-max standardization, which linearly transforms the data to the [0,1] interval. The formula is as follows:

[0114]

[0115] in, This is the original data. The minimum value of this feature data. The maximum value of this feature data. This is the standardized data. Standardization eliminates the influence of different feature units on the model.

[0116] Feature dimensionality reduction aims to reduce data dimensionality, remove redundant features, and improve model training efficiency. Principal Component Analysis (PCA) is used to map high-dimensional data to a low-dimensional space through linear transformation, while preserving key feature information. Let the original signal response data be... ,in For the sample size, The feature dimension is used to obtain the dimensionality-reduced data after PCA processing. , Principal components whose cumulative variance contribution rate reaches a preset threshold (e.g., 90%) are retained.

[0117] After data standardization and feature dimensionality reduction, a mathematical model of the signal-leakage response is established, yielding response model data. The mathematical model is based on physical mechanisms and signal processing theory, considering the physical relationships between pressure transmission, media penetration, and leakage, as well as the mapping relationship between signal characteristics and leakage parameters. For example, based on fluid dynamics theory, the leakage rate is related to parameters such as pressure difference and leakage orifice diameter, and changes in the pressure signal can reflect changes in the pressure difference. By establishing mathematical expressions to describe these relationships, a preliminary mathematical model is formed.

[0118] Machine learning training is performed based on response model data to construct a nonlinear mapping relationship between signal and leakage response, resulting in a signal-leakage response mapping model. Appropriate machine learning algorithms are selected, such as neural networks or support vector machines (AVM). Taking neural networks as an example, a network structure containing an input layer, hidden layers, and an output layer is constructed. The number of nodes in the input layer is equal to the dimensionality of the reduced features. The input is standardized signal characteristic data; the number of output layer nodes is determined according to the number of leakage response parameters, such as leakage rate and leakage location; the number of hidden layer nodes is determined by empirical formulas or cross-validation.

[0119] During training, the signal response data is divided into training and test sets. The training set is used for model parameter optimization, and the test set is used to evaluate model performance. Backpropagation is used to update network weights and biases to minimize the error between predicted and actual leakage response values. The loss function can be the mean squared error (MAE), as shown in the formula:

[0120]

[0121] in, The number of samples in the test set. This represents the actual leakage response value. These are the model's predicted values.

[0122] After constructing the signal-leakage response mapping model, the diffusion characteristics of the leaking medium are analyzed based on this model. The analysis includes three aspects: pressure transmission, medium penetration, and leakage diffusion. Pressure transmission analysis considers the changes in pressure over time and space during the leakage process, and analyzes the propagation characteristics of pressure waves inside the equipment based on Bernoulli's equation and continuity equation in fluid mechanics. Medium penetration analysis combines the permeability parameters of the equipment material to study the penetration rate and path of the medium through micropores or cracks. Leakage diffusion analysis considers the diffusion mode of the medium in the space surrounding the equipment after leakage, and predicts the diffusion range and concentration distribution of the leaking medium under the influence of factors such as ambient temperature, pressure, and wind speed, thereby obtaining diffusion characteristic data.

[0123] Based on diffusion characteristic data, spatially correlated leakage path calculations are performed to obtain leakage path data. The leakage path calculation includes an analysis of the influence of the leakage medium pressure attenuation law and the component's deformation resistance. The pressure attenuation law analysis calculates the degree of pressure attenuation with leakage distance based on the flow resistance of the medium along the leakage path, considering factors such as pipe friction and local resistance. The component deformation resistance analysis assesses the deformation of components under the pressure of the leakage medium based on their mechanical performance parameters, such as elastic modulus and yield strength. Deformation affects the shape and size of the leakage path, thus affecting the leakage rate. By comprehensively considering pressure attenuation and component deformation, the specific propagation path of the leakage medium within the equipment is determined.

[0124] By identifying key characteristics and abrupt change points in the leakage propagation dynamics data, dynamic data on leakage propagation is obtained. Key characteristics include the changing trend of the leakage rate and the expansion speed of the diffusion range; abrupt change points refer to nodes in the leakage process where the leakage rate suddenly increases or the diffusion direction changes significantly, such as when the leaking medium breaches a weak component, the leakage rate will change abruptly. Through signal analysis and model prediction, these key characteristics and abrupt change points are identified, providing a basis for subsequent analysis of factors affecting the leakage rate.

[0125] Through the detailed steps outlined above, from establishing the signal-leakage characteristic correspondence table to data standardization, feature dimensionality reduction, mathematical model construction and machine learning training, and then to the analysis of the leakage medium's diffusion characteristics and leakage path calculation, a signal-leakage response mapping model is ultimately constructed, yielding dynamic data on leakage propagation. The entire process combines physical mechanism analysis with data-driven methods, ensuring that the model accurately reflects the relationship between signal characteristics and leakage response, providing effective support for leakage rate calculation and prediction. The introduction of data standardization and loss function formulas clarifies the data processing and model optimization process from a mathematical perspective, making the implementation more rigorous and scientific.

[0126] Example 4:

[0127] When generating field coupling data through multiphysics coupling calculations in A4, the implementation method needs to be explained in detail with specific examples. Taking a liquefied petroleum gas (LPG) storage tank in a chemical plant as an example, the tank consists of a cylinder, end caps, flange interfaces, etc., and the medium is liquefied petroleum gas. The working pressure is 1.6 MPa, and the temperature is 25°C. The specific implementation process is as follows:

[0128] Based on dynamic data of leakage propagation and dual-source monitoring data, the influence of different signal combinations on the leakage rate was analyzed to obtain rate influence data. During the operation of the storage tank, if a leak occurs at the flange interface, the pressure sensor will detect periodic fluctuations in the pressure amplitude near the interface, while the displacement sensor will detect an abnormal increase in the displacement frequency of the flange sealing surface. By analyzing the combination of these pressure and displacement signals, such as the relationship between the decay rate of the pressure amplitude and the change in the displacement frequency, the influence of different signal combinations on the leakage rate was determined. For example, when the pressure amplitude decays from 1.6 MPa to 1.4 MPa within 10 minutes, while the displacement frequency increases from 5 Hz to 15 Hz, the leakage rate may show a rapid growth trend. These influence patterns were compiled into rate influence data.

[0129] Leakage rate performance is evaluated based on rate change patterns, leakage duration characteristics, and deformation propagation properties, yielding performance evaluation data. The leakage rate over time is analyzed to determine whether it exhibits linear, exponential, or initial-to-stabilization patterns. The duration from the onset of the leak to a specific stage is statistically analyzed to assess the leakage's development speed. Displacement sensor data is used to analyze the extent and rate of component deformation, such as whether deformation of the flange sealing surface extends from a localized area to the periphery. By combining these factors, the leakage rate performance is evaluated, for example, determining whether the current leak is in a rapid development or stable phase, thus obtaining performance evaluation data.

[0130] The evolution characteristics of leakage rate over time were analyzed using performance evaluation data to obtain rate impact characteristic data. A leakage rate versus time curve was plotted, and the slope, inflection points, and other characteristics of the curve were analyzed to determine the evolution pattern of the leakage rate. For example, in the initial stage of leakage, the rate may increase slowly; as component deformation intensifies, the rate gradually accelerates; and after reaching a certain level, the rate increase tends to level off due to factors such as pressure decay. These evolution characteristics were summarized to obtain rate impact characteristic data, providing a basis for subsequently defining the leakage impact range.

[0131] Numerical simulations of the leakage impact range were conducted based on rate influence characteristic data, and a multi-field coupled analysis model including pressure, displacement, and flow fields was established to obtain impact range data. Computational Fluid Dynamics (CFD) software was used, combined with a three-dimensional geometric model of the storage tank, to set boundary conditions, such as an inlet pressure of 1.6 MPa, an outlet as the leak point, and atmospheric pressure. The interaction of pressure, displacement, and flow fields was considered in the model: changes in the pressure field cause displacement deformation of tank components, such as radial expansion of the cylinder under pressure; displacement deformation, in turn, alters the flow field distribution, such as changes in the shape and size of the leak point after flange deformation, affecting the outflow velocity and direction of the medium; the flow of the medium in the flow field also has a reaction effect on the pressure and displacement fields, such as high-speed flowing medium generating scouring forces on components, exacerbating displacement deformation. Through numerical simulation, the diffusion range, pressure distribution, and component displacement of the leaking medium at different times were obtained, forming impact range data.

[0132] Boundary identification and spatial partitioning of the impact area data yielded leakage effect area data. Based on numerical simulation results, the boundary at which the concentration of the leaking medium reaches a certain threshold (e.g., 10% of the lower explosive limit of liquefied petroleum gas) was determined, dividing the leakage impact area into different regions, such as the core leakage zone, the diffusion influence zone, and the peripheral warning zone. The core leakage zone is near the leak point, with high medium concentration and drastic pressure and displacement changes; the diffusion influence zone is the area where the medium diffuses, with gradually decreasing concentration and relatively smaller pressure and displacement effects; the peripheral warning zone is the area requiring safety monitoring to prevent the spread of danger. Detailed descriptions of the spatial extent, medium characteristics, pressure, and displacement characteristics of each region were provided to obtain leakage effect area data.

[0133] Based on data from the leakage area, multiphysics coupling calculations are performed using pressure fluctuations, displacement deformation, and media properties. The evolution of the field and critical state characteristics are identified, generating field coupling data. Taking the core leakage area as an example, pressure fluctuations further exacerbate the displacement deformation of the flange interface, which in turn expands the leak opening, increasing the leakage rate. Simultaneously, the vaporization characteristics of liquefied petroleum gas absorb heat, causing a decrease in the surrounding temperature and affecting the flow state and pressure distribution of the medium. By establishing a multiphysics coupling calculation model, the interaction process between these physical fields is simulated, analyzing the evolution of the field over time, such as how the frequency and amplitude of pressure fluctuations change, how the expansion rate of displacement deformation is affected by pressure and media flow, and how the temperature and concentration distribution of the medium alter the flow and pressure fields. Critical state characteristics in the field are also identified, such as a sudden and sharp increase in the leakage rate when displacement deformation reaches a certain critical value, or the potential for component rupture and the formation of new leakage paths when pressure fluctuations exceed a certain threshold. These calculation results and analytical conclusions are compiled into field coupling data, providing crucial input data for the subsequent construction of a dual-channel calculation model.

[0134] During implementation, the complexity of actual operating conditions must also be considered. For example, other equipment or obstacles may exist around the storage tank, affecting the diffusion path of the medium. Therefore, an accurate geometric model of the surrounding environment needs to be constructed during numerical simulation. The properties of the medium, such as viscosity and density, change with temperature and pressure, requiring the use of appropriate equations of state to describe these changes. The mechanical properties of components may exhibit nonlinear characteristics, such as plastic deformation after exceeding the yield strength, necessitating the use of nonlinear constitutive models in the calculation. By fully considering these factors, the accuracy and reliability of multiphysics coupling calculations are ensured, thereby generating field coupling data that can realistically reflect the interaction of various physical fields during the leakage process.

[0135] Taking this liquefied petroleum gas storage tank as an example, through the detailed implementation steps described above, from the analysis of the rate influence law to performance evaluation, and then to the establishment of a multi-field coupling model, the partitioning of the influence range, and the field coupling calculation, the correlation analysis of the factors affecting the leakage rate and the definition of the leakage effect area were achieved, ultimately generating field coupling data. The entire process closely integrates the specific equipment and medium characteristics, fully considers the interaction between multiple physical fields, provides a solid foundation for the subsequent dynamic prediction of the leakage rate and the correction of monitoring parameters, ensures that the leakage monitoring system can accurately reflect the actual leakage state of the equipment, and provides effective technical support for the safe operation of industrial equipment.

[0136] Example 5:

[0137] When constructing a dual-channel calculation model and dynamically predicting leakage rates in A5, a natural gas pipeline compressor is used as an example for specific implementation. This compressor is a reciprocating structure, including components such as cylinders, pistons, piston rods, and sealing assemblies. The medium it transports is natural gas, with an operating pressure range of 3-5 MPa and a temperature range of 10-30℃. Dynamic monitoring of the leakage rate of its cylinder sealing surface is required.

[0138] Feature extraction was performed on the field-coupled data, and a training sample set was constructed for the dual-channel computational model. The field-coupled data encompassed multi-physics interaction data, including pressure fluctuations during compressor operation, piston rod displacement deformation, and natural gas medium characteristics (such as molar mass and compressibility factor). Taking a cylinder sealing surface leakage scenario as an example, features such as the pressure amplitude decay gradient (e.g., a pressure drop of 0.1 MPa every 10 seconds), abrupt changes in displacement frequency (e.g., the radial displacement frequency of the sealing surface jumps from 20 Hz to 45 Hz), and spatial distribution of medium flow velocity (e.g., a flow velocity of 15 m / A near the leak) were extracted as input features. Simultaneously, the actual measured leakage rate (e.g., 0.5 kg / A) was used as the rate label. Through the accumulation of historical monitoring data, 1000 sets of data under different leakage levels were collected and divided into training and test sets in an 8:2 ratio to form the training sample data.

[0139] A hybrid dual-channel computational network structure based on a feedforward neural network (FNN) and a recurrent neural network (RNN) was constructed using training sample data. The input layer has 6 nodes, corresponding to features such as pressure amplitude, displacement frequency, signal phase, medium temperature, flow velocity, and component wall thickness. The hidden layer consists of two layers: the first is a 12-node FNN layer for processing static features such as pressure amplitude, and the second is an 8-node RNN layer (using LATM units) for capturing the sequential features of displacement frequency over time. The output layer has 1 node, corresponding to the predicted leakage rate. During model training, the Adam optimizer was used, with a batch size of 32 pairs of training data for iterative training, and the number of iterations was set to 200, allowing the model to learn the mapping relationship between input features and leakage rate.

[0140] The rate prediction model was optimized for accuracy using cross-validation. The training set was further divided into five subsets. Four subsets were used to train the model each time, and one subset was used to validate the accuracy. This process was repeated five times, and the average error was used as the model evaluation metric. By adjusting the number of hidden layer nodes, the learning rate (e.g., from 0.01 to 0.005), and the regularization parameter (L2 regularization coefficient adjusted from 0.01 to 0.001), the model's ability to memorize time-series data and fit static features was optimized, reducing overfitting and yielding the calculated model data.

[0141] Dynamic prediction of leakage rate is achieved using computational model data. When a slight leak occurs at the compressor cylinder sealing surface, real-time input data such as the currently monitored pressure amplitude (4.2 MPa), displacement frequency (30 Hz), and medium temperature (25 °C) are used. The model outputs an instantaneous rate of 0.2 kg / A and a cumulative amount of 10 kg (calculated over 1 minute), generating a spatial distribution cloud map showing that the leakage impact is concentrated in a 0.5-meter area around the sealing surface. As the leak progresses, the input data is updated to a pressure amplitude of 3.8 MPa and a displacement frequency of 40 Hz. The model predicts that the instantaneous rate increases to 0.8 kg / A, the cumulative amount increases linearly over time, and the spatial distribution expands to a 1.2-meter area around the sealing surface.

[0142] The key stages of the leakage formation process were analyzed using the rate prediction data. In the initial leakage stage (0-5 minutes), the rate slowly increases from 0 to 0.3 kg / A, corresponding to the appearance of microcracks on the sealing surface, with relatively small pressure fluctuations. In the diffusion and development stage (5-15 minutes), the rate rapidly increases to 1.2 kg / A, displacement deformation intensifies, and cracks propagate. In the stable leakage stage (15-30 minutes), the rate remains at 1.0-1.3 kg / A, and pressure decay and component deformation reach a dynamic equilibrium. In the rate decay stage (after 30 minutes), due to the continuous decrease in system pressure, the rate gradually decreases to below 0.5 kg / A. In the leakage termination stage, when the pressure drops to near ambient pressure, the rate approaches 0. Rate evolution data was obtained by identifying the rate abrupt change characteristics of each stage (e.g., a sudden increase in the rate slope during the diffusion and development stage).

[0143] Based on the rate evolution data, monitoring parameters are corrected in real time for each stage. In the initial leakage stage, the sampling frequency of the pressure sensor is increased from 1Hz to 5Hz to capture minute pressure fluctuations. In the diffusion and development stage, the number of monitoring points for the displacement sensor is increased, with four additional measuring points evenly distributed around the sealing surface to ensure comprehensive deformation data. In the stable leakage stage, the input frequency of the medium characteristic parameters is adjusted from updating every 10 minutes to every 5 minutes to accommodate possible changes in medium composition. In the rate decay stage, the zero-point drift of the pressure sensor is closely monitored, and calibration is performed hourly. In the leakage termination stage, a backup sensor is activated for cross-validation to avoid misjudgments caused by a single sensor failure. Through these corrections, leakage monitoring and control feedback data is generated and transmitted to the compressor monitoring system, enabling dynamic adjustment of the monitoring strategy.

[0144] During implementation, the real-time performance and adaptability of the model must be considered. For example, when the compressor operating conditions change abruptly (such as the load increasing from 70% to 90%), the model needs to update the weights of the input features in real time, and fine-tune the model parameters using the latest 200 sets of data through an online learning mechanism. For changes in the composition of natural gas media (such as increased corrosivity due to increased hydrogen sulfide content), the media characteristic database needs to be updated synchronously to ensure the accuracy of the model input parameters. When abnormal noise occurs in the sensor, the data filtering mechanism is automatically triggered, using median filtering combined with Kalman filtering to process the original signal before inputting it into the model for calculation.

[0145] Taking this natural gas compressor as an example, a dual-channel computational model was constructed and applied through steps such as feature extraction, hybrid network construction, model optimization, dynamic prediction, stage analysis, and parameter correction. The entire process combines the equipment's structural characteristics and the properties of the medium, utilizing machine learning algorithms to process multi-source monitoring data. This enables dynamic prediction of leakage rates and optimization of monitoring strategies, providing an operable technical solution for leak early warning in industrial equipment. It ensures timely detection of potential leaks and adjustment of monitoring priorities during actual operation, thereby improving equipment operational safety.

[0146] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do 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 process, method, article, or apparatus.

[0147] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a leakage rate calculation model with dual channels of pressure and displacement, characterized in that, Includes the following steps: A1: Delineate the leakage risk area of ​​the target equipment and obtain structural data; determine key monitoring locations and deploy a multi-type sensor network; A2: Optimize sensor network signal stability, synchronously collect real-time pressure and displacement data to obtain dual-source monitoring data; generate feature correlation data through time-frequency joint processing of dual-source data. A3: Establish a signal-leakage response mapping model based on feature correlation data and structural data, analyze the diffusion characteristics of the leaking medium and trace the path to generate propagation dynamic data; A4: Analyze the factors affecting the leakage rate, generate data on the rate influence characteristics, define the scope of influence, perform multi-physics coupling calculations, and generate field coupling data. A5: Construct a dual-channel model to predict the leakage rate, identify the leakage stage based on the predicted data, correct the monitoring parameters, and generate monitoring and control feedback data.

2. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 1, characterized in that, A1 includes the following steps: A11: Perform multi-location structural detection on the target equipment to obtain material sample data of the equipment; A12: Identify the component types, connection methods, and wall thicknesses from the equipment material sample data to obtain the equipment structure data; A13: Based on the equipment structure data, perform mechanical property analysis on the equipment material sample data based on elastic modulus, yield strength and Poisson's ratio to obtain the component mechanical property data; A14: Analyze the flow characteristics of the medium based on the mechanical performance data of the components to obtain the flow characteristic data of the medium. The flow characteristic analysis of the medium includes the observation of the flow velocity inside the equipment, the permeability test of the material sample data of the equipment, and the analysis of the medium distribution based on the flow velocity and permeability. A15: Combine media flow characteristic data, component mechanical performance data, and equipment construction data into equipment structure data; A16: Determine key monitoring locations based on equipment structure data and deploy a multi-type sensor acquisition network.

3. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 2, characterized in that, A16 includes the following steps: A161: Based on the equipment structure data, perform abrupt change detection on the characteristics of the component connection interface to obtain interface abrupt change point data. Specifically, abrupt change detection involves identifying the location of significant changes in component type and mechanical properties. A162: Based on the equipment structure data, the stress state of the components is sensitively assessed to identify areas of large deformation and obtain data on deformation-sensitive areas. The deformation sensitivity assessment includes analysis of the resistance to deformation based on elastic modulus and yield strength, assessment of the degree of deformation based on wall thickness and Poisson's ratio, and assessment of the scouring characteristics of the permeable medium. A163: Determine the spatial distribution of key monitoring points based on preset equipment operating conditions and monitoring requirements data, interface mutation point data, and deformation-sensitive area data to obtain monitoring location layout data; A164: Select sensor types based on the data from the monitoring locations, and deploy a multi-type sensor acquisition network including pressure sensors, displacement sensors, strain sensors, and flow velocity sensors.

4. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 3, characterized in that, A2 includes the following steps: A21: Perform stability optimization processing based on signal filtering and noise isolation on the multi-type sensor acquisition network, and acquire real-time signals during equipment operation to obtain dual-source monitoring data; A22: The pressure signal in the dual-source monitoring data is segmented based on time series to establish a pressure-time correlation curve; A23: Calculate the frequency characteristics and amplitude distribution of pressure fluctuations based on the pressure-time correlation curve, and perform correlation analysis between pressure fluctuations and location to obtain pressure-location evolution characteristic data; A24: Based on the dual-source monitoring data, establish the relationship curve between displacement frequency and position, and calculate the frequency change rate at different position segments to obtain frequency spatiotemporal evolution characteristic data; A25: Real-time changes in signal phase are recorded based on dual-source monitoring data, and correlation characteristic analysis of signal phase and position is performed to obtain phase spatiotemporal evolution characteristic data. A26: Based on pressure-position evolution characteristic data, frequency spatiotemporal evolution characteristic data, and phase spatiotemporal evolution characteristic data, the comprehensive impact of signal combinations on leakage rate is analyzed, and the critical values ​​and effective ranges of key signal combinations are identified, thereby generating time-frequency evolution characteristic data of signal combinations. A27: Establish a dynamic evolution model of a two-dimensional signal space based on time-frequency evolution characteristic data, and extract feature changes to obtain feature correlation data.

5. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 4, characterized in that, A21 includes the following steps: The stability optimization process based on signal filtering and noise isolation is performed on the multi-type sensor acquisition network, and real-time signals during equipment operation are collected to obtain dual-source monitoring data. The signal filtering includes digital filtering algorithm processing of sensors, interference signal shielding of signal cables, and normalization calibration of sensor signals. The noise isolation includes setting up vibration isolation brackets, using flexible connection components, and optimizing the noise source avoidance of sensor installation positions.

6. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 5, characterized in that, A27 includes the following steps: A271: Construct a two-dimensional signal space with pressure amplitude, displacement frequency, and signal phase as coordinate axes based on time-frequency evolution characteristic data, thereby obtaining signal space coordinate data; A272: Establish characteristic change trajectories based on signal spatial coordinate data, and construct characteristic motion trajectory curves through spatial mapping of time-series sampling points, thereby obtaining characteristic trajectory data; A273: Perform location-based hierarchical processing on feature trajectory data to identify the characteristic change characteristics of different location segments, including change trends, change rates, and interrelationships, thereby obtaining hierarchical characteristic data; A274: Construct a dynamic model of a two-dimensional signal space based on hierarchical characteristic data, and establish a continuous expression of feature changes through spatial interpolation and numerical fitting to obtain dynamic evolution model data; A275: Extract features from dynamic evolution model data based on gradient characteristics, curvature characteristics, and frequency characteristics of feature changes to obtain model characteristic data; A276: Based on model characteristic data, perform signal correlation analysis based on the coupling relationship and feedback mechanism between signals to obtain signal correlation data; A277: Extract indicators that characterize the dynamic changes of features based on model characteristic data and signal correlation data, thereby generating feature correlation data.

7. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 6, characterized in that, A3 includes the following steps: A31: Based on the feature association data and equipment structure data, establish a signal-leakage characteristic correspondence table that includes the mapping relationship between signal changes and leakage response in different components, thereby obtaining signal response data; A32: Perform data standardization and feature dimensionality reduction on the signal response data, and establish a mathematical model of the signal-leakage response to obtain the response model data; A33: Based on response model data, machine learning training is performed to construct a nonlinear mapping relationship between signal and leakage response, thereby obtaining a signal-leakage response mapping model; A34: Based on the signal-leakage response mapping model, the diffusion characteristics of the leaking medium, including pressure transmission, medium penetration and leakage diffusion, are analyzed to obtain diffusion characteristic data; A35: Based on diffusion characteristic data, spatially related leakage path calculations are performed to obtain leakage path data. The leakage path calculation includes the influence analysis of the pressure decay law of the leakage medium and the deformation resistance of the components. A36: Identify key characteristics and abrupt change points in the leakage propagation dynamic data to obtain leakage propagation dynamic data.

8. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 7, characterized in that, A4 includes the following steps: A41: Based on the dynamic data of leakage propagation and the dual-source monitoring data, the influence of different signal combinations on the leakage rate is analyzed to obtain the rate influence data; A42: The leakage rate performance is evaluated based on the rate change law, leakage duration characteristics, and deformation propagation characteristics to obtain performance evaluation data. A43: Analyze the evolution characteristics of leakage rate over time from the performance evaluation data to obtain rate impact characteristic data; A44: Numerical simulation of the leakage influence range is performed based on the rate influence characteristic data. A multi-field coupled analysis model including pressure field, displacement field and flow field is established. Boundary identification and spatial partitioning of the leakage area are performed. The interaction between pressure fluctuation and displacement deformation is calculated. The field evolution law is identified by combining the viscosity, density and compressibility parameters of the medium, so as to obtain the influence range data. A45: Perform boundary identification and spatial partitioning on the impact range data to obtain the leakage effect area data; A46: Perform multi-physics field coupling calculations based on pressure fluctuations, displacement deformation, and medium properties according to the data of the leakage area, identify the field evolution law and critical state characteristics, and generate field coupling data.

9. The method for constructing the leakage rate calculation model with dual channels of pressure and displacement according to claim 8, characterized in that, A5 includes the following steps: A51: Extract features from field-coupled data and construct a training sample set for a dual-channel computation model to obtain training sample data, which includes input features and rate labels. A52: Construct a dual-channel computational network structure based on a hybrid architecture of feedforward neural network and recurrent neural network according to the training sample data, optimize the model prediction accuracy by using cross-validation method, and train the model to obtain the rate prediction model; A53: Optimize the prediction accuracy of the rate prediction model using a cross-validation method to obtain the computational model data; A54: Dynamically predict the leakage rate using computational model data to obtain rate prediction data, which includes instantaneous rate, cumulative amount, and spatial distribution. A55: The rate prediction data is analyzed to identify the key stages and rate mutation characteristics in the leakage formation process, thereby obtaining rate evolution data. The key stages include the initial leakage stage, the diffusion development stage, the stable leakage stage, the rate decay stage, and the leakage termination stage. In the initial stage, the pressure sampling frequency is increased to 5Hz. In the diffusion stage, displacement sensor points are added. In the stable stage, the medium parameter update cycle is shortened to 5 minutes. In the decay stage, the pressure sensor zero point is calibrated every hour. In the termination stage, the backup sensor cross-validation is initiated. A56: Based on the rate evolution data, the monitoring parameters at each stage are corrected in real time to obtain leakage monitoring and control feedback data.

10. The application of a pressure and displacement dual-channel leakage rate calculation model obtained by the construction method according to any one of claims 1-9 in an industrial equipment leakage dynamic monitoring system.

Images