Intelligent valve monitoring system based on industrial vision

Abstract

The invention discloses an intelligent valve monitoring system based on industrial vision, which comprises a multispectral imaging module which comprises a visible light sensor, a thermal infrared sensor and a short wave infrared sensor, synchronously acquires three-dimensional shape data of a valve at a preset sampling rate, processes light velocity, modulation frequency and phase difference calculation axial resolution through a three-dimensional reconstruction method, and obtains a three-dimensional shape data of the valve; ambient light and mechanical vibration interference are inhibited; the corrosion evaluation module is used for calculating a corrosion index based on the multispectral reflectivity difference value and the characteristic spectrum peak intensity, judging the corrosion grade according to a preset threshold value and triggering a maintenance instruction; and the predictive maintenance module is used for dynamically correcting the residual life prediction model through the multi-source characteristic parameters, dividing life stages and adjusting a monitoring strategy. According to the invention, the blank of multi-dimensional perception, dynamic prediction and autonomous optimization in the field of industrial vision and intelligent monitoring is filled, and better intelligent valve monitoring is realized.

Description

Technical Field

[0001] This invention relates to the field of monitoring systems, and more specifically to an intelligent valve monitoring system based on industrial vision. Background Technology

[0002] As a core component of fluid control systems, the health status of industrial valves directly impacts production safety and efficiency. Traditional valve monitoring primarily relies on manual inspections or single sensors, which has the following limitations:

[0003] Limited detection dimensions: It only monitors physical quantities and is difficult to detect surface defects such as rust and deformation.

[0004] Delayed response: Manual inspections take a long time, making it difficult to provide timely warnings of sudden malfunctions.

[0005] Data silos: The lack of integrated analysis of multi-source data makes it difficult to comprehensively assess the valve status.

[0006] With the advancement of industry and intelligent manufacturing, industrial vision technology has become a core means of equipment monitoring, achieving non-contact, multi-dimensional equipment status perception through technologies such as multispectral imaging, deep learning, and digital twins. However, existing solutions still face the following technical bottlenecks:

[0007] Changes in lighting and mechanical vibrations interfere with imaging accuracy, and changes in data distribution over time lead to a decrease in algorithm performance. Experience-based installation is prone to missing key monitoring areas. Therefore, a valve intelligent monitoring system based on industrial vision is proposed. Summary of the Invention

[0008] The technical problem to be solved by this invention is: how to solve the problems of poor adaptability to complex environments and unscientific sensor deployment in existing valve monitoring systems, and to provide a valve intelligent monitoring system based on industrial vision.

[0009] The present invention solves the above-mentioned technical problems through the following technical solutions, the present invention comprising:

[0010] The multispectral imaging module includes visible light, thermal infrared and short-wave infrared sensors, which synchronously acquire three-dimensional morphological data of the valve at a preset sampling rate, and process the axial resolution by calculating the light speed, modulation frequency and phase difference through a three-dimensional reconstruction method, and suppress the interference of ambient light and mechanical vibration.

[0011] The corrosion assessment module is used to calculate the corrosion index based on the difference in multispectral reflectance and the intensity of characteristic spectral peaks, determine the corrosion level according to the preset threshold, and trigger maintenance commands.

[0012] The predictive maintenance module is used to dynamically correct the remaining lifetime prediction model through multi-source feature parameters, divide the lifetime into stages, and adjust the monitoring strategy.

[0013] The autonomous calibration system calculates the KL divergence of the data distribution in real time and triggers model updates to suppress performance drift.

[0014] A distributed fiber optic strain sensor array is deployed in stages along the main stress direction of the valve. The installation position is optimized by combining stress gradient, temperature fluctuation and construction feasibility, and the monitoring blind zone is compensated by interpolation.

[0015] The millimeter-wave vibration monitoring unit captures the vibration spectrum of the target frequency band in a non-contact manner, avoiding installation interference;

[0016] Edge computing units run hardware-accelerated deep learning models to enable localized real-time analysis;

[0017] The cloud-based digital twin platform integrates real-time fluid simulation with a historical database to perform multiphysics coupling analysis.

[0018] Furthermore, the implementation process of the three-dimensional reconstruction method is as follows:

[0019]

[0020] In the formula: Δz is the axial resolution;

[0021] c is the speed of light, and f is the preset modulation frequency;

[0022] Where is the time-of-flight phase difference, and k is the structured light compensation coefficient;

[0023] d struct The spacing of the structured light stripes.

[0024] Furthermore, the multiphysics coupling process of the cloud-based digital twin platform is as follows:

[0025]

[0026] In the formula, σ eq The stress is the equivalent stress, where T is the current temperature and E is the elastic modulus of the valve material.

[0027] α t T0 is the coefficient of thermal expansion, and T0 is the reference temperature.

[0028] ε max Where is the maximum strain value, and C is the strain correction factor.

[0029] Furthermore, the corrosion assessment module calculates the corrosion index based on the difference in multispectral reflectance and the intensity of characteristic spectral peaks, and determines the corrosion level and triggers maintenance commands according to a preset threshold. The specific process is as follows:

[0030] Calculate the corrosion index:

[0031] R = λ1·(R) S1 -R S2 )+λ2·∑(I peak ·ω);

[0032] In the formula: R S1 and R S2 The reflectivity difference for the selected band;

[0033] I peak The intensity of the characteristic spectral peak of corrosion is given by ω, which is a weighting coefficient.

[0034] λ1 and λ2 are scaling factors calibrated through corrosion experiments;

[0035] After obtaining the corrosion index R, the corrosion level is determined based on the corrosion index R and the thresholds Γ1, Γ2, and Γ3.

[0036] Furthermore, the predictive maintenance module dynamically corrects the remaining lifetime prediction model in the remaining lifetime prediction model using multi-source feature parameters, as follows:

[0037]

[0038] In the formula, t base The baseline lifetime is given by k, where k is the decay rate coefficient.

[0039] S m For multi-source feature parameters, ω m For the corresponding feature weights;

[0040] The specific process of dynamically correcting the remaining lifetime prediction model using multi-source feature parameters is as follows: When the multi-source feature parameter S is detected... m When a mutation occurs, dynamic lifetime correction is performed.

[0041] Furthermore, the optimization process for the installation location of the distributed fiber Bragg grating strain sensor is as follows:

[0042] Calculate the location optimization index:

[0043]

[0044] In the formula: S s T is the stress sensitivity factor. g This represents the average temperature fluctuation.

[0045] E i V is the electromagnetic interference coefficient. d This represents the integral value of the vibration disturbance.

[0046] C m For construction feasibility, τ and N are calibration constants;

[0047] Sensors are deployed in stages according to the optimization index L, and blind spots are compensated by interpolation.

[0048] Furthermore, the construction feasibility C m The calculation process is as follows:

[0049]

[0050] Where: Da is the distance between the installation point and the obstacle, and Rc is the radius of curvature of the pipeline;

[0051] δ and ρ are engineering safety threshold parameters;

[0052] The feasibility assessment process is as follows: measure the installation environment parameters and obtain the real-time measurement values ​​of obstacle distance Da and pipeline curvature radius Rc;

[0053] Query the preset security thresholds δ and ρ, where δ = κ·Dp and ρ = λ·Rmin;

[0054] When the preset conditions are met, the area is determined to be a workable area;

[0055] According to C m The size triggers a tiered warning.

[0056] Furthermore, dynamic optimization of the installation location was performed. The specific process of dynamic optimization of the installation location is as follows:

[0057] First, construct the objective function:

[0058] min[ω1·(Δσ) 2 +ω2·ΔT+ω3·ξ];

[0059] in Let ξ be the stress gradient, ΔT be the temperature range, ξ be the vibration disturbance factor, and D be the stress gradient. s Install distance reference values ​​for the sensor, D p The diameter of the pipe;

[0060] ω1, ω2 and ω3 are weighting coefficients;

[0061] After setting geometric constraints, an optimization algorithm is used to solve the problem. The algorithm is solved iteratively through a heuristic optimization algorithm. The algorithm parameters include one or more of the following: population size, number of iterations, and convergence threshold.

[0062] Furthermore, the operation process of the autonomous calibration system is as follows:

[0063] Real-time calculation of the current data distribution P t KL divergence with the initial distribution P0:

[0064]

[0065] When D KL When the dynamic threshold is exceeded, the corresponding operation is triggered.

[0066] Compared with existing technologies, this invention has the following advantages: This intelligent valve monitoring system based on industrial vision, through a multispectral imaging module combined with visible light, thermal infrared, and short-wave infrared sensors, achieves synchronous high-precision monitoring of valve morphology, temperature, and surface condition, improving the comprehensiveness and reliability of data acquisition. Based on reflectivity difference and characteristic spectral analysis, it automatically quantifies the degree of corrosion and provides graded early warnings, supporting precise decisions from surface treatment to component replacement, avoiding human error. By dynamically correcting the remaining life prediction model through multi-source characteristic parameters, it divides the lifespan into stages and adjusts the monitoring strategy, extending the valve's service life and reducing the risk of sudden failures. It calculates the optimal installation location by combining multiple factors such as stress, temperature, and vibration, and deploys sensor arrays in stages to reduce monitoring blind spots, improve data coverage and accuracy, and enable real-time monitoring. Changes in data distribution trigger model updates, suppressing model drift issues during long-term operation and ensuring the continuous stability of system analysis. It integrates real-time fluid simulation with historical databases to build a digital twin platform, supporting remote monitoring and multi-physics coupled analysis, thus improving the scientific nature of maintenance decisions. Millimeter-wave technology captures vibration spectra in target frequency bands, avoiding installation interference from traditional contact sensors and enhancing monitoring flexibility. Combined with installation environment parameters and safety thresholds, it automatically assesses construction feasibility and triggers tiered early warnings, reducing engineering implementation risks. Hardware-accelerated deep learning models enable localized real-time analysis, reducing cloud dependence and improving response speed and data security. From data acquisition and status assessment to maintenance command generation, a closed-loop intelligent monitoring process is formed, reducing manual intervention and improving industrial operation and maintenance efficiency. Attached Figure Description

[0067] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0068] The embodiments of the present invention are described in detail below. These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments.

[0069] like Figure 1 As shown, this embodiment provides a technical solution: a valve intelligent monitoring system based on industrial vision, comprising:

[0070] The multispectral imaging module includes visible light, thermal infrared and short-wave infrared sensors, which synchronously acquire three-dimensional morphological data of the valve at a preset sampling rate, and process the axial resolution by calculating the light speed, modulation frequency and phase difference through a three-dimensional reconstruction method, and suppress the interference of ambient light and mechanical vibration.

[0071] The corrosion assessment module is used to calculate the corrosion index based on the difference in multispectral reflectance and the intensity of characteristic spectral peaks, determine the corrosion level according to the preset threshold, and trigger maintenance commands.

[0072] The predictive maintenance module is used to dynamically correct the remaining lifetime prediction model through multi-source feature parameters, divide the lifetime into stages, and adjust the monitoring strategy.

[0073] The autonomous calibration system calculates the KL divergence of the data distribution in real time and triggers model updates to suppress performance drift.

[0074] A distributed fiber optic strain sensor array is deployed in stages along the main stress direction of the valve. The installation position is optimized by combining stress gradient, temperature fluctuation and construction feasibility, and the monitoring blind zone is compensated by interpolation.

[0075] The millimeter-wave vibration monitoring unit captures the vibration spectrum of the target frequency band in a non-contact manner, avoiding installation interference;

[0076] Edge computing units run hardware-accelerated deep learning models to enable localized real-time analysis;

[0077] The cloud-based digital twin platform integrates real-time fluid simulation with a historical database to perform multiphysics coupling analysis.

[0078] The implementation process of the three-dimensional reconstruction method is as follows:

[0079]

[0080] In the formula: Δz is the axial resolution;

[0081] c is the speed of light, and f is the preset modulation frequency;

[0082] Where is the time-of-flight phase difference, and k is the structured light compensation coefficient;

[0083] d struct The spacing between structured light stripes;

[0084] By combining light speed, modulation frequency, and phase difference, the axial resolution is significantly improved, enabling the capture of minute deformations or defects on the valve surface and internal structure, providing a precise data foundation for subsequent corrosion assessment. The introduction of structured light compensation coefficient and fringe spacing effectively suppresses the interference of ambient light or mechanical vibration on imaging, ensuring stable measurement in complex industrial scenarios. Working in conjunction with multispectral sensors, it achieves spatial alignment of multi-dimensional data such as visible light and thermal infrared, enhancing the ability to analyze valve status from multiple angles.

[0085] In valve monitoring at chemical plants, the system scans the valve surface using a 3D reconstruction method. For example, when a valve undergoes slight deformation due to prolonged high pressure:

[0086] The multispectral imaging module emits modulated light at a preset frequency to simultaneously capture phase difference and structured light stripe spacing;

[0087] The axial resolution is calculated using a formula to generate a high-precision 3D model that clearly displays the local depressions at the valve flange connection.

[0088] The test results are automatically transmitted to the corrosion assessment module, which combines thermal infrared data to determine whether deformation is accompanied by abnormal temperature, providing early warning of potential seal failure risks and guiding maintenance personnel to carry out targeted repairs.

[0089] The process of multiphysics coupling analysis of the digital twin platform is as follows:

[0090]

[0091] In the formula:

[0092] σ eq The stress is the equivalent stress, where T is the current temperature and E is the elastic modulus of the valve material.

[0093] α t T0 is the coefficient of thermal expansion, and T0 is the reference temperature.

[0094] ε max Where C is the maximum strain value, and C is the strain correction factor:

[0095] By simultaneously integrating the effects of temperature changes (thermal expansion) and mechanical strain on valves, multi-physics coupling analysis is achieved, avoiding the limitations of single-factor analysis, significantly improving the accuracy of stress prediction, accurately calculating equivalent stress, identifying potential stress concentration areas of valves under alternating high and low temperatures or high pressure conditions, and providing early warning of crack or fatigue risks. Combined with real-time temperature and strain data, the stress model is dynamically adjusted to adapt to transient operating conditions in complex industrial environments. Through precise input of elastic modulus and coefficient of thermal expansion, data support is provided for valve material selection or design improvement.

[0096] In the monitoring of steam pipeline valves in oil refineries, the system applies multiphysics coupling analysis:

[0097] The valve operates under high-temperature steam (150°C) and periodic fluid impact.

[0098] The current temperature T is 150℃, and the reference temperature T0 is 20℃.

[0099] The maximum strain ε was measured max It is 0.002.

[0100] The valve material has an elastic modulus E of 200 GPa and a coefficient of thermal expansion α. t 12×10 -6 / ℃.

[0101] σeq=200*10 9 Pa*[12×10 -6 [ / ℃*(150-20)℃+0.002 / C]:

[0102] Assuming the strain correction factor C is 1.2, σ is calculated. eq ≈312MPa:

[0103] The system compares the material yield strength (e.g., 300 MPa) and finds σ eq The threshold has been reached, triggering an alert:

[0104] It is recommended to lower the steam temperature to below 130°C or shorten the valve maintenance cycle.

[0105] The valve flange connection area is marked as a high-risk area and should be given priority for non-destructive testing.

[0106] The corrosion assessment method includes:

[0107] Calculate the corrosion index: R = λ1·(R S1 -R S2 )+λ2·∑I peak ·ω;

[0108] In the formula: R S1 With R S2 The reflectivity difference for the selected band;

[0109] I peak The intensity of the characteristic spectral peak of corrosion is given by ω, which is a weighting coefficient.

[0110] λ1 and λ2 are scaling factors calibrated through corrosion experiments;

[0111] After obtaining the corrosion index R, the corrosion state is determined according to the following rules:

[0112] When R≤Γ1, it is determined to be a normal state, and routine monitoring is performed;

[0113] When Γ1<R≤Γ2, it is judged as mild corrosion, and surface treatment is recommended.

[0114] When Γ2<R≤Γ3, it is judged as moderate corrosion, and an anti-corrosion coating repair instruction is generated;

[0115] When R > Γ3, it is determined to be severe corrosion, and a component replacement warning is activated;

[0116] The thresholds Γ1, Γ2, and Γ3 were calibrated through accelerated corrosion tests and satisfy Γ3 = 2.5Γ2 and Γ2 = 1.8Γ1.

[0117] By comprehensively calculating the difference in multispectral reflectance and the intensity of characteristic spectral peaks, the subjectivity of a single indicator is avoided, and the degree of corrosion is scientifically quantified. The corrosion level (normal, light, moderate, and severe) is automatically classified according to the corrosion index, triggering differentiated maintenance instructions, avoiding a one-size-fits-all maintenance strategy, improving resource utilization efficiency, and providing timely warnings in the early stages of corrosion (such as the slight oxidation stage) by combining the threshold calibrated by corrosion experiments, preventing the spread of corrosion and structural damage. By adjusting the weighting coefficients and calibration factors, the model can be adapted to different valve materials or environmental conditions, thereby improving the generalization ability of the model.

[0118] For example, carbon steel valves in chemical plants are exposed to high humidity and salt spray environments for extended periods.

[0119] The multispectral imaging module detected the valve surface in the short-wave infrared band (R... S1 ) and visible light band (R S2 The reflectance difference of ) increased significantly, while the characteristic spectral peak of corrosion (I) peak Intensity increases.

[0120] The system automatically substitutes the reflectance difference, spectral peak intensity, and calibration parameters (λ1=0.6, λ2=0.4, ω=1.2) to calculate the corrosion index R=2.8.

[0121] The judgment and response process is as follows:

[0122] Preset thresholds: Γ1 = 1.0, Γ2 = 1.8, Γ3 = 4.5 (calibrated through accelerated corrosion experiments).

[0123] The assessment result is R = 2.8 (between Γ2 and Γ3), triggering a moderate corrosion warning.

[0124] The system responds by generating a corrosion-resistant coating repair instruction and pushing it to the maintenance personnel's handheld terminal, prompting them to prioritize the valve flange connection.

[0125] The predictive maintenance module dynamically corrects the remaining lifetime prediction model in the remaining lifetime prediction model using multi-source feature parameters.

[0126] In the formula, t base The baseline lifetime is given by k, where k is the decay rate coefficient.

[0127] S m For multi-source feature parameters, ω m For the corresponding feature weights;

[0128] The remaining useful life assessment process includes: calculating the predicted remaining useful life (RUL) value and dividing the life into three phases.

[0129] When RUL>τ1·t baseWhen the current time is reached, it indicates that the device is in its normal lifespan, and the current monitoring cycle should be maintained.

[0130] When τ2·t base ≤RUL≤τ1·t base When this period begins, the maintenance and observation period begins, and the monitoring cycle is shortened to 1 / ηm of the original interval.

[0131] When RUL < τ2·t_base, the emergency replacement period is triggered, and the spare parts allocation process is initiated.

[0132] When multi-source feature parameter S is detected m When a mutation occurs, perform dynamic lifetime correction:

[0133]

[0134] In the formula, τ1 and τ2 are the stage threshold coefficients, ηm is the monitoring cycle adjustment factor, and σ S This is a parameter related to the sensitivity to changes in features.

[0135] By real-time monitoring and weighted calculation of multi-source characteristic parameters (such as vibration, temperature, and pressure), the remaining life is dynamically adjusted to avoid the lag of traditional fixed-life models and improve prediction accuracy. The remaining life is divided into normal period, observation period, and emergency period, and the monitoring frequency and maintenance priority are automatically adjusted to optimize resource allocation and operation and maintenance costs. When the characteristic parameters change abruptly (such as a sudden increase in vibration), the model immediately corrects the remaining life and triggers emergency measures to prevent sudden failures from causing production interruptions. Through weighting coefficients and sensitivity parameters, the model can be adapted to different valve types or operating conditions to enhance its versatility.

[0136] For example, if a steam valve operates in a high-pressure and high-temperature environment, the system continuously monitors parameters such as vibration, temperature, and sealing pressure.

[0137] The system sets a baseline lifespan (t) based on historical data. base =8 years), combined with current parameters (such as stable vibration and normal temperature), the remaining lifespan RUL is calculated to be 6 years, which is determined to be the normal lifespan period, and a monitoring cycle of once a month is maintained.

[0138] A sudden increase in vibration amplitude (ΔS) was detected on a certain day. m If the value increases significantly, the system will immediately initiate dynamic correction.

[0139] Calculate the new remaining lifetime using the formula:

[0140]

[0141] Assume ΔS m =3 (abnormal vibration intensity), σ S =1.5, then the corrected RUL new ≈3.2 years.

[0142] The system determines that RUL has entered the "maintenance observation period", shortens the monitoring cycle to once a week, and pushes maintenance suggestions;

[0143] Maintenance personnel discovered that the valve bearing was worn. After timely replacement, the vibration parameters returned to normal, the system recalculated RUL and restored the original monitoring cycle, thus avoiding potential downtime risks.

[0144] By using dynamic prediction and graded response mechanisms, valve life management is transformed from passive maintenance to proactive prevention, significantly improving the reliability and economy of industrial equipment operation.

[0145] The process for evaluating the installation location of the distributed fiber Bragg grating strain sensor includes:

[0146] Calculate the location optimization index:

[0147] In the formula: S s T is the stress sensitivity factor. g This represents the average temperature fluctuation.

[0148] E i V is the electromagnetic interference coefficient. d This represents the integral value of the vibration disturbance.

[0149] C m For construction feasibility, τ and N are calibration constants;

[0150] The application methods of the location optimization index L include:

[0151] a) Establish a three-dimensional coordinate system and discretize the valve surface into a gridded set of candidate points;

[0152] b) Calculate the location optimization index L value point by point to generate a heat map distribution matrix;

[0153] c) Implement a tiered deployment strategy:

[0154] The main sensor array is deployed at grid points where L > Λ1, with a spacing of Δd1.

[0155] Auxiliary sensors are deployed at grid points with Λ2≤L≤Λ1, with a spacing of Δd2=2Δd1;

[0156] Grid points with L less than Λ2 are marked as monitoring blind spots, and interpolation between adjacent nodes is used for compensation.

[0157] Where Λ1 and Λ2 are optimization thresholds, satisfying Λ1 = 1.5, Λ2, and Δd1 are adaptively adjusted according to the material stress gradient;

[0158] Δd1 and Δd2 are both sensor spacings;

[0159] By integrating multi-dimensional parameters such as stress sensitivity, temperature fluctuation, electromagnetic interference, vibration, and construction feasibility, the advantages and disadvantages of installation locations are scientifically quantified to avoid monitoring deviations caused by single factors. The main and auxiliary sensors are deployed in a hierarchical manner according to the location optimization index, and the spacing is dynamically adjusted in combination with stress gradient to balance monitoring accuracy and cost. Areas with low L values ​​are marked as monitoring blind spots, and data gaps are filled by interpolation of adjacent nodes to ensure full coverage of the monitoring network. Through construction feasibility calculation, installation obstacles (such as pipe bends or obstacles that are too close) are identified in advance, triggering hierarchical early warnings and reducing the risks of project implementation.

[0160] A chemical plant needs to deploy fiber optic grating sensors on the valves of a high-pressure reactor to monitor stress and deformation in real time.

[0161] System measurement environmental parameters:

[0162] Stress sensitivity factor (S) s =0.8 (high stress areas are highly sensitive);

[0163] Average temperature fluctuation (T) g =50℃ (temperature fluctuations are severe near the valve);

[0164] Electromagnetic interference coefficient (E) i =0.3 (slight interference from nearby motors);

[0165] Vibration disturbance integral value (V) d = 2.5m / s 2 (Periodic fluid impact causes vibration);

[0166] Construction feasibility (C) m =1.2 (Installation point is far from obstacles, radius of curvature meets the standard);

[0167] Substitute into the formula to calculate the position optimization index:

[0168]

[0169] The system discretizes the valve surface into a grid and calculates the L value point by point to generate a heat map (areas with high L values ​​are displayed in red).

[0170] Execute the strategy based on the thresholds (Λ1 = 3.0, Λ2 = 2.0):

[0171] Main sensor array: grids with L>3.0 (such as at flange connections), densely deployed at a spacing of Δd1=20cm.

[0172] Auxiliary sensors: grids with a diameter of 2.0 ≤ L ≤ 3.0 (such as in the middle of the valve body), deployed at a spacing of Δd2 = 40cm.

[0173] Blind spot compensation: For grids with L < 2.0 (such as the top of a valve), virtual monitoring points are generated by interpolating data from adjacent nodes.

[0174] During construction, the system detects the distance (D) between a certain installation point and the pipe wall. a Too close (C) m =0.8 < As0 = 1.0), triggering a red alert, prohibiting installation and recommending adjustments.

[0175] During operation and maintenance, due to the aging of valve materials causing changes in stress gradient, the system automatically adjusts Δd1 from 20cm to 15cm to increase the monitoring density in high-risk areas.

[0176] Traditional sensor deployment relies on experience and may lead to data distortion due to overlooking electromagnetic interference or vibration. In contrast, this project utilizes quantitative calculations and hierarchical deployment to precisely place the main sensor in the high-stress area of ​​the flange, while interpolation is used in blind zones to ensure data integrity, significantly improving the reliability and cost-effectiveness of the monitoring network.

[0177] The construction feasibility C m The calculation process is as follows:

[0178]

[0179] In the formula: D a R is the distance between the installation point and the obstacle. c The radius of curvature of the pipeline;

[0180] δ and ρ are engineering safety threshold parameters;

[0181] Construction feasibility assessment process:

[0182] Measure the installation environment parameters:

[0183] Obtain real-time measurements of obstacle distance Da and pipeline curvature radius Rc;

[0184] Query the preset security thresholds δ and ρ, where δ = κ·Dp and ρ = λ·Rmin;

[0185] (Dp is the pipe diameter, Rmin is the minimum bending radius of the optical fiber);

[0186] A zone is considered a workable area when the following conditions are met:

[0187] C m ≥As, and

[0188] Triggering a tiered warning:

[0189] When As0≤C m When <As, a yellow alert is activated, and the use of auxiliary fixing devices is recommended;

[0190] When C m When As is less than 0, a red alert is triggered, and construction of the current plan is prohibited;

[0191] Where As0 and As are the feasibility classification thresholds, and Ω is the comprehensive constraint coefficient;

[0192] The formula converts the distance between the installation point and the obstacle and the radius of curvature of the pipeline into a quantifiable feasibility of construction, avoiding the bias of subjective experience judgment. Based on the feasibility value, yellow and red warnings are triggered to clearly guide construction personnel to take countermeasures, such as using auxiliary devices or prohibiting construction, thereby improving project safety. By flexibly setting the safety threshold, it can adapt to the engineering needs of different pipeline sizes and optical fiber materials, identify risks such as insufficient pipeline bending or obstacles being too close in advance, and avoid damage to optical fibers due to excessive bending or collision.

[0193] If you need to have a diameter D p Fiber optic sensors are installed on steam pipelines with a strength of 0.5.

[0194] Parameter measurement:

[0195] Distance D between installation point and obstacle a =0.3, Pipe curvature radius R c =0.8.

[0196] Minimum bending radius R of optical fiber min =0.1 The safety factor is set as κ=0.2 and λ=2.

[0197] Calculate the safety threshold:

[0198] δ = 0.2 × 0.5 = 0.1 meters, ρ = 2 × 0.1 = 0.2 meters;

[0199]

[0200] Judgment and Response:

[0201] Preset threshold, A s0 =1.0, A s =1.5;

[0202] Because C m ≈0.845 s0 This triggers a red alert, prohibiting construction of the current plan. The system suggests adjusting the installation location or using a flexible fixing bracket.

[0203] Optimized and recalculated:

[0204] Adjust the installation point to D a =0.5 meters, R c = 1.0 meter;

[0205] ​

[0206] C m =1.0 satisfies As0≤C m <A s This triggers a yellow alert; it is recommended to use auxiliary fixing devices before construction.

[0207] Traditional installation methods rely on manual experience and may overlook subtle risks such as pipe bends or obstruction distances. This method uses quantitative calculations to identify red risks during the initial installation, preventing fiber optic cable breakage or data distortion caused by forced construction. Adjusted yellow warning prompts additional measures, balancing safety and construction efficiency.

[0208] The installation location was also dynamically optimized. The specific process for dynamic optimization of the installation location is as follows:

[0209] First, construct the objective function: min[ω1·(Δσ)] 2 +ω2·ΔT+ω3·ξ];

[0210] in Let ξ be the stress gradient, ΔT be the temperature range, ξ be the vibration disturbance factor, and D be the stress gradient. s Install distance reference values ​​for the sensor, D p The diameter of the pipe;

[0211] ω1, ω2 and ω3 are weighting coefficients;

[0212] Define geometric constraints:

[0213] Installation distance is sufficient, k1D s ≤Installation distance≤k2D s ;

[0214] Among them, D s The reference installation distance is k1 and k2, which are proportional coefficients based on the structural strength of the valve.

[0215] The radius of curvature of the pipeline satisfies R c ≥k3Dp;

[0216] Among them, D p Where k is the pipe diameter, and k3 is the safety factor (calibrated based on the bending characteristics of optical fiber);

[0217] Optimization algorithm solution: Iterative solution is obtained through heuristic optimization algorithms (such as genetic algorithms or particle swarm optimization). The algorithm parameters include population size, number of iterations and convergence threshold, which are configured according to the actual working conditions.

[0218] Through mathematical modeling, the effects of stress gradient, temperature fluctuation and vibration interference are scientifically balanced to avoid measurement errors caused by improper sensor installation.

[0219] For example, at the flange connection of a high-pressure valve, the optimized sensor can more accurately capture local stress concentrations and avoid missing tiny cracks.

[0220] Set geometric constraints, such as installation distance range and lower limit of radius of curvature, to ensure that sensor deployment complies with engineering safety standards and reduce fiber breakage or data distortion caused by improper installation.

[0221] In areas with insufficient bending radius, the system automatically prohibits installation to prevent damage to the optical fiber due to excessive bending.

[0222] Dynamic optimization algorithms, such as genetic algorithms, can be used to adapt to transient changes in valve operating status in real time, such as sudden pressure increases or temperature fluctuations.

[0223] When the valve vibrates more intensely due to fluid impact, the system automatically adjusts the sensor position to avoid areas of high interference.

[0224] A tiered deployment strategy, combining main sensor arrays and auxiliary sensors, balances monitoring density and cost, reducing redundant deployments.

[0225] In low-risk areas, such as the top of the valve body, the number of sensors can be reduced, and data can be compensated only through interpolation, thus saving hardware costs.

[0226] If a steam pipeline valve is subjected to high temperatures (150℃) and periodic fluid impacts for extended periods, a fiber optic grating sensor needs to be deployed to monitor stress distribution. Traditional experience-based installation leads to the following problems:

[0227] The sensors are concentrated near the flange, ignoring the influence of the temperature gradient in the middle of the valve body. Some sensors frequently fail due to vibration interference, and blind areas (such as the top of the valve) cannot be covered.

[0228] Construct the objective function: min[0.5(Δσ)² + 0.3ΔT + 0.2ξ];

[0229] Set geometric constraints: Installation distance: 0.8D s ≤Installation location≤1.2D s (Ds=0.5m);

[0230] Pipe curvature radius: R c ≥3D p (Dp=0.3m);

[0231] A genetic algorithm (population size 100, 50 iterations) was used to select the optimal candidate points:

[0232] Main sensor: At the flange connection (Δσ=300MPa / m), deployed at 20cm intervals;

[0233] Auxiliary sensors: located in the middle of the valve body (ΔT=50℃), deployed at a spacing of 40cm4;

[0234] Blind zone: Virtual monitoring points are generated at the top of the valve using cubic spline interpolation.

[0235] This enables the detection of abnormal temperature gradients (ΔT = 80℃) in the middle of the valve body, detects gasket aging earlier than traditional methods, reduces the number of sensors used, and lowers the failure rate. It solves the problems of strong reliance on experience, incomplete coverage, and weak anti-interference ability in traditional sensor installation, and significantly improves the reliability, economy, and intelligence level of industrial valve monitoring systems.

[0236] The operation process of the autonomous calibration system is as follows:

[0237] Real-time calculation of the current data distribution P t KL divergence with the initial distribution P0:

[0238]

[0239] In the formula, P t P(x) is the probability mass function of the current data distribution. It represents the probability of data point x appearing at system runtime t. P0(x) is the probability mass function of the initial data distribution, representing the probability of data point x appearing during the initial calibration phase (e.g., before commissioning).

[0240] When D KL When the dynamic threshold is exceeded (increasing with runtime), the following action is triggered:

[0241] Collect historical data and filter outliers;

[0242] Optimize the parameters of the corrosion assessment and life prediction model;

[0243] Update the weights of the deep learning model;

[0244] If calibration fails, activate the backup model and request manual intervention;

[0245] Assuming the temperature data of the valve monitoring system initially follows a normal distribution N(25℃, 2℃) at the P0 level, after running for a period of time, the current distribution P... t It becomes N (30℃, 5℃);

[0246] The KL divergence is calculated as follows:

[0247] Simplified calculation for univariate variables (temperature only):

[0248] If D KL =1.2 indicates a significant shift in temperature distribution, which may be due to valve aging or environmental changes;

[0249] The system will trigger a calibration process to update the scaling factors λ1 and λ2 in the corrosion assessment model;

[0250] KL divergence is a core indicator of autonomous calibration systems, used to dynamically determine whether the model needs to be updated.

[0251] Triggering condition: When DKL>θ(1+νt), calibration is started, where θ is the initial threshold and ν is the time decay coefficient;

[0252] The calibration results are used to inversely optimize the modulation frequency f and structured light compensation coefficient k of the multispectral imaging module;

[0253] Calibrated data distribution P t It will be synchronized to the digital twin platform to correct the elastic modulus E and thermal expansion coefficient αt in the multiphysics coupling model, improve simulation accuracy, and thus ensure the system continues to operate stably in complex industrial environments.

[0254] The system evaluates valve performance using the environmental attenuation factor η:

[0255] η=η0·exp[-(α·T 2 +β·Dust+γ·RH)], where T is the ambient temperature, Dust is the dust concentration, and RH is the relative humidity;

[0256] A, β, and γ are environmental coupling coefficients calibrated experimentally;

[0257] η0 is the nominal performance parameter of the valve.

[0258] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0259] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0260] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A valve intelligent monitoring system based on industrial vision, characterized in that, include: The multispectral imaging module includes visible light, thermal infrared and short-wave infrared sensors, which synchronously acquire three-dimensional morphological data of the valve at a preset sampling rate, and process the axial resolution by calculating the light speed, modulation frequency and phase difference through a three-dimensional reconstruction method, and suppress the interference of ambient light and mechanical vibration. The corrosion assessment module is used to calculate the corrosion index based on the difference in multispectral reflectance and the intensity of characteristic spectral peaks, determine the corrosion level according to the preset threshold, and trigger maintenance commands. The predictive maintenance module is used to dynamically correct the remaining lifetime prediction model through multi-source feature parameters, divide the lifetime into stages, and adjust the monitoring strategy. The autonomous calibration system calculates the KL divergence of the data distribution in real time and triggers model updates to suppress performance drift. A distributed fiber optic strain sensor array is deployed in stages along the main stress direction of the valve. The installation position is optimized by combining stress gradient, temperature fluctuation and construction feasibility, and the monitoring blind zone is compensated by interpolation. The millimeter-wave vibration monitoring unit captures the vibration spectrum of the target frequency band in a non-contact manner, avoiding installation interference; Edge computing units run hardware-accelerated deep learning models to enable localized real-time analysis; The cloud-based digital twin platform integrates real-time fluid simulation with a historical database to perform multiphysics coupling analysis.

2. The valve intelligent monitoring system based on industrial vision according to claim 1, characterized in that: The implementation process of the three-dimensional reconstruction method is as follows: In the formula: Δz is the axial resolution; c is the speed of light, and f is the preset modulation frequency; Where is the time-of-flight phase difference, and k is the structured light compensation coefficient; d struct The spacing of the structured light stripes.

3. The valve intelligent monitoring system based on industrial vision according to claim 1, characterized in that: The multiphysics coupling process of the cloud-based digital twin platform is as follows: In the formula, σ eq The stress is the equivalent stress, where T is the current temperature and E is the elastic modulus of the valve material. α t T0 is the coefficient of thermal expansion, and T0 is the reference temperature. ε max Where is the maximum strain value, and C is the strain correction factor.

4. The valve intelligent monitoring system based on industrial vision according to claim 1, characterized in that: The corrosion assessment module calculates the corrosion index based on the difference in multispectral reflectance and the intensity of characteristic spectral peaks, determines the corrosion level according to a preset threshold, and triggers a maintenance command. The specific process is as follows: Calculate the corrosion index: R=λ1·(R S1 -R S2 )+λ2·∑(I peak ·oh); In the formula: R S1 and R S2 The reflectivity difference for the selected band; I peak The intensity of the characteristic spectral peak of corrosion is given by ω, which is a weighting coefficient. λ1 and λ2 are scaling factors calibrated through corrosion experiments; After obtaining the corrosion index R, the corrosion level is determined based on the corrosion index R and the thresholds Γ1, Γ2, and Γ3.

5. The valve intelligent monitoring system based on industrial vision according to claim 1, characterized in that: The predictive maintenance module dynamically corrects the remaining lifetime prediction model in the remaining lifetime prediction model using multi-source feature parameters. In the formula, t base The baseline lifetime is given by k, where k is the decay rate coefficient. S m For multi-source feature parameters, ω m For the corresponding feature weights; The specific process of dynamically correcting the remaining lifetime prediction model using multi-source feature parameters is as follows: When the multi-source feature parameter S is detected... m When a mutation occurs, dynamic lifetime correction is performed.

6. The valve intelligent monitoring system based on industrial vision according to claim 1, characterized in that: The optimization process for the installation location of the distributed fiber Bragg grating strain sensor is as follows: Calculate the location optimization index: In the formula: S s T is the stress sensitivity factor. g This represents the average temperature fluctuation. E i V is the electromagnetic interference coefficient. d This is the integral value of the vibration disturbance. C m For construction feasibility, τ and N are calibration constants; Sensors are deployed in stages according to the optimization index L, and blind spots are compensated by interpolation.

7. The valve intelligent monitoring system based on industrial vision according to claim 6, characterized in that: The construction feasibility C m The calculation process is as follows: Where: Da is the distance between the installation point and the obstacle, and Rc is the radius of curvature of the pipeline; δ and ρ are engineering safety threshold parameters; The feasibility assessment process is as follows: measure the installation environment parameters and obtain the real-time measurement values ​​of obstacle distance Da and pipeline curvature radius Rc; Query the preset security thresholds δ and ρ, where δ = κ·Dp and ρ = λ·Rmin; When the preset conditions are met, the area is determined to be a workable area; According to C m The size triggers a tiered warning.

8. The valve intelligent monitoring system based on industrial vision according to claim 6, characterized in that: Dynamic optimization of the installation location was also performed. The specific process of dynamic optimization of the installation location is as follows: First, construct the objective function: min[ω1(Δσ) 2 +ω2·ΔT+ω3·ξ]; in Let denoted as stress gradient, ΔT as temperature range, ξ as vibration disturbance factor, and D as the stress gradient. s Install distance reference values ​​for the sensor, D p The diameter of the pipe; ω1, ω2 and ω3 are weighting coefficients; After setting geometric constraints, an optimization algorithm is used to solve the problem. The algorithm is solved iteratively through a heuristic optimization algorithm. The algorithm parameters include one or more of the following: population size, number of iterations, and convergence threshold.

9. The valve intelligent monitoring system based on industrial vision according to claim 1, characterized in that: The operation process of the autonomous calibration system is as follows: Real-time calculation of the current data distribution P t KL divergence with the initial distribution P0: When D KL When the dynamic threshold is exceeded, the corresponding operation is triggered.

Images