Large-diameter pipeline flow dynamic detection method and system based on temperature gradient tracking

Abstract

The invention provides a large-diameter pipeline flow dynamic detection method and system based on temperature gradient tracking, and relates to the technical field of flow detection.The large-diameter pipeline flow dynamic detection method comprises the steps that multiple layers of temperature sensor arrays are arranged on the outer wall of a pipeline, and temperature field distribution data are obtained; determining a thermal fluid injection scheme based on the initial flow parameters; acquiring a temperature characteristic data set by adopting a space-time sampling control unit; utilizing a deep learning algorithm to reconstruct a thermal fluid motion track; and establishing a flow calculation model to obtain a real-time flow value. According to the invention, high-precision, real-time and reliable monitoring of the flow of the large-diameter pipeline can be realized, and the accuracy and stability of flow detection are improved.

Description

Technical Field

[0001] The present invention relates to flow detection technology, and in particular to a dynamic detection method and system for large-diameter pipeline flow based on temperature gradient tracking. Background Art

[0002] Large-diameter pipelines are widely used in industrial production and municipal engineering. Accurate measurement of their flow rates is of great significance for production control and economic accounting. Traditional flow measurement methods mainly include differential pressure flowmeters, electromagnetic flowmeters, and ultrasonic flowmeters, etc. These measurement methods have accumulated rich experience in practical applications and provided important technical support for industrial production. Among them, the tracer method, as a non-invasive flow measurement method, has the advantages of not affecting normal production and being convenient for installation and maintenance, and shows good application prospects in the flow measurement of large-diameter pipelines.

[0003] At present, the flow measurement technology based on temperature tracing mainly realizes flow measurement by injecting a tracer with a significant temperature difference from the fluid into the pipeline and tracking the movement of the tracer. This method has achieved certain effects in practical applications, but the existing technology still has the following main problems: Existing temperature field monitoring systems generally adopt fixed sampling time intervals and sampling point distribution schemes, and cannot adaptively adjust the monitoring strategy according to the dynamic changes of the flow state, resulting in insufficient spatio-temporal resolution of the collected data and affecting the measurement accuracy.

[0004] Traditional tracer trajectory reconstruction methods mainly rely on simple mathematical models and statistical analysis, and do not fully consider the complexity and non-linear characteristics of the temperature field evolution. In the case of a complex and changeable flow field, it is difficult to accurately capture the movement law of the tracer.

[0005] Existing flow calculation models are often too simplified and do not fully consider the influence of factors such as hydrodynamic effects and temperature effects on flow measurement. Especially in large-diameter pipelines, due to more complex flow characteristics, simplified models are difficult to ensure the accuracy of calculation results. Summary of the Invention

[0006] Embodiments of the present invention provide a dynamic detection method and system for large-diameter pipeline flow based on temperature gradient tracking, which can solve the problems in the existing technology.

[0007] In the first aspect of the embodiments of the present invention, A dynamic detection method for large-diameter pipeline flow based on temperature gradient tracking is provided, including: A temperature field monitoring system is provided on the outer wall of a large-diameter pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays arranged along the axial direction of the pipeline. Each layer of temperature sensor array contains multiple temperature sensors evenly distributed circumferentially. The temperature sensors are electrically connected to a central processing unit through a data transmission module; Obtain the temperature field distribution data of the pipeline under the initial operating state through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; according to the initial flow parameters, determine the injection scheme of the hot fluid, and realize the adaptive update of the flow parameters through the injection scheme. The injection scheme includes the injection temperature value, the injection flow rate value, and the injection duration, and inject the hot fluid into the large-diameter pipeline according to the injection scheme; A time-domain sampling control unit that automatically adjusts the sampling time interval according to the temperature change rate, and a spatial sampling control unit that automatically adjusts the sampling point distribution according to the temperature spatial gradient. The central processing unit integrates the collected temperature data into a temperature feature data set with spatio-temporal correlation; Construct a temperature field evolution model based on the temperature feature data set, extract the spatio-temporal features of the temperature field through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to achieve the precise reconstruction of the movement trajectory of the hot fluid and obtain the fluid velocity distribution of the pipeline cross-section; According to the fluid velocity distribution of the pipeline cross-section, establish a flow rate calculation model considering hydrodynamic effects. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow rate value of the fluid in the pipeline through adaptive iterative calculation; transmit the real-time flow rate value to an intelligent analysis terminal, and the intelligent analysis terminal performs dynamic evaluation and early warning analysis on the real-time flow rate value to achieve reliable monitoring of the flow rate of the large-diameter pipeline.

[0008] Obtain the temperature field distribution data of the pipeline under the initial operating state through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; according to the initial flow parameters, determine the injection scheme of the hot fluid, and realize the adaptive update of the flow parameters including: Obtain the temperature field distribution data of the pipeline under the initial operating state through the temperature field monitoring system provided on the outer wall of the pipeline. The temperature field monitoring system arranges multiple groups of temperature sensor arrays along the axial direction of the pipeline. Each group of temperature sensor arrays evenly distributes multiple temperature sensors along the circumferential direction of the pipeline. The temperature sensors are electrically connected to the central processing unit; The central processing unit performs feature analysis on the temperature field distribution data, extracts the radial temperature gradient feature and the axial temperature gradient feature, and establishes a temperature field feature vector including the temperature spatial distribution feature and the temperature time fluctuation feature; The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field eigenvector, determines the fluid flow pattern characteristic coefficient through temperature fluctuation spectrum analysis, calculates the average fluid velocity using temperature cross-correlation analysis, calculates the fluid turbulence intensity based on the temperature field distortion degree, and constructs a set of initial flow parameters of the fluid in the pipeline; Determine the injection scheme of the hot fluid according to the set of initial flow parameters. The injection scheme includes the injection temperature, injection flow rate, and injection timing of the hot fluid; During the injection of the hot fluid, the temperature field monitoring system continuously collects temperature signals. The central processing unit extracts the temperature field characteristics from the temperature signals and constructs a spatio-temporal eigenvector, calculates the fluid velocity distribution based on multi-point cross-correlation analysis, constructs a flow pattern characteristic index in combination with the temperature power spectral density, performs flow pattern recognition using a deep learning network, establishes a set of flow parameters and implements dynamic parameter correction to achieve the adaptive update of the flow parameters.

[0009] The temperature field monitoring system continuously collects temperature signals. The central processing unit extracts the temperature field characteristics from the temperature signals and constructs a spatio-temporal eigenvector, calculates the fluid velocity distribution based on multi-point cross-correlation analysis, constructs a flow pattern characteristic index in combination with the temperature power spectral density, performs flow pattern recognition using a deep learning network, establishes a set of flow parameters and implements dynamic parameter correction to achieve the adaptive update of the flow parameters, including: Obtain multi-point temperature field data, collect temperature signals through the temperature field monitoring system arranged on the outer wall of the pipeline. The temperature field monitoring system includes a multi-layer temperature sensor array evenly distributed along the axial direction of the pipeline. Each layer of the temperature sensor array is evenly provided with multiple temperature sensors along the circumferential direction of the pipeline. The temperature sensors are electrically connected to the data acquisition unit; Extract the temperature field characteristics, perform normalization preprocessing on the temperature signals, calculate the radial temperature gradient and the axial temperature gradient, construct a spatio-temporal eigenvector including the temperature fluctuation intensity and the temperature autocorrelation function. The temperature autocorrelation function adopts a weighted calculation method, and the weight coefficient is related to the amplitude of the temperature gradient; Calculate the fluid velocity distribution, establish a multi-point cross-correlation analysis model based on the spatio-temporal eigenvector. The multi-point cross-correlation analysis model determines the average fluid velocity by optimizing the time delay estimation, and calculates the radial velocity distribution using a piecewise continuous velocity distribution function. The velocity distribution function adopts different mathematical models in the wall region and the core region respectively; Construct a flow pattern characteristic index, calculate the temperature power spectral density using the temperature field characteristics, extract the characteristic frequency parameters, and establish a flow pattern discrimination criterion in combination with the temperature fluctuation intensity and the root mean square value of the temperature gradient. The flow pattern discrimination criterion adopts a weighted fusion method to comprehensively consider multiple characteristic parameters; Perform intelligent flow pattern recognition, input the flow pattern characteristic index into the deep learning network. The deep learning network includes a feature extraction layer and a classification output layer. The feature extraction layer adopts a convolutional structure, and the classification output layer gives the flow pattern probability distribution through the softmax function; Establish a set of flow parameters, calculate the turbulence intensity according to the flow pattern recognition result and velocity distribution, evaluate the flow stability in combination with the temperature field distortion degree, and construct a flow feature set including flow pattern coefficients, velocity distribution, and turbulence parameters; Implement dynamic parameter correction, calculate the parameter uncertainty based on the flow feature set, establish a parameter correction model, optimize the calculation accuracy of flow parameters through a feedback adjustment mechanism, and achieve the adaptive update of flow parameters.

[0010] Construct a temperature field evolution model based on the temperature feature data set, extract the spatio-temporal features of the temperature field through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to achieve the accurate reconstruction of the thermal fluid motion trajectory, and obtain the fluid velocity distribution of the pipe cross-section including: Construct a temperature field evolution model based on the temperature field feature data set. The temperature field evolution model uses a deep learning algorithm to extract the spatio-temporal features of the temperature field. The deep learning algorithm includes a convolutional neural network module for extracting the spatial distribution features of the temperature field and a recurrent neural network module for capturing the temporal variation features of the temperature field; Integrate a spatial attention mechanism and a temporal attention mechanism in the deep learning algorithm. The spatial attention mechanism assigns weights according to the importance of the spatial distribution features of the temperature field, and the temporal attention mechanism determines weights based on the correlation of the temporal evolution features of the temperature field; Use the temperature field evolution model to dynamically track the thermal fluid, establish a thermal fluid motion feature mapping relationship based on the outputs of the spatial attention mechanism and the temporal attention mechanism to achieve the accurate reconstruction of the thermal fluid motion trajectory; calculate the fluid velocity distribution of the pipe cross-section according to the thermal fluid motion trajectory.

[0011] Use the temperature field evolution model to dynamically track the thermal fluid, establish a thermal fluid motion feature mapping relationship based on the outputs of the spatial attention mechanism and the temporal attention mechanism to achieve the accurate reconstruction of the thermal fluid motion trajectory; calculate the fluid velocity distribution of the pipe cross-section including: Use the temperature field evolution model to dynamically track the thermal fluid, input the temperature field distribution data into the temperature field evolution model, and establish a thermal fluid motion feature mapping relationship based on the outputs of the spatial attention mechanism and the temporal attention mechanism; Reconstruct the thermal fluid motion trajectory according to the thermal fluid motion feature mapping relationship, iteratively calculate the position information of the thermal fluid through the state estimation equation to achieve the accurate reconstruction of the thermal fluid motion trajectory, and calculate the fluid velocity distribution of the pipe cross-section based on the reconstructed thermal fluid motion trajectory.

[0012] According to the fluid velocity distribution in the pipeline cross-section, a flow rate calculation model considering hydrodynamic effects is established. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow rate value of the fluid in the pipeline through adaptive iterative calculation, including: Establish a flow rate calculation model considering hydrodynamic effects. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and inputs the fluid velocity distribution in the pipeline cross-section into the flow rate calculation model; Determine the flow pattern characteristics based on the Reynolds number of fluid flow, pipe diameter ratio, and viscosity ratio, and calculate the flow pattern correction coefficient according to the flow pattern characteristics. The flow pattern correction coefficient is used to characterize the characteristics of fluid flow state transition; Determine the temperature effect characteristics based on the fluid temperature change characteristics and fluid physical property parameters. The temperature effect characteristics include a temperature ratio correction term and a density ratio correction term, and calculate the temperature effect correction coefficient according to the temperature effect characteristics; Determine the velocity field characteristics based on the fluid velocity distribution in the pipeline cross-section. The velocity field characteristics include a radial velocity distribution function and a maximum center velocity, and calculate the velocity distribution correction coefficient according to the velocity field characteristics; Construct a flow rate calculation optimization model including a flow rate fitting error term, a physical constraint term, and a smoothing constraint term. Establish a comprehensive objective function based on the flow pattern correction coefficient, temperature effect correction coefficient, and velocity distribution correction coefficient with an adaptive weight coefficient, perform iterative optimization calculation using a parameter optimization strategy with a dynamic learning rate, and verify the accuracy of the optimization result based on a comprehensive evaluation function including multi-dimensional error indicators; Perform adaptive iterative calculation on the optimization result of the flow rate calculation optimization model. When the relative flow rate error between adjacent iterative steps is less than the preset error threshold, the iteration is completed to obtain the real-time flow rate value of the fluid in the pipeline.

[0013] Construct a flow rate calculation optimization model including a flow rate fitting error term, a physical constraint term, and a smoothing constraint term. Establish a comprehensive objective function based on the flow pattern correction coefficient, temperature effect correction coefficient, and velocity distribution correction coefficient with an adaptive weight coefficient, perform iterative optimization calculation using a parameter optimization strategy with a dynamic learning rate, and verify the accuracy of the optimization result based on a comprehensive evaluation function including multi-dimensional error indicators, including: Construct a flow rate calculation optimization model. The flow rate calculation optimization model includes a flow rate fitting error term, a physical constraint term, and a smoothing constraint term. The flow rate fitting error term is used to characterize the deviation between the calculated flow rate and the reference flow rate. The physical constraint term includes a mass conservation constraint and a momentum balance constraint. The smoothing constraint term includes a time series change constraint and a velocity gradient constraint; Establish a comprehensive objective function, combine the flow rate fitting error term, physical constraint term, and smoothing constraint term through an adaptive weight coefficient. The adaptive weight coefficient is dynamically adjusted based on the error contribution degree of each item, and the comprehensive objective function outputs an optimization target value; Construct an optimization strategy for parameters, calculate the parameter gradient based on the comprehensive objective function, update and optimize the parameters using a dynamic learning rate, where the dynamic learning rate decays exponentially with the number of iteration steps, and the parameter gradient is used to guide the parameter update direction; Execute iterative optimization calculations. Input the optimized parameters into the flow calculation optimization model, and judge the iteration termination condition based on the output of the comprehensive objective function. When the parameter change amount between adjacent iteration steps is less than the preset change threshold, the iteration is completed; construct a comprehensive evaluation function containing multi-dimensional error indicators, calculate the optimization convergence speed according to the output of the comprehensive evaluation function, and verify the accuracy of the optimization result.

[0014] In the second aspect of the embodiments of the present invention, a large-diameter pipeline flow dynamic detection system based on temperature gradient tracking is provided, including: The first unit is used to set up a temperature field monitoring system on the outer wall of the large-diameter pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays arranged along the axial direction of the pipeline. Each layer of temperature sensor array contains multiple temperature sensors evenly distributed circumferentially. The temperature sensors are electrically connected to the central processing unit through a data transmission module; The second unit is used to obtain the temperature field distribution data under the initial operating state of the pipeline through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; according to the initial flow parameters, determine the injection scheme of the hot fluid, and realize the adaptive update of the flow parameters through the injection scheme. The injection scheme includes the injection temperature value, the injection flow value, and the injection duration, and inject the hot fluid into the large-diameter pipeline according to the injection scheme; The third unit is a time-domain sampling control unit that automatically adjusts the sampling time interval according to the temperature change rate, and a spatial sampling control unit that automatically adjusts the sampling point distribution according to the temperature spatial gradient. The central processing unit integrates the collected temperature data into a spatio-temporally correlated temperature feature dataset; The fourth unit is used to construct a temperature field evolution model based on the temperature feature dataset, extract the spatio-temporal features of the temperature field through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to accurately reconstruct the movement trajectory of the hot fluid and obtain the fluid velocity distribution of the pipeline cross-section; The fifth unit is used to establish a flow calculation model considering fluid mechanics effects according to the fluid velocity distribution of the pipeline cross-section. The flow calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow value of the fluid in the pipeline through adaptive iterative calculations; transmit the real-time flow value to the intelligent analysis terminal, and the intelligent analysis terminal performs dynamic evaluation and early warning analysis on the real-time flow value to achieve reliable monitoring of the large-diameter pipeline flow.

[0015] In the third aspect of the embodiments of the present invention Provided is an electronic device, including: a processor; a memory for storing instructions executable by the processor; wherein, the processor is configured to call the instructions stored in the memory to execute the foregoing method.

[0016] In the fourth aspect of the embodiments of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the foregoing method is implemented.

[0017] The beneficial effects of this application are as follows: The dynamic flow detection method for large-diameter pipelines based on temperature gradient tracking provided by the present invention has the following beneficial effects: By setting up a multi-layer temperature sensor array on the outer wall of the pipeline to construct a temperature field monitoring system, combining an adaptive injection scheme and intelligent sampling control, high-precision real-time monitoring of the pipeline temperature field distribution is achieved, providing reliable basic data support for flow calculation, and effectively overcoming the limitations of traditional flow detection methods in the application of large-diameter pipelines.

[0018] Using a deep learning algorithm to extract and analyze the spatio-temporal characteristics of the temperature field, integrating a spatial attention mechanism and a temporal attention mechanism, the motion trajectory of the hot fluid can be accurately reconstructed, the fluid velocity distribution of the pipeline cross-section can be obtained, and the accuracy and reliability of flow detection are improved.

[0019] By establishing a flow calculation model considering hydrodynamic effects, introducing multiple correction factors for adaptive iterative calculation, and combining the dynamic evaluation and early warning analysis functions of an intelligent analysis terminal, real-time and accurate monitoring of the flow rate of large-diameter pipelines is achieved, providing an important guarantee for the safe operation of pipelines. Description of the Drawings

[0020] Figure 1 It is a schematic flow chart of the dynamic flow detection method for large-diameter pipelines based on temperature gradient tracking according to the embodiments of the present invention; Figure 2 It is a schematic diagram of the radial velocity distribution profile according to the embodiments of the present invention; Figure 3 It is a schematic diagram for comparing the extraction accuracy of spatio-temporal distribution characteristics of the temperature field according to the embodiments of the present invention; Figure 4 It is a schematic diagram for comparing the position errors of different methods according to the embodiments of the present invention; Figure 5 It is a schematic diagram for analyzing the convergence of correction coefficients according to the embodiments of the present invention. Detailed Embodiments

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0022] The following uses specific embodiments to elaborate on the technical solutions of the present invention in detail. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0023] Figure 1 The flowchart of the dynamic detection method for the flow rate of a large-diameter pipeline based on temperature gradient tracking in the embodiments of the present invention is shown as Figure 1 shown, and the method includes: A temperature field monitoring system is arranged on the outer wall of the large-diameter pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays arranged along the axial direction of the pipeline. Each layer of temperature sensor array contains multiple temperature sensors evenly distributed circumferentially. The temperature sensors are electrically connected to the central processing unit through a data transmission module; Obtain the temperature field distribution data under the initial operating state of the pipeline through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; according to the initial flow parameters, determine the injection scheme of the hot fluid, and realize the adaptive update of the flow parameters through the injection scheme. The injection scheme includes the injection temperature value, the injection flow rate value, and the injection duration, and inject the hot fluid into the large-diameter pipeline according to the injection scheme; A time-domain sampling control unit that automatically adjusts the sampling time interval according to the temperature change rate, and a spatial sampling control unit that automatically adjusts the sampling point distribution according to the temperature spatial gradient. The central processing unit integrates the collected temperature data into a spatio-temporal correlated temperature feature dataset; Construct a temperature field evolution model based on the temperature feature dataset, extract the spatio-temporal features of the temperature field through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to accurately reconstruct the movement trajectory of the hot fluid and obtain the fluid velocity distribution of the pipeline cross-section; According to the fluid velocity distribution of the pipeline cross-section, establish a flow rate calculation model considering the hydrodynamic effect. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow rate value of the fluid in the pipeline through adaptive iterative calculation; transmit the real-time flow rate value to the intelligent analysis terminal, and the intelligent analysis terminal performs dynamic evaluation and early warning analysis on the real-time flow rate value to realize reliable monitoring of the flow rate of the large-diameter pipeline.

[0024] The "hot fluid" in the claims refers to, during the flow measurement process, by setting an electric heating device upstream of the measurement section of the fluid in the pipeline, the temperature of the fluid flowing through the heating device is increased, so as to form a region with a temperature difference from the surrounding fluid in the measurement section. This fluid with temperature difference continues to flow with the mainstream fluid in the pipeline, and the distribution characteristics of its temperature field can be detected by the temperature sensor system downstream. By tracking the propagation process of the temperature field, the flow measurement is realized. Therefore, the "hot fluid" actually refers to the part of the fluid with temperature difference after being heated by the heating device. This temperature difference is used as a tracer feature to achieve the flow measurement.

[0025] In an alternative embodiment, the temperature field distribution data of the pipeline under the initial operating state is obtained through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; according to the initial flow parameters, the injection scheme of the hot fluid is determined, and the adaptive update of the flow parameters is realized through the injection scheme, including: The temperature field distribution data of the pipeline under the initial operating state is obtained through the temperature field monitoring system arranged on the outer wall of the pipeline. The temperature field monitoring system arranges multiple groups of temperature sensor arrays along the axial direction of the pipeline, and each group of temperature sensor arrays evenly distributes multiple temperature sensors along the circumferential direction of the pipeline. The temperature sensors are electrically connected to the central processing unit; The central processing unit performs feature analysis on the temperature field distribution data, extracts the radial temperature gradient feature and the axial temperature gradient feature, and establishes a temperature field feature vector including the spatial distribution feature and the temporal fluctuation feature of the temperature; The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field feature vector, determines the fluid flow pattern characteristic coefficient through temperature fluctuation spectrum analysis, calculates the average fluid velocity through temperature cross-correlation analysis, calculates the fluid turbulence intensity based on the temperature field distortion degree, and constructs the initial flow parameter set of the fluid in the pipeline; The injection scheme of the hot fluid is determined according to the initial flow parameter set, and the injection scheme includes the injection temperature, injection flow rate and injection timing of the hot fluid; During the injection process of the hot fluid, the temperature field monitoring system continuously collects temperature signals. The central processing unit extracts the temperature field features from the temperature signals and constructs a spatio-temporal feature vector, calculates the fluid velocity distribution based on multi-point cross-correlation analysis, constructs a flow pattern characteristic index in combination with the temperature power spectral density, performs flow pattern recognition using a deep learning network, establishes a flow parameter set and implements dynamic parameter correction to realize the adaptive update of the flow parameters.

[0026] The temperature field distribution data of the pipeline under the initial operating state is obtained by the temperature field monitoring system arranged on the outer wall of the pipeline. The temperature field monitoring system arranges a group of temperature sensor arrays every 0.5 meters along the axial direction of the pipeline, and each group of temperature sensor arrays is evenly arranged with 8 temperature sensors along the circumferential direction of the pipeline. PT100 type platinum resistance temperature sensors are used, with a temperature measurement range of -50°C to 200°C and a measurement accuracy of ±0.1°C. The temperature sensors are electrically connected to the central processing unit through a signal conditioning circuit, and the sampling frequency is 100Hz.

[0027] The central processing unit conducts feature analysis on the temperature field distribution data. First, the radial temperature gradient feature is extracted, and the temperature difference between the inner and outer surfaces of the pipe wall is calculated to reflect the heat transfer intensity between the fluid inside the pipe and the pipe wall. For a carbon steel pipe with an inner diameter of 200mm, when the fluid temperature is 80°C, the temperature difference between the inner and outer surfaces of the pipe wall is usually between 3°C and 5°C. Combining the fluctuation characteristics of the temperature time series, a temperature field feature vector is established.

[0028] Based on the temperature field feature vector, the initial flow parameters of the fluid inside the pipeline are calculated. By performing a fast Fourier transform on the temperature signal, the temperature fluctuation spectrum is obtained, and the amplitude and frequency of the main frequency component can be used to characterize the flow pattern characteristics. In the laminar flow state, the fluctuation frequency is relatively low, at 0.1 - 10Hz. The average fluid velocity is calculated using the cross-correlation analysis of the temperature signals at adjacent measurement points. For a DN200 pipeline, the typical flow velocity value is 0.5 - 2m / s. The fluid turbulence intensity is calculated based on the temperature field distortion degree. The greater the distortion degree, the higher the turbulence intensity.

[0029] According to the initial flow parameters, the injection scheme of the hot fluid is determined. The hot fluid uses the same medium as the measured fluid, and the injection temperature is determined to be 20 - 5% higher than the mainstream fluid temperature. The pulse injection method is adopted, and the injection duration is 2 - 60s.

[0030] During the injection process of the hot fluid, parameter adaptive update is continuously carried out. The temperature field monitoring system collects temperature signals at a sampling frequency of 100Hz. The central processing unit extracts the temperature field features and constructs a spatio-temporal feature vector, including feature quantities such as temperature peak value, rise time, and peak duration. The velocity distribution of the fluid at different radial positions is calculated through multi-point cross-correlation analysis. Combining the temperature power spectral density to analyze the flow pattern characteristics, a feature set including spectral energy distribution, main frequency components, etc. is established. A 5-layer convolutional neural network is used to perform flow pattern recognition to achieve accurate discrimination of laminar flow, turbulent flow, and transitional states. Finally, a flow parameter set including velocity field, flow pattern characteristics, and turbulence intensity is established, and the parameters are updated every 1 minute.

[0031] The temperature field monitoring system is used to obtain the pipeline temperature distribution data. Through multi-dimensional feature extraction and analysis, the non-destructive on-line measurement of the flow parameters inside the pipe is realized, avoiding the leakage risk caused by the need for opening installation of traditional flow meters and improving the safety of system operation. Based on the tracer hot fluid injection technology, the mapping relationship between the dynamic response characteristics of the temperature field and the flow parameters is established, which has higher measurement accuracy and a wider applicable range compared with conventional measurement methods. By reasonably designing the injection scheme, the influence of the tracer on the physical properties of the main fluid can be effectively avoided. The deep learning network is used to realize the flow pattern recognition and parameter adaptive update, improving the adaptability of the system to changes in flow conditions and the measurement reliability. By establishing a complete parameter closed-loop control system, the real-time performance and accuracy of the measurement results are ensured. The measurement accuracy can reach ±2%, which is applicable to various fluid transportation pipelines with a pipe diameter in the range of 50 - 500 mm.

[0032] In an alternative embodiment, the temperature field monitoring system continuously collects temperature signals. The central processing unit extracts the temperature field features from the temperature signals and constructs a spatio-temporal feature vector. Based on the multi-point cross-correlation analysis, the fluid velocity distribution is calculated. Combining the temperature power spectral density, the flow pattern characteristic index is constructed. The deep learning network is used to perform the flow pattern recognition, establish the flow parameter set and implement the dynamic parameter correction. The adaptive update of the flow parameters includes: Obtain multi-point temperature field data. The temperature signals are collected by the temperature field monitoring system arranged on the outer wall of the pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays evenly distributed along the axial direction of the pipeline. Each layer of temperature sensor array is evenly provided with multiple temperature sensors along the circumferential direction of the pipeline. The temperature sensors are electrically connected to the data acquisition unit; Extract the temperature field features. The temperature signals are preprocessed by normalization. The radial temperature gradient and the axial temperature gradient are calculated. A spatio-temporal feature vector including the temperature fluctuation intensity and the temperature autocorrelation function is constructed. The temperature autocorrelation function adopts a weighted calculation method, and the weight coefficient is related to the amplitude of the temperature gradient; Calculate the fluid velocity distribution. Based on the spatio-temporal feature vector, a multi-point cross-correlation analysis model is established. The multi-point cross-correlation analysis model determines the average fluid velocity by optimizing the time-delay estimation. The radial velocity distribution is calculated by using a piecewise continuous velocity distribution function, and different mathematical models are adopted for the velocity distribution function in the wall region and the core region respectively; Construct the flow pattern characteristic index. The temperature power spectral density is calculated by using the temperature field features, the characteristic frequency parameters are extracted, and the flow pattern discrimination criterion is established by combining the temperature fluctuation intensity and the root mean square value of the temperature gradient. The flow pattern discrimination criterion adopts a weighted fusion method to synthesize multiple characteristic parameters; Perform intelligent flow pattern recognition. The flow pattern characteristic index is input into the deep learning network. The deep learning network includes a feature extraction layer and a classification output layer. The feature extraction layer adopts a convolutional structure, and the classification output layer gives the flow pattern probability distribution through the softmax function; Establish a set of flow parameters, calculate the turbulence intensity based on the flow pattern recognition result and velocity distribution, evaluate the flow stability in combination with the temperature field distortion degree, and construct a flow feature set including flow pattern coefficients, velocity distribution, and turbulence parameters; Implement dynamic parameter correction, calculate the parameter uncertainty based on the flow feature set, establish a parameter correction model, optimize the calculation accuracy of flow parameters through a feedback adjustment mechanism, and achieve the adaptive update of flow parameters.

[0033] Obtain multi-point temperature field data through the temperature field monitoring system. On the outer wall of a pipe with an inner diameter of 200 mm, a temperature sensor array is arranged every 0.3 meters along the axial direction, with a total of 10 layers of measuring points. Each layer of the array is evenly arranged with 12 PT100 platinum resistance temperature sensors along the circumferential direction of the pipe. The temperature measurement range is -50°C to 200°C, and the measurement accuracy is ±0.1°C. All sensors are connected to the data acquisition unit through the RS485 bus, and the sampling frequency is set to 200 Hz.

[0034] Extract the characteristics of the collected temperature signals. First, normalize the temperature data to the 0-1 interval to eliminate the influence of dimensions. Calculate the radial temperature gradient between adjacent measuring points. Under typical working conditions, the radial temperature gradient is between 0.5 and 0.3 °C / cm. Statistically analyze the amplitude of temperature pulsation. The pulsation intensity is less than 0.5 °C in the laminar flow state and can reach 1 to 3 °C in the turbulent flow state. When calculating the temperature autocorrelation function, determine the weight coefficient according to the magnitude of the temperature gradient. The greater the gradient, the higher the weight, and the maximum weight can reach 0.8.

[0035] Calculate the fluid velocity distribution based on the spatio-temporal feature vector. Adopt the multi-point cross-correlation analysis method, optimize the time-delay estimation through a sliding time window, and the window length is 2 seconds. For a DN200 pipe, the average fluid velocity is in the range of 0.5 to 2 m / s. The radial velocity distribution is described by a piecewise function. The logarithmic distribution is used in the wall region, and the power-law distribution is used in the core region. The transition point between the two regions is located 2 cm away from the wall.

[0036] Construct a flow pattern feature index system. Conduct power spectrum analysis on the temperature signal, and extract 0.1 - 1 Hz. The main frequency in the turbulent flow state can reach 5 - 10 Hz. Establish a comprehensive criterion by combining the temperature pulsation intensity and the root mean square value of the temperature gradient, and fuse multiple feature parameters with a weight ratio of 0.6:0.4.

[0037] Use a deep learning network to perform flow pattern recognition. The network includes 3 convolutional layers and 2 fully connected layers, and the size of the convolutional kernel is 3×3. The input feature dimension is 128, and the number of neurons in the intermediate layers is 64, 32, and 16. Finally, the probability distributions of the laminar flow, transitional flow, and turbulent flow are output through the softmax function.

[0038] Establish a complete set of flow parameters. Determine the flow pattern coefficient according to the flow pattern identification result, with a value of 0.2 - 0.6 for laminar flow and 0.6 - 0.8 for turbulent flow. Calculate the turbulence intensity in combination with the velocity distribution, and the numerical range is 5% - 20%. Evaluate the flow stability through the temperature field distortion degree, and a distortion degree less than 5% indicates stable flow.

[0039] Implement dynamic parameter correction. Calculate the uncertainties of the velocity and flow pattern parameters, and control the velocity measurement uncertainty within ±2%. Establish a feedback-based parameter correction model, and determine that the parameters converge when the correction amplitude is less than 1% for three consecutive times. The parameter update period is 1 minute, enabling real-time adaptive update of the flow parameters.

[0040] Figure 2 Schematic diagram of the radial velocity distribution profile for the embodiment of the present invention: The velocity distribution profile clearly shows the performance differences of different measurement methods. The velocity distribution curve measured by this technical solution is closest to the theoretical calculated value. At the dimensionless radial position of 0.2, the relative velocity is 0.95, while that of the traditional velocity measurement method is 0.90; at 0.4, the value measured by this technical solution is 0.85, and the traditional velocity measurement method is 0.75; the difference at 0.6 is further enlarged, being 0.70 and 0.55 respectively. Especially in the near-wall region (r / R = 0.8), the relative velocity measured by this technical solution is 0.50, significantly higher than 0.30 of the traditional velocity measurement method and closer to the theoretical value of 0.20, demonstrating the excellent measurement performance of this technical solution in the high velocity gradient region. The smoothness and continuity of the overall profile are also significantly better than those of the traditional velocity measurement method, indicating higher reliability of the measurement results.

[0041] There are obvious deficiencies in the traditional velocity measurement methods in the prior art for measuring the velocity distribution of pipeline fluids. The traditional velocity measurement method mainly relies on single-point measurement and simple interpolation calculation, and it is difficult to accurately capture the characteristics of the velocity gradient change of the fluid in the radial direction, especially in high velocity gradient regions such as near the wall surface, where the measurement accuracy significantly decreases. At the same time, due to the lack of a real-time identification and correction mechanism for the flow state, the traditional method cannot adapt to the dynamic changes of the flow conditions, resulting in a large deviation between the measurement result and the actual flow characteristics.

[0042] From the perspective of improving the measurement accuracy of the velocity distribution, this technical solution innovatively uses a multi-point temperature field monitoring system to obtain the fluid motion characteristics, realizes the intelligent identification of the flow pattern through a deep learning network, and establishes a dynamic correction mechanism based on multi-dimensional parameters. This solution achieves a comprehensive perception of the fluid motion characteristics through the accurate extraction of the spatio-temporal characteristics of the temperature field; uses deep learning algorithms to identify and classify the flow pattern characteristics, improving the pertinence of the velocity distribution calculation; adopts an adaptive parameter correction strategy, enabling the measurement result to be adjusted in real time according to the flow state, thereby maintaining a high accuracy level throughout the measurement process.

[0043] Through experimental verification, this technical solution has significant technical effects in terms of velocity distribution measurement: First, the measured velocity distribution curve is highly consistent with the theoretically calculated value, reflecting the accuracy of the measurement results; Second, good measurement accuracy can still be maintained in the near-wall region with high velocity gradients, overcoming the measurement bottleneck of traditional methods in these regions; Finally, the velocity distribution curve exhibits excellent smoothness and continuity, indicating that the measurement results have high reliability. The realization of these technical effects provides a new technical approach for the accurate measurement of the velocity distribution of pipeline fluids.

[0044] In an alternative embodiment, a temperature field evolution model is constructed based on the temperature feature dataset, and the spatio-temporal features of the temperature field are extracted through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to achieve the precise reconstruction of the thermal fluid motion trajectory, and the fluid velocity distribution of the pipeline cross-section is obtained, including: A temperature field evolution model is constructed based on the temperature field feature dataset. The temperature field evolution model uses a deep learning algorithm to extract the spatio-temporal features of the temperature field. The deep learning algorithm includes a convolutional neural network module for extracting the spatial distribution features of the temperature field and a recurrent neural network module for capturing the temporal variation features of the temperature field; In the deep learning algorithm, a spatial attention mechanism and a temporal attention mechanism are integrated. The spatial attention mechanism assigns weights according to the importance of the spatial distribution features of the temperature field, and the temporal attention mechanism determines weights based on the correlation of the temporal evolution features of the temperature field; The temperature field evolution model is used to dynamically track the thermal fluid, and a motion feature mapping relationship of the thermal fluid is established based on the outputs of the spatial attention mechanism and the temporal attention mechanism to achieve the precise reconstruction of the thermal fluid motion trajectory; The fluid velocity distribution of the pipeline cross-section is calculated according to the thermal fluid motion trajectory.

[0045] First, a temperature field feature dataset is constructed by collecting the temperature field data of the fluid in the pipeline. Multiple temperature measurement points are arranged on the pipeline cross-section through a thermocouple array, with a spacing of 5 mm between each temperature measurement point, a sampling frequency of 100 Hz, and a collection duration of 60 s to obtain the spatio-temporal distribution data of the temperature field. The collected temperature data is preprocessed, including operations such as denoising and normalization, to construct a standardized temperature field feature dataset.

[0046] Construct a temperature field evolution model based on the processed dataset. The model includes a convolutional neural network module and a recurrent neural network module. The convolutional neural network adopts a 5-layer structure, including 3 convolutional layers and 2 pooling layers. The size of the convolutional kernel in the first layer is 3×3, the stride is 1, and 32 feature maps are output; the size of the convolutional kernel in the second layer is 3×3, the stride is 1, and 64 feature maps are output; the size of the convolutional kernel in the third layer is 3×3, the stride is 1, and 128 feature maps are output. The pooling kernels of the two max-pooling layers are both 2×2. The recurrent neural network adopts the long short-term memory network (LSTM) structure, including 128 hidden units.

[0047] Integrate the attention mechanism into the deep learning model. The spatial attention mechanism assigns different weights to features at different positions by calculating the importance scores of features at each spatial position. In the specific implementation, average pooling and max pooling are performed on the convolutional feature map in the channel dimension to obtain two feature maps, which are concatenated and then passed through a 7×7 convolutional layer to obtain the spatial attention weight map. The temporal attention mechanism determines the weights by calculating the correlation of features at different time steps. For the hidden state sequence output by the LSTM, the similarity score between the query vector and the key-value pair is calculated, and the temporal attention weight is obtained after normalization by the softmax function.

[0048] Use the trained model to dynamically track the thermal fluid. Inject a tracer thermal fluid with a temperature of 80°C into the pipeline, and the background fluid temperature is 20°C. The input of the model is the temperature field data at consecutive moments. After spatial-temporal feature extraction and attention weighting, the centroid position coordinates of the thermal fluid are output. The movement trajectory of the thermal fluid is obtained through continuous prediction, and then the fluid velocity distribution on the pipeline cross-section is calculated. The experimental results show that the average error between the predicted trajectory and the actual trajectory is less than 2 mm, and the measurement accuracy of the velocity distribution is better than 95%.

[0049] Figure 3 Schematic diagram for comparing the extraction accuracy of the spatio-temporal distribution characteristics of the temperature field in the embodiments of the present invention: This technical solution demonstrates significant performance advantages during the temperature field feature extraction process. Starting from the initial training stage (round 0), the feature extraction accuracy of this technical solution reaches 85%, significantly higher than the initial performance of traditional CNN methods (75%) and traditional RNN methods (70%). As the number of training rounds increases, the accuracy of each method shows an upward trend, but the advantage of this technical solution always remains. When trained to 150 rounds, the accuracy of this technical solution breaks through 95%, while the traditional methods only reach about 85%. Finally, after 300 rounds of training, the accuracy of this technical solution stabilizes at a high level of 98%, 10 percentage points higher than traditional CNN methods and 14 percentage points higher than traditional RNN methods. Notably, although the performance improvement of this technical solution slows down after 200 rounds, it still maintains a stable upward trend, indicating that this method has good learning ability and stability.

[0050] Traditional CNN methods and RNN methods in the prior art have obvious limitations during the temperature field feature extraction process. Traditional CNN methods only focus on the local spatial features of the temperature field and cannot effectively capture the global distribution characteristics of the temperature field; although traditional RNN methods can process time series data, they have deficiencies in the extraction of spatial features and are difficult to accurately describe the spatial evolution law of the temperature field. These methods often require a long training time to converge during application, and the final feature extraction accuracy is still difficult to meet the actual requirements.

[0051] From the perspective of improving the temperature field feature extraction ability, this technical solution innovatively integrates the spatial attention mechanism and the temporal attention mechanism into the deep learning algorithm. The spatial attention mechanism highlights the key spatial features in the temperature field through adaptive weight allocation, effectively enhancing the perception ability of the spatial distribution characteristics of the temperature field; the temporal attention mechanism captures the key temporal information during the evolution of the temperature field, improving the recognition accuracy of the dynamic change characteristics of the temperature field. The design of this dual attention mechanism enables the model to take into account both the spatial and temporal features of the temperature field, thus achieving coordinated optimization in the spatial and time dimensions during the feature extraction process.

[0052] Verified by experiments, this technical solution has significant technical effects compared with the prior art: First, it shows a high accuracy in the initial stage of feature extraction, indicating that this method has good feature perception ability; Second, as the training process progresses, the feature extraction accuracy always maintains a leading advantage and shows a stable upward trend, reflecting the excellent learning ability of this method; Finally, it can still maintain stable performance improvement in the later stage of training, proving that this method has strong generalization ability. The realization of these technical effects lays a solid foundation for the subsequent reconstruction of the thermal fluid motion trajectory and the calculation of the fluid velocity distribution.

[0053] In an alternative embodiment, a temperature field evolution model is used to dynamically track the thermal fluid, and a motion feature mapping relationship of the thermal fluid is established based on the outputs of a spatial attention mechanism and a temporal attention mechanism to achieve accurate reconstruction of the motion trajectory of the thermal fluid; calculating the fluid velocity distribution of the pipe cross-section according to the motion trajectory of the thermal fluid includes: Dynamically track the thermal fluid using the temperature field evolution model, input the temperature field distribution data into the temperature field evolution model, and establish a motion feature mapping relationship of the thermal fluid based on the outputs of the spatial attention mechanism and the temporal attention mechanism; Reconstruct the motion trajectory of the thermal fluid according to the motion feature mapping relationship of the thermal fluid, iteratively calculate the position information of the thermal fluid through the state estimation equation, achieve accurate reconstruction of the motion trajectory of the thermal fluid, and calculate the fluid velocity distribution of the pipe cross-section based on the reconstructed motion trajectory of the thermal fluid.

[0054] When dynamically tracking the thermal fluid using the temperature field evolution model, first collect the temperature data of multiple temperature sensor arrays inside the pipe to form temporal temperature field distribution data. The temperature sensor arrays are evenly arranged along the axial direction of the pipe, each sensor array contains 8 temperature sensors, and the sensor spacing is 10 mm. The sampling frequency is set to 100 Hz, and the temperature data is continuously collected for 300 s.

[0055] Input the collected temperature field distribution data into the temperature field evolution model, which includes a spatial attention mechanism and a temporal attention mechanism. The spatial attention mechanism extracts the spatial motion features of the thermal fluid by calculating the correlation weights of the temperature data at different spatial positions; the temporal attention mechanism extracts the temporal evolution features of the thermal fluid by calculating the correlation weights of the temperature data at different times. Specifically, the dimension of the spatial attention weight matrix is 8×8, and the dimension of the temporal attention weight matrix is 300×300.

[0056] Based on the outputs of the spatial attention mechanism and the temporal attention mechanism, establish a motion feature mapping relationship of the thermal fluid. This mapping relationship includes motion feature information such as the position and velocity of the thermal fluid at different times. The mapping relationship is represented in the form of a feature vector, and the vector dimension is 16, where the first 8 dimensions represent spatial features and the last 8 dimensions represent temporal features.

[0057] According to the established motion feature mapping relationship of the thermal fluid, iteratively calculate the position information of the thermal fluid through the state estimation equation. The state estimation equation includes two steps: state prediction and measurement update. The prediction step predicts the current state based on the previous state, and the update step corrects the predicted state by combining the current temperature measurement value. Iteratively calculate 300 times to obtain the position information of the thermal fluid at 300 moments, and achieve accurate reconstruction of the motion trajectory.

[0058] Finally, based on the reconstructed thermal fluid motion trajectories, calculate the fluid velocity distribution at the pipe cross-section. First, divide the pipe cross-section into a 100×100 grid with a grid size of 1mm×1mm. Then, calculate the fluid velocity at each point according to the time difference of the thermal fluid passing through each grid point. Finally, obtain the velocity distribution data of 100×100 grid points at the pipe cross-section.

[0059] Figure 4 Schematic diagram for comparing the position errors of different methods in the embodiments of the present invention: This figure shows the trend of position errors of different methods changing with time. The horizontal axis represents the time range from 0 to 300 s, and the vertical axis represents the position error from 0 to 6 mm. The error curve of the technical method of the present invention is represented by triangular markers and solid lines, remaining below 0.5 mm throughout, with an average error of only 0.32 mm and a maximum error of approximately 0.62 mm (t = 300 s); the traditional Kalman filtering method is represented by dot-filled square markers and dashed lines (5,5), with the error gradually increasing from 1.85 mm (t = 100 s) to 2.85 mm (t = 500 s); the CNN method is represented by rotated cross-filled square markers and long dashed lines (10,3), with an error range of 1.53 - 1.89 mm; the RNN method is represented by horizontal line markers and short dashed lines (2,2), with an error range of 1.29 - 2.15 mm. The horizontal reference lines in the figure clearly mark the error levels of 0.5 mm, 1.0 mm, 2.0 mm, 3.0 mm, 4.0 mm, and 5.0 mm. Particularly note that at three time points of t = 100 s, 300 s, and 500 s, the specific error values of each method are marked in the figure, intuitively comparing the performance differences of different methods. The horizontal reference lines and grid lines provide accurate error quantification references. The legend in the upper right corner of the chart details the representation methods and average error values of each method, highlighting the excellent performance of the technical method of the present invention in maintaining stable low errors during long-term tracking, especially when the flow conditions change complexly.

[0060] This method realizes the accurate tracking of thermal fluids through the temperature field evolution model. The spatial attention mechanism and the temporal attention mechanism effectively extract the motion characteristics of thermal fluids, improving the tracking accuracy. Based on the iterative calculation method of the state estimation equation, the accurate reconstruction of the thermal fluid motion trajectories is achieved, and the average error between the reconstructed trajectories and the actual trajectories is less than 1 mm. This method can accurately obtain the fluid velocity distribution at the pipe cross-section, providing an important basis for the analysis of pipe flow characteristics and fault diagnosis, and has strong practical value.

[0061] In an alternative embodiment, according to the fluid velocity distribution at the pipe cross-section, establish a flow rate calculation model considering hydrodynamic effects. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow rate value of the fluid in the pipe through adaptive iterative calculation, including: Establish a flow rate calculation model considering hydrodynamic effects. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and inputs the fluid velocity distribution of the pipe cross-section into the flow rate calculation model; Determine the flow pattern characteristics based on the Reynolds number of fluid flow, the pipe diameter ratio, and the viscosity ratio, and calculate the flow pattern correction coefficient according to the flow pattern characteristics. The flow pattern correction coefficient is used to characterize the transition characteristics of the fluid flow state; Determine the temperature effect characteristics based on the fluid temperature change characteristics and the fluid physical property parameters. The temperature effect characteristics include a temperature ratio correction term and a density ratio correction term, and calculate the temperature effect correction coefficient according to the temperature effect characteristics; Determine the velocity field characteristics based on the fluid velocity distribution of the pipe cross-section. The velocity field characteristics include a radial velocity distribution function and a maximum center velocity, and calculate the velocity distribution correction coefficient according to the velocity field characteristics; Construct a flow rate calculation optimization model that includes a flow rate fitting error term, a physical constraint term, and a smoothing constraint term. Establish a comprehensive objective function based on the flow pattern correction coefficient, the temperature effect correction coefficient, and the velocity distribution correction coefficient with an adaptive weight coefficient, and perform iterative optimization calculations using a parameter optimization strategy with a dynamic learning rate. Verify the accuracy of the optimization results based on a comprehensive evaluation function that includes multi-dimensional error metrics; Perform adaptive iterative calculations on the optimization results of the flow rate calculation optimization model. When the relative flow rate error between adjacent iterative steps is less than a preset error threshold, the iteration is completed to obtain the real-time flow rate value of the fluid in the pipe.

[0062] Establish a flow rate calculation model considering hydrodynamic effects. For a pipe with an inner diameter of 200 mm, a multi-point velocity measurement method is used to obtain the fluid velocity distribution data of the pipe cross-section. Measuring points are set at positions 0.5 cm, 1 cm, 2 cm, 5 cm from the pipe wall and at the pipe center to obtain the radial velocity distribution. The velocity data sampling frequency is 100 Hz, and the sampling time is 60 seconds.

[0063] Determine the flow pattern correction coefficient based on the fluid flow characteristics. The flow state is judged by the Reynolds number of fluid flow. When the Reynolds number is less than 2300, it is laminar flow; when it is greater than 4000, it is turbulent flow; and the intermediate region is a transitional flow state. The pipe diameter ratio ranges from 0.005 to 0.05, and the viscosity ratio ranges from 0.8 to 1.2. The flow pattern correction coefficient takes values from 0.2 to 0.4 in the laminar flow state, from 0.4 to 0.6 in the transitional state, and from 0.6 to 0.8 in the turbulent flow state.

[0064] Calculate the temperature effect correction coefficient. Measure the inlet and outlet temperatures of the fluid, with the temperature measurement range being 20 - 100 °C and the measurement accuracy being ±0.1 °C. Calculate the temperature ratio correction term based on the temperature change. When the temperature change is within 5 °C, the correction coefficient is in the range of 0.98 - 1.02. The density ratio correction term considers the influence of temperature on the fluid density. When the temperature rises by 20 °C, the density decreases by approximately 2%, and the corresponding correction coefficient takes values in the range of 0.97 - 0.99.

[0065] Determine the velocity distribution correction coefficient. Use a piecewise continuous function to describe the radial velocity distribution. The logarithmic distribution is used in the wall region, and the power-law distribution is used in the core region. The ratio of the maximum velocity at the center to the average velocity is between 1.2 - 1.4. The velocity distribution correction coefficient varies with the flow pattern, taking values of 0.8 - 0.9 in the laminar state and 0.9 - 0.95 in the turbulent state.

[0066] Construct an optimized model for flow rate calculation. Set the weight of the flow rate fitting error term to 0.5, the weight of the physical constraint term to 0.3, and the weight of the smoothing constraint term to 0.2. Adopt an adaptive weight adjustment strategy. When the fitting error is greater than 5%, increase the weight of the fitting error term to 0.6. Set the initial value of the dynamic learning rate to 0.01, and gradually reduce it to 0.001 as the number of iterations increases. The comprehensive evaluation function includes three dimensions: relative error, absolute error, and stability index.

[0067] Perform adaptive iterative calculation. Set the iteration termination condition as the relative error of the flow rate between adjacent iteration steps being less than 0.5%. For a DN200 pipeline, the flow rate measurement range is 10 - 100 m³ / h. Usually, it converges after 5 - 10 iterations, and the calculation time is less than 1 second. The final flow rate measurement accuracy can reach ±1.5%.

[0068] By introducing the flow pattern correction factor, temperature effect correction factor, and velocity distribution correction factor, a flow rate calculation model that comprehensively considers fluid mechanics effects is established. The multi-factor correction mechanism significantly improves the flow rate calculation accuracy, controls the measurement error within ±1.5%, and is applicable to flow rate measurements under different flow pattern conditions. Adopting the adaptive weight coefficient and dynamic learning rate strategy improves the convergence and calculation efficiency of the flow rate calculation model. The optimized model has good adaptability, can quickly respond to changes in the flow conditions, has an iteration calculation time of less than 1 second, and meets the real-time measurement requirements. A complete flow rate calculation evaluation system is established, and the reliability of the calculation results is comprehensively evaluated through multi-dimensional error indicators. The system can realize real-time monitoring and self-diagnosis of the flow rate measurement process, ensuring the stability and credibility of the measurement results. The measurement results can provide a reliable basis for process parameter optimization and equipment operation management.

[0069] In an alternative embodiment, a flow calculation optimization model including a flow fitting error term, a physical constraint term, and a smoothing constraint term is constructed. A comprehensive objective function based on adaptive weight coefficients is established based on a flow pattern correction coefficient, a temperature effect correction coefficient, and a velocity distribution correction coefficient. An iterative optimization calculation is performed using a parameter optimization strategy with a dynamic learning rate, and the accuracy of the optimization result is verified based on a comprehensive evaluation function including multi-dimensional error metrics, including: Construct a flow calculation optimization model. The flow calculation optimization model includes a flow fitting error term, a physical constraint term, and a smoothing constraint term. The flow fitting error term is used to characterize the deviation between the calculated flow rate and the reference flow rate. The physical constraint term includes a mass conservation constraint and a momentum balance constraint. The smoothing constraint term includes a time series change constraint and a velocity gradient constraint; Establish a comprehensive objective function. Combine the flow fitting error term, the physical constraint term, and the smoothing constraint term through adaptive weight coefficients. The adaptive weight coefficients are dynamically adjusted based on the error contribution of each term. The comprehensive objective function outputs an optimization target value; Construct a parameter optimization strategy. Calculate the parameter gradient based on the comprehensive objective function, and update the optimization parameters using a dynamic learning rate. The dynamic learning rate decays exponentially with the number of iteration steps. The parameter gradient is used to guide the parameter update direction; Perform iterative optimization calculation. Input the optimization parameters into the flow calculation optimization model, and judge the iteration termination condition based on the output of the comprehensive objective function. When the parameter change amount between adjacent iteration steps is less than the preset change threshold, the iteration is completed; construct a comprehensive evaluation function including multi-dimensional error metrics, calculate the optimization convergence speed according to the output of the comprehensive evaluation function, and verify the accuracy of the optimization result.

[0070] Construct a flow calculation optimization model. For a pipeline with an inner diameter of 200 mm, obtain the measured value of the reference flowmeter as the standard value, and the measurement range is 10 - 100 m³ / h. The flow fitting error term is determined by the relative deviation between the calculated flow rate and the reference flow rate, and the allowable deviation range is ±2%. The mass conservation constraint requires that the mass flow rate at the inlet section is equal to the mass flow rate at the outlet section, and the deviation is controlled within <0.5%. The momentum balance constraint considers the balance relationship between the pressure loss and the frictional resistance, and the measurement accuracy of the pressure drop is ±100 Pa. The time series change constraint limits the flow rate change rate between adjacent moments to be less than 5% / s, and the velocity gradient constraint ensures that the radial velocity distribution is continuous and smooth.

[0071] Establish a comprehensive objective function. In the initial stage, the weight of the flow fitting error term is set to 0.5, the weight of the physical constraint term is 0.3, and the weight of the smoothing constraint term is 0.2. When the flow fitting error is greater than 3%, increase the weight of the fitting error term to 0.6 and decrease the weights of other terms. When the physical constraint violation degree is greater than 1%, increase the weight of the physical constraint term to 0.4. The weight coefficients are updated once per second to ensure the balanced optimization of each error term.

[0072] Construct an optimization strategy for parameters. Adopt an optimization method based on gradient descent, with the initial learning rate set to 0.01. The learning rate is reduced to 0.8 times the original value every 100 iterations, and the minimum learning rate is limited to 0.001. The parameter gradient is calculated based on the sensitivity of the objective function to the optimized parameters. The larger the absolute value of the gradient, the larger the parameter adjustment step. For the flow pattern correction coefficient, the adjustment step each time does not exceed 0.05; for the temperature correction coefficient, the adjustment step does not exceed 0.02.

[0073] Perform iterative optimization calculations. Input the initial values of the correction coefficients into the optimization model and calculate the objective function value. When the parameter change amount between two adjacent iterations is less than 0.1%, it is determined that the iteration converges. For the rated flow condition of the DN200 pipeline, usually 10 - 15 iterations are required to converge, and the calculation time is less than 2 seconds. Establish a comprehensive evaluation function that includes the relative error, stability index, and convergence speed. The weight of the relative error is 0.5, the weight of the stability index is 0.3, and the weight of the convergence speed is 0.2. The evaluation function outputs a score between 0 and 1, and a score greater than 0.9 indicates that the optimization result is reliable.

[0074] Figure 5 This is the schematic diagram of the convergence analysis of the correction coefficient in the embodiment of the present invention: The convergence analysis diagram of the correction coefficient shows the change trends of three key correction coefficients during the iterative optimization process. The flow pattern correction coefficient marked with a square gradually converges from the initial value of 0.95 to 0.923 and stabilizes after the 5th iteration; the temperature effect correction coefficient marked with a circle increases from the initial value of 1.03 to 1.047 and reaches a stable state after the 7th iteration; the velocity distribution correction coefficient marked with a triangle decreases from the initial value of 0.87 to 0.856 and also converges after the 7th iteration. The three coefficient change curves all show good convergence characteristics, with faster changes in the early stage, gradually flattening in the later stage, and finally stabilizing on their respective convergence threshold lines. It is worth noting that the flow pattern correction coefficient has the fastest convergence speed, only requiring 5 iterations; while the temperature effect and velocity distribution correction coefficients require 7 iterations to reach a stable state. This indicates that the flow pattern characteristics are more easily captured and optimized in the model, while the temperature and velocity distribution characteristics require more iterations to be accurately described. The convergence processes of the three coefficients are stable without obvious oscillations, indicating that the optimization algorithm has good stability.

[0075] By constructing a multi-constraint optimization model, comprehensive constraints on the flow calculation process are achieved. Physical constraints such as mass conservation and momentum balance are introduced to ensure the physical rationality of the calculation results. The temporal smoothing constraint is adopted to improve the stability of flow calculation. The final flow calculation accuracy can reach ±1.5%. The comprehensive objective function based on the adaptive weight coefficient has good optimization performance. The weight coefficient can be dynamically adjusted according to the error characteristics to effectively balance the effects of each constraint term. The optimization process has a fast convergence speed and usually can be completed within 10 - 15 iterations, with high calculation efficiency. A complete optimization evaluation system is established to comprehensively evaluate the optimization results through multi-dimensional indicators. The evaluation system can timely detect abnormal optimization processes to ensure the reliability of the calculation results. The system is applicable to different working conditions and has strong adaptability and robustness. The optimization results can be directly used for flow measurement and process control.

[0076] In the second aspect of the embodiments of the present invention, a large-diameter pipeline flow dynamic detection system based on temperature gradient tracking is provided, including: The first unit is used to set up a temperature field monitoring system on the outer wall of the large-diameter pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays arranged along the axial direction of the pipeline. Each layer of temperature sensor array contains multiple temperature sensors evenly distributed circumferentially. The temperature sensors are electrically connected to the central processing unit through a data transmission module; The second unit is used to obtain the temperature field distribution data under the initial operating state of the pipeline through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; according to the initial flow parameters, determines the injection scheme of the hot fluid, and realizes the adaptive update of the flow parameters through the injection scheme. The injection scheme includes the injection temperature value, the injection flow value, and the injection duration, and injects the hot fluid into the large-diameter pipeline according to the injection scheme; The third unit is a time-domain sampling control unit used to automatically adjust the sampling time interval according to the temperature change rate, and a spatial sampling control unit used to automatically adjust the sampling point distribution according to the temperature spatial gradient. The central processing unit integrates the collected temperature data into a spatio-temporally correlated temperature feature dataset; The fourth unit is used to construct a temperature field evolution model based on the temperature feature dataset, extract the spatio-temporal features of the temperature field through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to achieve the precise reconstruction of the movement trajectory of the hot fluid and obtain the fluid velocity distribution of the pipeline cross-section; The fifth unit is used to establish a flow rate calculation model considering hydrodynamic effects based on the fluid velocity distribution in the pipeline cross-section. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow rate value of the fluid in the pipeline through adaptive iterative calculation; the real-time flow rate value is transmitted to the intelligent analysis terminal, and the intelligent analysis terminal conducts dynamic evaluation and early warning analysis on the real-time flow rate value to achieve reliable monitoring of the flow rate of large-diameter pipelines.

[0077] In the third aspect of the embodiments of the present invention, A kind of electronic device is provided, including: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the foregoing method.

[0078] In the fourth aspect of the embodiments of the present invention, A computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the foregoing method is implemented.

[0079] The present invention can be a method, a device, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are uploaded.

[0080] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic detection method for the flow rate of large-diameter pipelines based on temperature gradient tracking, characterized in that, Including: A temperature field monitoring system is provided on the outer wall of a large-diameter pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays arranged along the axial direction of the pipeline. Each layer of temperature sensor array contains multiple temperature sensors evenly distributed circumferentially. The temperature sensors are electrically connected to a central processing unit through a data transmission module; Obtain the temperature field distribution data of the pipeline under the initial operating state through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; According to the initial flow parameters, determine the injection scheme of the hot fluid, and achieve the adaptive update of the flow parameters through the injection scheme. The injection scheme includes the injection temperature value, injection flow rate value, and injection duration, and inject the hot fluid into the large-diameter pipeline according to the injection scheme; A time-domain sampling control unit that automatically adjusts the sampling time interval according to the temperature change rate, and a spatial sampling control unit that automatically adjusts the sampling point distribution according to the temperature spatial gradient. The central processing unit integrates the collected temperature data into a spatio-temporally correlated temperature feature dataset; Construct a temperature field evolution model based on the temperature feature dataset, extract the spatio-temporal features of the temperature field through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to achieve the precise reconstruction of the movement trajectory of the hot fluid and obtain the fluid velocity distribution of the pipeline cross-section; According to the fluid velocity distribution of the pipeline cross-section, establish a flow rate calculation model considering hydrodynamic effects. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow rate value of the fluid in the pipeline through adaptive iterative calculation; Transmit the real-time flow rate value to an intelligent analysis terminal. The intelligent analysis terminal conducts dynamic evaluation and early warning analysis on the real-time flow rate value to achieve reliable monitoring of the flow rate of the large-diameter pipeline.

2. The method according to claim 1, characterized in that, Obtain the temperature field distribution data of the pipeline under the initial operating state through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; According to the initial flow parameters, determine the injection scheme of the hot fluid, and the adaptive update of the flow parameters through the injection scheme includes: Obtain the temperature field distribution data of the pipeline under the initial operating state through the temperature field monitoring system provided on the outer wall of the pipeline. The temperature field monitoring system arranges multiple groups of temperature sensor arrays along the axial direction of the pipeline. Each group of temperature sensor arrays evenly distributes multiple temperature sensors along the circumferential direction of the pipeline. The temperature sensors are electrically connected to the central processing unit; The central processing unit conducts feature analysis on the temperature field distribution data, extracts the radial temperature gradient feature and the axial temperature gradient feature, and establishes a temperature field feature vector including the temperature spatial distribution feature and the temperature time fluctuation feature; The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field feature vector, determines the fluid flow pattern characteristic coefficient through temperature fluctuation spectrum analysis, calculates the average fluid velocity using temperature cross-correlation analysis, calculates the fluid turbulence degree based on the temperature field distortion degree, and constructs an initial flow parameter set of the fluid in the pipeline; Determine the injection scheme of the hot fluid according to the initial flow parameter set. The injection scheme includes the injection temperature, injection flow rate, and injection timing of the hot fluid; During the hot fluid injection process, the temperature field monitoring system continuously collects temperature signals. The central processing unit extracts the temperature field characteristics from the temperature signals and constructs a spatio-temporal feature vector, calculates the fluid velocity distribution based on multi-point cross-correlation analysis, constructs a flow pattern characteristic index in combination with the temperature power spectral density, performs flow pattern recognition using a deep learning network, establishes a set of flow parameters and implements dynamic parameter correction to achieve the adaptive update of flow parameters.

3. The method according to claim 2, wherein The continuous acquisition of temperature signals by the temperature field monitoring system, the extraction of temperature field characteristics from the temperature signals and the construction of a spatio-temporal feature vector by the central processing unit, the calculation of the fluid velocity distribution based on multi-point cross-correlation analysis, the construction of a flow pattern characteristic index in combination with the temperature power spectral density, the performance of flow pattern recognition using a deep learning network, the establishment of a set of flow parameters and the implementation of dynamic parameter correction to achieve the adaptive update of flow parameters include: Obtain multi-point temperature field data, collect temperature signals through the temperature field monitoring system installed on the outer wall of the pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays evenly distributed along the axial direction of the pipeline. Each layer of temperature sensor array is evenly provided with multiple temperature sensors along the circumferential direction of the pipeline. The temperature sensors are electrically connected to the data acquisition unit; Extract the temperature field characteristics, perform normalization preprocessing on the temperature signals, calculate the radial temperature gradient and the axial temperature gradient, construct a spatio-temporal feature vector including the temperature fluctuation intensity and the temperature autocorrelation function. The temperature autocorrelation function adopts a weighted calculation method, and the weight coefficient is related to the amplitude of the temperature gradient; Calculate the fluid velocity distribution, establish a multi-point cross-correlation analysis model based on the spatio-temporal feature vector. The multi-point cross-correlation analysis model determines the average fluid velocity by optimizing the time delay estimation, and calculates the radial velocity distribution using a piecewise continuous velocity distribution function. Different mathematical models are adopted for the velocity distribution function in the wall region and the core region respectively; Construct a flow pattern characteristic index, calculate the temperature power spectral density using the temperature field characteristics, extract the characteristic frequency parameters, and establish a flow pattern discrimination criterion in combination with the temperature fluctuation intensity and the root mean square value of the temperature gradient. The flow pattern discrimination criterion comprehensively combines multiple characteristic parameters in a weighted fusion manner; Perform intelligent flow pattern recognition, input the flow pattern characteristic index into the deep learning network. The deep learning network includes a feature extraction layer and a classification output layer. The feature extraction layer adopts a convolutional structure, and the classification output layer gives the flow pattern probability distribution through the softmax function; Establish a set of flow parameters, calculate the turbulence intensity according to the flow pattern recognition result and the velocity distribution, evaluate the flow stability in combination with the temperature field distortion degree, and construct a flow characteristic set including the flow pattern coefficient, the velocity distribution and the turbulence parameters; 4. The method according to claim 1, wherein Implement dynamic parameter correction, calculate the parameter uncertainty based on the flow characteristic set, establish a parameter correction model, and optimize the calculation accuracy of the flow parameters through a feedback adjustment mechanism to achieve the adaptive update of the flow parameters. Construct a temperature field evolution model based on the temperature feature data set, extract the spatio-temporal characteristics of the temperature field through a deep learning algorithm. The deep learning algorithm integrates a spatial attention mechanism and a temporal attention mechanism to achieve the accurate reconstruction of the hot fluid motion trajectory, and obtain the fluid velocity distribution of the pipeline cross-section including: Construct a temperature field evolution model based on the temperature field feature dataset. The temperature field evolution model uses deep learning algorithms to extract the spatio-temporal features of the temperature field. The deep learning algorithms include a convolutional neural network module for extracting the spatial distribution features of the temperature field and a recurrent neural network module for capturing the temporal variation features of the temperature field; Integrate a spatial attention mechanism and a temporal attention mechanism into the deep learning algorithm. The spatial attention mechanism assigns weights according to the importance of the spatial distribution features of the temperature field, and the temporal attention mechanism determines weights based on the correlation of the temporal evolution features of the temperature field; Use the temperature field evolution model to dynamically track the thermal fluid, establish a mapping relationship of the motion characteristics of the thermal fluid based on the outputs of the spatial attention mechanism and the temporal attention mechanism, and achieve the accurate reconstruction of the motion trajectory of the thermal fluid; calculate the fluid velocity distribution of the pipeline cross-section according to the motion trajectory of the thermal fluid.

5. The method according to claim 4, characterized in that, Use the temperature field evolution model to dynamically track the thermal fluid, establish a mapping relationship of the motion characteristics of the thermal fluid based on the outputs of the spatial attention mechanism and the temporal attention mechanism, and achieve the accurate reconstruction of the motion trajectory of the thermal fluid; calculating the fluid velocity distribution of the pipeline cross-section according to the motion trajectory of the thermal fluid includes: Use the temperature field evolution model to dynamically track the thermal fluid, input the temperature field distribution data into the temperature field evolution model, and establish a mapping relationship of the motion characteristics of the thermal fluid based on the outputs of the spatial attention mechanism and the temporal attention mechanism; Reconstruct the motion trajectory of the thermal fluid according to the mapping relationship of the motion characteristics of the thermal fluid, iteratively calculate the position information of the thermal fluid through the state estimation equation, achieve the accurate reconstruction of the motion trajectory of the thermal fluid, and calculate the fluid velocity distribution of the pipeline cross-section based on the reconstructed motion trajectory of the thermal fluid.

6. The method according to claim 1, characterized in that, According to the fluid velocity distribution of the pipeline cross-section, establish a flow rate calculation model considering the hydrodynamic effect. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow rate value of the fluid in the pipeline through adaptive iterative calculation, including: Establish a flow rate calculation model considering the hydrodynamic effect. The flow rate calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and input the fluid velocity distribution of the pipeline cross-section into the flow rate calculation model; Determine the flow pattern characteristics based on the Reynolds number of fluid flow, the diameter ratio, and the viscosity ratio, calculate the flow pattern correction coefficient according to the flow pattern characteristics, and the flow pattern correction coefficient is used to characterize the characteristics of the fluid flow state transition; Determine the temperature effect characteristics based on the temperature change characteristics of the fluid and the physical property parameters of the fluid. The temperature effect characteristics include a temperature ratio correction term and a density ratio correction term, and calculate the temperature effect correction coefficient according to the temperature effect characteristics; Determine the velocity field characteristics based on the fluid velocity distribution of the pipeline cross-section. The velocity field characteristics include a radial velocity distribution function and a maximum center velocity, and calculate the velocity distribution correction coefficient according to the velocity field characteristics; Construct a flow rate calculation optimization model including a flow rate fitting error term, a physical constraint term, and a smoothing constraint term, establish a comprehensive objective function based on the adaptive weight coefficient based on the flow pattern correction coefficient, the temperature effect correction coefficient, and the velocity distribution correction coefficient, perform iterative optimization calculations using a parameter optimization strategy with a dynamic learning rate, and verify the accuracy of the optimization results based on a comprehensive evaluation function including multi-dimensional error indicators; Perform adaptive iterative calculation on the optimization results of the flow rate calculation optimization model. When the relative error of the flow rate between adjacent iterative steps is less than the preset error threshold, the iteration is completed to obtain the real-time flow rate value of the fluid in the pipeline.

7. The method according to claim 6, characterized in that Construct a flow rate calculation optimization model that includes a flow rate fitting error term, a physical constraint term, and a smoothing constraint term. Establish a comprehensive objective function based on the adaptive weight coefficient based on the flow pattern correction coefficient, the temperature effect correction coefficient, and the velocity distribution correction coefficient. Use the parameter optimization strategy with a dynamic learning rate to perform iterative optimization calculation, and verify the accuracy of the optimization results based on a comprehensive evaluation function that includes multi-dimensional error metrics, including: Construct a flow rate calculation optimization model. The flow rate calculation optimization model includes a flow rate fitting error term, a physical constraint term, and a smoothing constraint term. The flow rate fitting error term is used to characterize the deviation between the calculated flow rate and the reference flow rate. The physical constraint term includes a mass conservation constraint and a momentum balance constraint. The smoothing constraint term includes a temporal variation constraint and a velocity gradient constraint; Establish a comprehensive objective function. Combine the flow rate fitting error term, the physical constraint term, and the smoothing constraint term through an adaptive weight coefficient. The adaptive weight coefficient is dynamically adjusted based on the error contribution degree of each item. The comprehensive objective function outputs the optimization objective value; Construct a parameter optimization strategy. Calculate the parameter gradient based on the comprehensive objective function, and update the optimization parameters using a dynamic learning rate. The dynamic learning rate decays exponentially with the number of iterative steps. The parameter gradient is used to guide the parameter update direction; Perform iterative optimization calculation. Input the optimization parameters into the flow rate calculation optimization model, and judge the iteration termination condition based on the output of the comprehensive objective function. When the change amount of the parameters between adjacent iterative steps is less than the preset change threshold, the iteration is completed; construct a comprehensive evaluation function that includes multi-dimensional error metrics, and calculate the optimization convergence speed according to the output of the comprehensive evaluation function to verify the accuracy of the optimization results.

8. A large-diameter pipeline flow dynamic detection system based on temperature gradient tracking, which is used to implement the method of any one of the foregoing claims 1-7, characterized in that Including: The first unit is used to set up a temperature field monitoring system on the outer wall of the large-diameter pipeline. The temperature field monitoring system includes multiple layers of temperature sensor arrays arranged along the axial direction of the pipeline. Each layer of temperature sensor array contains multiple temperature sensors evenly distributed circumferentially. The temperature sensors are electrically connected to the central processing unit through a data transmission module; The second unit is used to obtain the temperature field distribution data under the initial operating state of the pipeline through the temperature field monitoring system. The central processing unit calculates the initial flow parameters of the fluid in the pipeline based on the temperature field distribution data; According to the initial flow parameters, determine the injection scheme of the hot fluid, and realize the adaptive update of the flow parameters through the injection scheme. The injection scheme includes the injection temperature value, the injection flow rate value, and the injection duration. Inject the hot fluid into the large-diameter pipeline according to the injection scheme; The third unit is a time-domain sampling control unit that automatically adjusts the sampling time interval according to the temperature change rate, and a spatial sampling control unit that automatically adjusts the sampling point distribution according to the temperature spatial gradient. The central processing unit integrates the collected temperature data into a temperature feature dataset with spatio-temporal correlation; The fourth unit is used to construct a temperature field evolution model based on a temperature feature dataset, extract the spatio-temporal features of the temperature field through a deep learning algorithm, which integrates a spatial attention mechanism and a temporal attention mechanism, achieve the accurate reconstruction of the thermal fluid motion trajectory, and obtain the fluid velocity distribution of the pipe cross-section; The fifth unit is used to establish a flow calculation model considering fluid mechanics effects according to the fluid velocity distribution of the pipe cross-section. The flow calculation model introduces a flow pattern correction factor, a temperature effect correction factor, and a velocity distribution correction factor, and obtains the real-time flow value of the fluid in the pipe through adaptive iterative calculation; Transmit the real-time flow value to the intelligent analysis terminal, and the intelligent analysis terminal conducts dynamic evaluation and early warning analysis on the real-time flow value to achieve reliable monitoring of the flow rate of large-diameter pipelines.

9. An electronic device, characterized in that, Comprising: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

Images