Large-pipe-diameter air volume accurate measurement system based on AI and digital twinning

Abstract

The invention discloses a large-pipe-diameter air volume accurate measurement system based on AI and digital twinning, which relates to the technical field of air volume measurement and comprises an acquisition module, a processing transmission module, a twinning model module, an analysis module, a man-machine interaction module, a historical abnormal data storage module and an early warning module. The acquisition module is located in the pipeline and used for acquiring airflow material parameter information in the pipeline in real time; the processing and transmitting module is used for preprocessing the collected data and uploading the processed data to the analysis and prediction module; the twinborn model module is based on the actual geometric structure, the material characteristics and the physical law of internal airflow flowing of the large-diameter pipeline. According to the method, through mass data input and algorithm training, the influence of various complex factors on the air volume is fully considered, the accuracy of air volume calculation is greatly improved, data collection is conducted through the sensors, the use limitation is avoided, collection is conducted through multiple distributed sensors, the accuracy of data collection and analysis can be improved, and errors are reduced.

Description

Technical Field

[0001] The present invention relates to the field of air volume measurement technology, and in particular to a large-diameter air volume precision measurement system based on AI and digital twins. Background Art

[0002] In many fields such as industrial production, ventilation systems, HVAC, etc., accurate measurement of large-diameter air volume is extremely critical to the system's operating efficiency, energy consumption control, and environmental regulation. The development of large-scale industry is accompanied by the concentrated and large-scale discharge of three wastes. The concentrated discharge has brought many conveniences to environmental protection treatment. However, the accurate detection and monitoring of large-scale emissions has become an urgent need for enterprises and environmental protection departments. Especially for waste gas, accurate measurement of emissions and emission factors can not only grasp the emission of pollutants and carbon dioxide of enterprises, but also have important significance in the accurate peak regulation of coal-fired power. my country's total pollutant emission reduction work mainly focuses on studying the composition and concentration of waste gas emissions. The analytical test of measuring gas flow in large-diameter pipelines is inaccurate and has large errors. For large-diameter pipelines, industrial waste gas includes a large amount of moisture, dust and corrosive pollutants, which are very harmful to the flow meter and cause errors in subsequent measurements. As the nominal diameter of the pipeline increases, the calibration of the flow signal becomes more difficult. In actual projects, the use of large pipelines often has limitations. Large pipelines have various shapes, including round, square and rectangular. At the same time, there is a serious shortage of straight pipe sections, and the gas flow field distribution in the pipeline is extremely uneven, resulting in generally low accuracy and a general error of more than 5% to 7%. In this regard, we have proposed a large-diameter air volume precision measurement system based on AI and digital twins. Summary of the invention

[0003] In order to solve the above technical problems, a large-diameter air volume precision measurement system based on AI and digital twins is provided. This technical solution solves the problems of measurement errors, limitations in the use of large pipes, and low accuracy.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is: a large-diameter air volume precision measurement system based on AI and digital twins, including: an acquisition module, a processing and transmission module, a twin model module, an analysis module, a human-computer interaction module, a historical abnormal data storage module and an early warning module;

[0005] The acquisition module is located inside the pipeline and is used to collect the air flow material parameter information in the pipeline in real time;

[0006] The processing and transmission module is used to pre-process the collected data and upload the processed data to the analysis and prediction module;

[0007] The twin model module builds a digital twin model of large-diameter pipelines based on the actual geometric structure, material properties and physical laws of internal airflow of large-diameter pipelines using computational fluid dynamics simulation technology, and trains and optimizes the model.

[0008] The analysis module receives the uploaded data, inputs the data into the twin model module, analyzes and outputs the data based on the artificial intelligence algorithm, calculates the large-diameter air volume, and inputs the predicted calculated value into the historical abnormal data storage module;

[0009] The historical abnormal data storage module obtains historical data, obtains the normal airflow threshold range based on machine learning algorithm training, compares the input calculated airflow value with the threshold, and determines whether the current large-diameter air volume is abnormal;

[0010] After the early warning module determines that the large-diameter air volume is abnormal, it sends an early warning message to the human-computer interaction module to warn the operator that the large-diameter air volume is abnormal and take action;

[0011] The human-computer interaction module provides operators with a visual interactive interface and is equipped with an encryption security unit to prevent outside users from accessing the system to ensure security.

[0012] Preferably, the acquisition module includes a wind speed sensor, a pressure sensor, a temperature sensor and a humidity sensor, and the sensors are fixedly installed in a large-diameter pipeline to collect data in real time.

[0013] Preferably, the processing and transmission module preprocesses the collected data, and the preprocessing includes data cleaning, data filtering and data normalization processing. Data cleaning includes missing value processing, abnormal value processing and repeated value processing, so as to remove missing values, abnormal values ​​and repeated values; data filtering processing is performed by simple moving average filtering processing, and the data is smoothed by calculating the average value of the data in the window. The original data sequence is set to x(n), and the window size is m. Then the data after moving average filtering is:

[0014]

[0015] Where n is the serial number of the data point, m is an odd number, and i is a specific value. After calculation and processing, the filtered data is obtained;

[0016] The normalization processing method uses linear normalization processing to linearly map the original data to the specified area.

[0017] Preferably, the twin model module measures and records the diameter, length, curvature and cross-sectional geometric parameters of the pipeline, restores the spatial form of the pipeline, records the roughness, thermal conductivity and elastic modulus physical properties of the pipeline material, and after mastering the geometric structure and material properties, uses computational fluid dynamics simulation technology to set boundary conditions that meet the actual situation, including the velocity, temperature, pressure distribution of the inlet airflow and the pressure conditions of the outlet, to simulate the complex flow state of the airflow in the pipeline;

[0018] After the model is built, the airflow parameter data obtained from actual measurements is input into the model. By comparing the differences between the simulation results output by the model and the actual measured data, the optimization algorithm is used to adjust the parameters in the model to reduce the simulation error and optimize the model.

[0019] Preferably, the optimization algorithm is a genetic algorithm, which simulates the biological evolution process, encodes the model parameters as chromosomes, and continuously generates new parameter combinations through selection, crossover, and mutation operations. The difference between the model output corresponding to the combination and the measured data is evaluated, and the parameter combinations with better performance are retained. After multiple generations of evolution, the optimized model parameters are obtained, so that the model simulation results have the highest match with the measured data.

[0020] Preferably, the analysis module uses the data uploaded by the data collection end as the feature vector A, A = (a 1 , a 2 , ...a b ), where a 1 is the collected wind speed, a 2 is the collected pressure, calculated based on the following formula:

[0021] y=β 0 +β 1 a 1 +β 2 a 2 +…+β b a b +∈

[0022] where β 0 , β 1 , ... β b is the regression coefficient, ∈ is the error term, and y is the air volume value of the largest pipe diameter. During training, the least squares method is used to fit the data to find the regression coefficient that minimizes the sum of squares of the error between the predicted air volume and the actual air volume. The collected data is input into the formula for calculation, and the air volume calculation result is finally output.

[0023] Preferably, the historical abnormal data storage module obtains historical data from the database, and the historical data includes detailed data records of wind speed, pressure, temperature, humidity at different times, and the corresponding air volume data. The machine learning algorithm is used for in-depth training and analysis to determine the normal airflow threshold range. During the training process, a clustering algorithm is used to cluster the airflow parameter data in the historical data into different categories based on the similarity characteristics of the data, calculate the distance measurement between the data points, and divide the data points with similar characteristics into the same category. After repeated iterative calculations and optimizations, clustering is completed, and the cluster center and the range boundary of the cluster formed by the airflow parameter data under normal operating conditions are analyzed to determine the normal airflow threshold range.

[0024] Preferably, the distance metric between data points is calculated by Euclidean distance calculation, assuming that two data points Ci = (d i1 , d i2 , ..., d ie ), Cf=(d f1 , d f2 , ...d fe ), where e represents the feature dimension of the data, including airflow parameter data, wind speed, pressure and temperature, e is the number of features, and the Euclidean distance G (Ci, Cf) between two points is calculated as:

[0025]

[0026] After calculating the distance value, the clustering algorithm steps are to initialize the cluster center, assign data points to clusters, update the cluster center and iterate the convergence judgment. After the clustering is completed, the statistics of the data points in each dimension within the cluster are finally calculated to determine the threshold range;

[0027] When the calculated value output is greater than the threshold, it is judged that the large diameter air volume is abnormal, otherwise it is in a normal state.

[0028] Preferably, after determining that the current large-diameter air volume is abnormal, the early warning module starts the early warning process, sends early warning information, and presents it to the operator in an intuitive and eye-catching manner. A warning pop-up window with a red background and a flashing warning icon pops up on the operation interface.

[0029] Preferably, the visual interactive interface is in the core area of ​​the main interface, which clearly presents the key parameters of the airflow in the large-diameter pipe in a dynamic and real-time updated manner, and is equipped with trend charts to depict the change trajectory, allowing operators to intuitively grasp the dynamic changes of the air volume.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention fully considers the influence of various complex factors on air volume through a large amount of data input and algorithm training, thereby greatly improving the accuracy of air volume calculation. Data collection by sensors avoids the limitation of use. Collection by multiple distributed sensors can improve the accuracy of data collection and analysis and reduce the generation of errors. The collaborative work of the historical abnormal data storage module and the analysis module can quickly and accurately detect abnormal situations of large-diameter air volume and remind management personnel to deal with them in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a framework diagram of the measurement system of the present invention. DETAILED DESCRIPTION

[0033] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0034] Reference Figure 1 As shown in the figure, the large-diameter air volume precision measurement system based on AI and digital twins includes: acquisition module, processing and transmission module, twin model module, analysis module, human-computer interaction module, historical abnormal data storage module and early warning module;

[0035] The acquisition module is located inside the pipeline and is used to collect the air flow material parameter information in the pipeline in real time;

[0036] The processing and transmission module is used to pre-process the collected data and upload the processed data to the analysis and prediction module;

[0037] The twin model module builds a digital twin model of large-diameter pipelines based on the actual geometric structure, material properties and physical laws of internal airflow of large-diameter pipelines using computational fluid dynamics simulation technology, and trains and optimizes the model.

[0038] The analysis module receives the uploaded data, inputs the data into the twin model module, analyzes and outputs the data based on the artificial intelligence algorithm, calculates the large-diameter air volume, and inputs the predicted calculated value into the historical abnormal data storage module;

[0039] The historical abnormal data storage module obtains historical data, obtains the normal airflow threshold range based on machine learning algorithm training, compares the input calculated airflow value with the threshold, and determines whether the current large-diameter air volume is abnormal;

[0040] After the early warning module determines that the large-diameter air volume is abnormal, it sends an early warning message to the human-computer interaction module to warn the operator that the large-diameter air volume is abnormal and take action;

[0041] The human-computer interaction module provides operators with a visual interactive interface and is equipped with an encryption security unit to prevent outside users from accessing the system to ensure security.

[0042] The acquisition module of this application is located inside the pipeline and can directly contact the airflow, and collect airflow material parameter information in real time, including wind speed, pressure, temperature, humidity, and multi-dimensional data of impurity content; the digital twin model constructed by the twin model module is based on the actual geometric structure, material characteristics and physical laws of airflow of the pipeline, and can highly restore the complex flow state of airflow in the pipeline by using computational fluid dynamics simulation technology, whether it is the bending, diameter change, branching geometric characteristics of the pipeline, or the roughness of the material, The influence of thermal conductivity physical properties on the airflow can be accurately simulated;

[0043] The historical abnormal data storage module uses machine learning algorithm training to obtain the normal airflow threshold range. By comparing it with the real-time calculated airflow value, it can automatically and accurately determine whether the current large-diameter air volume is abnormal. After detecting the abnormality, the early warning module quickly sends early warning information to the human-computer interaction module, so that the operator can know the system abnormality in time and take corresponding measures. This helps to prevent a series of safety accidents caused by abnormal air volume, such as equipment overheating and damage due to insufficient air volume, and pipeline rupture or leakage due to excessive air volume;

[0044] The human-computer interaction module provides a visual interactive interface, allowing operators to intuitively view various data information, real-time air volume data, airflow status simulation diagrams, historical data curves, and conveniently perform system parameter settings and equipment control operations.

[0045] The acquisition module includes wind speed sensors, pressure sensors, temperature sensors and humidity sensors. The sensors are fixedly installed in large-diameter pipes to collect data in real time.

[0046] The four sensors in the present application work together, and the comprehensive data collected can build a complete portrait of the airflow state in the pipeline. By comprehensively analyzing the wind speed, pressure, temperature and humidity data, the state of the airflow can be more accurately evaluated. These data can be used to build more complex fluid mechanics models or energy balance models, and conduct more in-depth research and analysis on the physical processes in the pipeline. This multi-dimensional data fusion analysis can discover potential problems or abnormal conditions that a single sensor cannot detect. By comparing the changing trends of temperature and pressure, combined with wind speed and humidity data, it can be determined whether water vapor condensation has occurred in the pipeline. Since the sensor is fixedly installed in a large-diameter pipeline and can collect data in real time, the system can instantly grasp the dynamic changes of the airflow in the pipeline. During industrial production or ventilation system operation, the state of the airflow may change rapidly due to the start and stop of equipment, environmental changes or process adjustments.

[0047] The processing and transmission module preprocesses the collected data. The preprocessing includes data cleaning, data filtering and data normalization. Data cleaning includes missing value processing, abnormal value processing and repeated value processing, so as to remove missing values, abnormal values ​​and repeated values. Data filtering is processed by simple moving average filtering. The data is smoothed by calculating the average value of the data in the window. The original data sequence is set to x(n) and the window size is m. The data after moving average filtering is:

[0048]

[0049] Where n is the serial number of the data point, m is an odd number, and i is a specific value. After calculation and processing, the filtered data is obtained;

[0050] The normalization processing method uses linear normalization processing to linearly map the original data to the specified area.

[0051] After completing the normalization process, this application needs to verify whether the normalized data set meets expectations. The verification can be done by checking whether the normalized data range is within the target range and whether the data distribution pattern is reasonable. The original data is linearly transformed to accurately map it to the specified area. When implementing linear normalization, the original data must first be comprehensively and deeply analyzed, and the minimum and maximum values ​​corresponding to each variable in the original data must be carefully identified and determined. This step is like the cornerstone of a building and plays a decisive role in subsequent normalization operations.

[0052] The twin model module measures and records the diameter, length, curvature and cross-sectional geometric parameters of the pipeline, restores the spatial form of the pipeline, and records the roughness, thermal conductivity, and elastic modulus physical properties of the pipeline material. After mastering the geometric structure and material properties, it uses computational fluid dynamics simulation technology to set boundary conditions that meet actual conditions, including the velocity, temperature, and pressure distribution of the inlet airflow and the pressure conditions of the outlet, to simulate the complex flow state of the airflow in the pipeline.

[0053] After the model is built, the airflow parameter data obtained from actual measurements is input into the model. By comparing the differences between the simulation results output by the model and the actual measured data, the optimization algorithm is used to adjust the parameters in the model to reduce the simulation error and optimize the model.

[0054] After fully understanding these geometric structures and material properties, we began to use computational fluid dynamics simulation technology to build models. When setting boundary conditions, the velocity distribution setting of the inlet airflow is particularly critical. According to actual measurement data or engineering experience, if the inlet airflow is a relatively uniform airflow provided by the fan equipment, it can be set to a uniform velocity distribution. In the Cartesian coordinate system, the inlet velocity is a constant value in the x direction, and the velocities in the y and z directions are zero. However, if there is a certain turbulence or unevenness in the inlet airflow, such as in some industrial air inlets, due to interference from the surrounding environment or the structural characteristics of the air intake device, there are differences in the spatial distribution of the airflow velocity. In this case, it is necessary to make detailed settings based on the velocity profile data obtained from actual measurements. It may be a complex function form about the x, y, and z coordinates. The inlet temperature distribution is also set according to the actual situation. If it is an airflow introduced from a constant temperature environment, it can be set to a uniform temperature, but if it is heated Or the airflow that is cooled, or the airflow that is affected by the thermal radiation of the surrounding environment, needs to be set according to the measured temperature distribution function. For example, at the inlet of some chemical pipelines, due to the thermal effect of the upstream reaction process, the airflow temperature shows a distribution of high center and low edge in the inlet section. It is necessary to accurately input this distribution function into the model. The setting of the inlet pressure distribution must also take into account the actual source. If it is connected to a high-pressure gas source, the pressure distribution needs to be determined according to the pressure characteristics of the gas source and the pipeline connection method. It may be a uniform pressure or a distribution form with a certain pressure gradient. For the outlet pressure conditions, it is set according to the connection conditions of the pipeline outlet and the downstream environmental pressure. If the outlet leads directly to the atmospheric environment, it can usually be set to the local atmospheric pressure; if the outlet is connected to other pressure equipment or is in a specific pressure environment, such as the pipeline outlet in a vacuum system, the corresponding pressure value needs to be set according to the actual pressure environment.

[0055] The optimization algorithm is a genetic algorithm, which simulates the biological evolution process. The model parameters are encoded as chromosomes. Through selection, crossover, and mutation operations, new parameter combinations are continuously generated. The differences between the model outputs corresponding to the combinations and the measured data are evaluated, and the parameter combinations with better performance are retained. After multiple generations of evolution, the optimized model parameters are obtained, so that the model simulation results are most closely matched with the measured data.

[0056] The optimization algorithm used in this application is a genetic algorithm, which is a highly innovative and scientific algorithm strategy. Its core principle is to cleverly simulate the biological evolution process. First, it is necessary to encode and process many parameters in the model. These parameters can cover various physical property parameters related to large-diameter pipes, boundary condition parameters, and key coefficients in computational fluid dynamics simulation models and artificial intelligence models. These parameters are encoded in the form of chromosomes, just like chromosomes that carry genetic information in organisms. Each chromosome represents a specific set of model parameter combinations. In the initial stage of the algorithm operation, a certain number of chromosome populations will be randomly generated. These initial parameter combinations are like primitive organisms in the process of biological evolution. Individuals, each with different characteristics, then enter the key iterative evolution process, in which the selection operation is based on the fitness function. The fitness function is defined here as a quantitative evaluation index of the difference between the model output and the measured data. The model parameters corresponding to each chromosome are substituted into the model for calculation to obtain the simulation results of the model, and then a detailed comparative analysis is performed with the actual measured data to calculate the difference between the two. The smaller the difference, the higher the fitness of the parameter combination represented by the chromosome. Based on the fitness, chromosomes with higher fitness are selected from the current population. These selected chromosomes have a better chance of passing on the "excellent genes" they carry to the next generation.

[0057] The crossover operation simulates the gene recombination phenomenon in the biological genetic process. Among the selected chromosomes, according to a certain crossover probability, certain positions on the chromosomes are randomly selected for gene exchange.

[0058] The analysis module uses the data uploaded by the data acquisition terminal as the feature vector A, A = (a 1 , a 2 , ...a b ), where a 1 is the collected wind speed, a 2 is the collected pressure, calculated based on the following formula:

[0059] y=β 0 +β 1 a 1 +β 2 a 2 +…+β b a b +∈

[0060] where β 0 , β 1 , ... β bis the regression coefficient, ∈ is the error term, and y is the air volume value of the largest pipe diameter. During training, the least squares method is used to fit the data to find the regression coefficient that minimizes the sum of squares of the error between the predicted air volume and the actual air volume. The collected data is input into the formula for calculation, and the air volume calculation result is finally output.

[0061] This application uses the least squares method to perform data fitting and train regression coefficients, making full use of a large amount of historical data and real-time collected data. By minimizing the sum of squares of errors between predicted air volume and actual air volume, it can automatically learn from the data the intrinsic relationship between air volume and various characteristic parameters without assuming complex physical models in advance or relying on a large number of empirical formulas.

[0062] The historical abnormal data storage module obtains historical data from the database. The historical data includes detailed data records of wind speed, pressure, temperature, humidity at different times, and the corresponding air volume data. The machine learning algorithm is used for in-depth training and analysis to determine the normal airflow threshold range. During the training process, the clustering algorithm is used to cluster the airflow parameter data in the historical data into different categories based on the similarity characteristics of the data, calculate the distance measurement between data points, and divide the data points with similar characteristics into the same category. After repeated iterative calculations and optimizations, clustering is completed, and the cluster center and the range boundary of the cluster formed by the airflow parameter data under normal operating conditions are analyzed to determine the normal airflow threshold range.

[0063] This application uses a machine learning clustering algorithm to allow the characteristics of the data itself to dominate the determination of the threshold range. The clustering algorithm automatically divides the data into different categories based on the similarity of the data. There is no need to assume a complex airflow anomaly model in advance or rely on a lot of manual experience to set the threshold. For some complex industrial ventilation ducts, where the airflow may be affected by a variety of unknown factors, the traditional threshold setting method based on physical formulas and experience is difficult to accurately define the normal range. The machine learning clustering algorithm can automatically discover the inherent clustering pattern of normal airflow data from a large amount of historical data, thereby determining a threshold that is more in line with the actual situation.

[0064] The distance metric between data points is calculated by Euclidean distance. Let two data points Ci = (d i1 , d i2 , ..., d ie ), Cf=(d f1 , d f2 , ...d fe ), where e represents the feature dimension of the data, including airflow parameter data, wind speed, pressure and temperature, e is the number of features, and the Euclidean distance G (Ci, Cf) between two points is calculated as:

[0065]

[0066] After calculating the distance value, the clustering algorithm steps are to initialize the cluster center, assign data points to clusters, update the cluster center and iterate the convergence judgment. After the clustering is completed, the statistics of the data points in each dimension within the cluster are finally calculated to determine the threshold range;

[0067] When the calculated value output is greater than the threshold, it is judged that the large diameter air volume is abnormal, otherwise it is in a normal state.

[0068] After determining that the current large-diameter air volume is abnormal, the early warning module starts the early warning process, sends early warning information, and presents it to the operator in an intuitive and eye-catching manner. A warning pop-up window with a red background and a flashing warning icon will pop up on the operation interface.

[0069] This application uses red as the background color of the warning pop-up window. Red has extremely high recognition and warning properties in human visual perception. It can quickly stand out in the operator's field of vision. Even when the operator is paying attention to multiple monitoring interfaces at the same time or is in a busy working state, it can attract their attention in the first time and make them quickly focus on the abnormal warning information. In large industrial control rooms, the monitoring screens of many devices display various information at the same time. The red warning pop-up window will be like a striking signal, immediately alerting the operator to avoid ignoring the critical situation of abnormal air volume due to too much information.

[0070] The visual interactive interface is located in the core area of ​​the main interface. It clearly presents the key parameters of airflow in large-diameter pipes in a dynamic and real-time updated manner, and is paired with trend charts to depict the trajectory of changes, allowing operators to intuitively grasp the dynamic changes in air volume.

[0071] The trend chart provided with this application provides operators with an intuitive understanding of the history and future trends of air volume changes. By observing the trajectory of air volume changes over time, operators can analyze the periodic changes in air volume. For example, in some industrial production processes, due to the periodic switching of production procedures, the air volume will show regular fluctuations. Understanding these patterns will help to prepare and adjust in advance and optimize system operation.

[0072] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention, and these changes and improvements fall within the scope of the present invention claimed.

Claims

1. Large-diameter air volume precision measurement system based on AI and digital twins, characterized by: include: Acquisition module, processing and transmission module, twin model module, analysis module, human-computer interaction module, historical abnormal data storage module and early warning module; The acquisition module is located inside the pipeline and is used to collect the air flow material parameter information in the pipeline in real time; The processing and transmission module is used to pre-process the collected data and upload the processed data to the analysis and prediction module; The twin model module builds a digital twin model of large-diameter pipelines based on the actual geometric structure, material properties and physical laws of internal airflow of large-diameter pipelines using computational fluid dynamics simulation technology, and trains and optimizes the model. The analysis module receives the uploaded data, inputs the data into the twin model module, analyzes and outputs the data based on the artificial intelligence algorithm, calculates the large-diameter air volume, and inputs the predicted calculated value into the historical abnormal data storage module; The historical abnormal data storage module obtains historical data, obtains the normal airflow threshold range based on machine learning algorithm training, compares the input calculated airflow value with the threshold, and determines whether the current large-diameter air volume is abnormal; After the early warning module determines that the large-diameter air volume is abnormal, it sends an early warning message to the human-computer interaction module to warn the operator that the large-diameter air volume is abnormal and take action; The human-computer interaction module provides operators with a visual interactive interface and is equipped with an encryption security unit to prevent outside users from accessing the system to ensure security.

2. According to claim 1, the large-diameter air volume precision measurement system based on AI and digital twins is characterized in that: The acquisition module includes wind speed sensors, pressure sensors, temperature sensors and humidity sensors. The sensors are fixedly installed in large-diameter pipes to collect data in real time.

3. The large-diameter air volume precision measurement system based on AI and digital twins according to claim 1 is characterized in that: The processing and transmission module preprocesses the collected data. The preprocessing includes data cleaning, data filtering and data normalization. Data cleaning includes missing value processing, abnormal value processing and duplicate value processing, so as to remove missing values, abnormal values ​​and duplicate values. Data filtering is done by simple moving average filtering. The data is smoothed by calculating the average value of the data in the window. The original data sequence is set to x(n) and the window size is m. The data after moving average filtering is: Where n is the serial number of the data point, m is an odd number, and i is a specific value. After calculation and processing, the filtered data is obtained; The normalization processing method uses linear normalization processing to linearly map the original data to the specified area.

4. According to claim 1, the large-diameter air volume precision measurement system based on AI and digital twins is characterized in that: The twin model module measures and records the diameter, length, curvature and cross-sectional geometric parameters of the pipeline, restores the spatial form of the pipeline, and records the roughness, thermal conductivity, and elastic modulus physical properties of the pipeline material. After mastering the geometric structure and material properties, it uses computational fluid dynamics simulation technology to set boundary conditions that meet actual conditions, including the velocity, temperature, and pressure distribution of the inlet airflow and the pressure conditions of the outlet, to simulate the complex flow state of the airflow in the pipeline. After the model is built, the airflow parameter data obtained from actual measurements is input into the model. By comparing the differences between the simulation results output by the model and the actual measured data, the optimization algorithm is used to adjust the parameters in the model to reduce the simulation error and optimize the model.

5. The large-diameter air volume precision measurement system based on AI and digital twins according to claim 4 is characterized in that: The optimization algorithm is a genetic algorithm, which simulates the biological evolution process. The model parameters are encoded as chromosomes. Through selection, crossover, and mutation operations, new parameter combinations are continuously generated. The differences between the model outputs corresponding to the combinations and the measured data are evaluated, and the parameter combinations with better performance are retained. After multiple generations of evolution, the optimized model parameters are obtained, so that the model simulation results are most closely matched with the measured data.

6. The large-diameter air volume precision measurement system based on AI and digital twins according to claim 1 is characterized in that: The analysis module uses the data uploaded by the data acquisition terminal as the feature vector A, A = (a1, a2, ... a b ), where a1 is the collected wind speed and a2 is the collected pressure, calculated based on the following formula: y=β0+β1a1+β2a2+…+β b a b +∈ where β0, β1, ... β b is the regression coefficient, ∈ is the error term, and y is the air volume value of the largest pipe diameter. During training, the least squares method is used to fit the data to find the regression coefficient that minimizes the sum of squares of the error between the predicted air volume and the actual air volume. The collected data is input into the formula for calculation, and the air volume calculation result is finally output.

7. The large-diameter air volume precision measurement system based on AI and digital twins according to claim 1 is characterized in that: The historical abnormal data storage module obtains historical data from the database. The historical data includes detailed data records of wind speed, pressure, temperature, humidity at different times, and the corresponding air volume data. The machine learning algorithm is used for in-depth training and analysis to determine the normal airflow threshold range. During the training process, the clustering algorithm is used to cluster the airflow parameter data in the historical data into different categories based on the similarity characteristics of the data, calculate the distance measurement between data points, and divide the data points with similar characteristics into the same category. After repeated iterative calculations and optimizations, clustering is completed, and the cluster center and the range boundary of the cluster formed by the airflow parameter data under normal operating conditions are analyzed to determine the normal airflow threshold range.

8. The large-diameter air volume precision measurement system based on AI and digital twins according to claim 7 is characterized in that: The distance metric between data points is calculated by Euclidean distance. Let two data points Ci = (d i1 , d i2 , ..., d ie ), Cf=(d f1 , d f2 , ...d fe ), where e represents the feature dimension of the data, including airflow parameter data, wind speed, pressure and temperature, e is the number of features, and the Euclidean distance G (Ci, Cf) between two points is calculated as follows: After calculating the distance value, the clustering algorithm steps are to initialize the cluster center, assign data points to clusters, update the cluster center and iterate the convergence judgment. After the clustering is completed, the statistics of the data points in each dimension within the cluster are finally calculated to determine the threshold range; When the calculated value output is greater than the threshold, it is judged that the large diameter air volume is abnormal, otherwise it is in a normal state.

9. The large-diameter air volume precision measurement system based on AI and digital twins according to claim 1 is characterized in that: After determining that the current large-diameter air volume is abnormal, the early warning module starts the early warning process, sends early warning information, and presents it to the operator in an intuitive and eye-catching manner. A warning pop-up window with a red background and a flashing warning icon will pop up on the operation interface.

10. The large-diameter air volume precision measurement system based on AI and digital twins according to claim 1 is characterized in that: The visual interactive interface is located in the core area of ​​the main interface. It clearly presents the key parameters of airflow in large-diameter pipes in a dynamic and real-time updated manner, and is paired with trend charts to depict the trajectory of changes, allowing operators to intuitively grasp the dynamic changes in air volume.

Images