Vortex flowmeter monitoring method, device, medium and product based on deep learning

Abstract

A monitoring method, device, medium and product for a vortex flowmeter based on deep learning, which relates to the field of electronic digital data processing. The method includes: obtaining historical time-series data of the vortex flowmeter and preprocessing it to obtain standardized data; dividing the standardized data into a training set and a validation set in a preset proportion; according to the flow characteristics of the vortex flowmeter, dividing the training set into segments according to a preset flow threshold interval, and respectively constructing deep learning models to obtain a flow model set; using the validation set to verify multiple deep learning models in the flow model set and optimizing the flow model set; collecting real-time data of the vortex flowmeter and performing model calculation according to the flow threshold interval to which the real-time data belongs to obtain a flow correction value; when the performance of the flowmeter calculated based on the flow correction value is lower than a preset performance threshold, generating an abnormal warning message. Implementing this application can improve the timeliness of monitoring the vortex flowmeter and reduce its maintenance cost.

Description

Technical Field

[0001] This application relates to the field of electrical digital data processing, and particularly to a monitoring method, device, medium and product for vortex flowmeters based on deep learning. Background Art

[0002] During industrial production processes, the monitoring of the operating status and performance evaluation of flowmeters are important links to ensure production safety and metering accuracy. With the development of industrial automation and intelligent manufacturing, the monitoring of flowmeters has evolved from traditional periodic inspections to continuous monitoring, which can not only detect metering anomalies in a timely manner but also predict equipment failures, and is of great significance for improving production efficiency and reducing maintenance costs. Especially in fields such as chemical industry, oil and natural gas, the monitoring quality of vortex flowmeters is directly related to the accuracy of process control and material statistics.

[0003] In related technologies, the monitoring methods for vortex flowmeters mainly adopt the methods of periodic verification and comparison. Staff use portable standard devices to conduct on-site verification of in-use flowmeters, or conduct comparative measurements by installing standard flowmeters in series online. These methods evaluate the performance status of the flowmeter by comparing the deviation between the measured value and the standard value. At the same time, by recording the output signal characteristics of the flowmeter, such as parameters like signal amplitude and signal frequency, to determine whether the equipment is in a normal working state.

[0004] However, this periodic detection method has the problem of monitoring lag. Due to the lack of effective monitoring during the detection interval, the performance degradation or failure of the flowmeter is often discovered only after it has affected the production quality. This passive way of discovering problems will cause significant economic losses. Summary of the Invention

[0005] This application provides a monitoring method, device, medium and product for vortex flowmeters based on deep learning, which is used to improve the timeliness of monitoring of vortex flowmeters and reduce their maintenance costs.

[0006] In a first aspect, the present application provides a monitoring method for a vortex flowmeter based on deep learning, which is applied to a monitoring device. The method includes: obtaining historical time-series data of the vortex flowmeter, and preprocessing the historical time-series data to obtain standardized data; based on a preset random number seed, dividing the standardized data into a training set and a validation set in a preset ratio by using the Bernoulli distribution; according to the flow characteristics of the vortex flowmeter, segmenting the training set according to a preset flow threshold interval, and respectively constructing deep learning models based on the data within each flow threshold interval to obtain a flow model set; using the validation set to validate multiple deep learning models in the flow model set, and optimizing the flow model set based on the validation error results; collecting real-time data of the vortex flowmeter, and according to the flow threshold interval to which the real-time data belongs, calling the corresponding deep learning model in the flow model set for calculation to obtain a flow correction value; when the performance of the flowmeter calculated based on the flow correction value is lower than a preset performance threshold, generating an abnormal warning message.

[0007] In the above embodiment, the monitoring device realizes the intelligent real-time monitoring of the vortex flowmeter through the complete processes of obtaining historical data and preprocessing, dividing the data set, constructing the flow model set, validating and optimizing the model, real-time monitoring and warning. By modeling different flow intervals based on the deep learning model, it can accurately capture the performance characteristics of the vortex flowmeter under different working conditions, and timely detect abnormal situations, improving the accuracy and timeliness of monitoring.

[0008] In combination with some embodiments of the first aspect, in some embodiments, the steps of obtaining historical time-series data of the vortex flowmeter and preprocessing the historical time-series data to obtain standardized data specifically include: collecting the original data of the vortex flowmeter within a preset time period, where the original data includes flow values, temperature values, pressure values, and signal frequencies; performing outlier processing on the original data, and calculating the local mean, and filling the identified outliers based on the local mean to obtain initialized data; segmenting the initialized data according to a preset time window length to obtain multiple data segments; the preset time window length is determined based on the natural frequency characteristics of the vortex flowmeter; performing standardization processing on each data segment to obtain standardized data; the standardization processing includes maximum-minimum normalization and Z-score normalization.

[0009] In the above embodiment, the monitoring device ensures the quality and feature expression ability of the input data by performing outlier processing, local mean filling, and standardization processing on the original data, and determining the time window in combination with the natural frequency characteristics of the vortex flowmeter.

[0010] In some embodiments in combination with some embodiments of the first aspect, the step of segmenting the training set according to a preset flow threshold interval based on the flow characteristics of the vortex flowmeter and constructing a deep learning model respectively based on the data within each flow threshold interval to obtain a flow model set specifically includes: dividing the working range of the vortex flowmeter into a laminar flow interval, a transition interval, and a turbulent flow interval based on fluid mechanics theory; within each interval, determining a plurality of flow threshold points according to the corresponding relationship between the Reynolds number and the Strouhal number; segmenting the data of the training set according to the determined flow threshold points to obtain a plurality of sub-datasets; constructing a deep neural network model including a convolutional layer, a recurrent layer, and a fully connected layer respectively for each sub-dataset to obtain a flow model set; the neural network structures of each interval are set with different convolutional kernel sizes and the number of recurrent units according to different fluid characteristics.

[0011] In the above embodiments, the monitoring device realizes the accurate modeling of the vortex flowmeter under different flow states by dividing the working range into different flow state intervals and designing corresponding neural network structures according to fluid mechanics characteristics.

[0012] In some embodiments in combination with some embodiments of the first aspect, before the step of constructing a deep neural network model including a convolutional layer, a recurrent layer, and a fully connected layer respectively for each sub-dataset to obtain a flow model set, the method further includes: extracting the feature matrix of the vortex street signal within each flow threshold interval; performing singular value decomposition on the feature matrix to obtain the main eigenvector; determining the network structure parameters of each interval according to the distribution characteristics of the main eigenvector.

[0013] In the above embodiments, the monitoring device realizes the optimized configuration of network parameters by extracting the feature matrix of the vortex street signal, performing singular value decomposition, and obtaining the main eigenvector to guide the network structure design.

[0014] In some embodiments in combination with some embodiments of the first aspect, before the step of validating a plurality of deep learning models in the flow model set using a validation set and optimizing the flow model set based on the validation error results, the method further includes: constructing a loss function based on the physical characteristics of the vortex flowmeter; the loss function includes a flow prediction error term, a signal frequency consistency constraint term, and a physical law constraint term; the flow prediction error term represents the deviation between the model prediction value and the actual value, the signal frequency consistency constraint term represents the correlation between the prediction result and the frequency of the vortex street signal, and the physical law constraint term is constructed based on Bernoulli's equation and the continuity equation in fluid mechanics; adopting a cross-validation method to evaluate the generalization performance of the flow model set under different data segmentation schemes to obtain a generalization evaluation result; adjusting and optimizing the network structure and hyperparameters of each deep learning model in the flow model set based on the generalization evaluation result.

[0015] In the above embodiments, the monitoring device optimizes the deep learning model by constructing a loss function containing multiple constraint terms and using the cross-validation method to evaluate the model generalization performance.

[0016] Combined with some embodiments of the first aspect, in some embodiments, after the steps of collecting the real-time data of the vortex flowmeter, and according to the flow threshold interval to which the real-time data belongs, calling the corresponding deep learning model in the flow model set for calculation to obtain the flow correction value, the method further includes: constructing a state evaluation model based on time series, using the flow correction values, temperature values and pressure values within a continuous plurality of time windows as inputs to determine the device state score; based on the historical failure case library, performing probability prediction on the potential failure types of the device state score to obtain the failure prediction result; when it is determined that the vortex flowmeter has a performance degradation trend according to the device state score, generating the maintenance time and maintenance suggestions according to the failure prediction result.

[0017] In the above embodiments, the monitoring device realizes the dynamic evaluation of the device state and the failure prediction by constructing a state evaluation model and combining with the historical failure case library.

[0018] Combined with some embodiments of the first aspect, in some embodiments, before the steps of collecting the real-time data of the vortex flowmeter, and according to the flow threshold interval to which the real-time data belongs, calling the corresponding deep learning model in the flow model set for calculation to obtain the flow correction value, the method further includes: obtaining the vortex signal frequency sequence and the fluid parameter sequence within a preset time window; using Fourier transform to calculate the power spectrum of the vortex signal frequency sequence and extracting the main frequency component; based on the corresponding relationship between the main frequency component and the fluid parameter sequence, calculating the Strouhal number under the current working condition; dynamically adjusting the flow threshold interval according to the change trend of the Strouhal number.

[0019] In the above embodiments, the monitoring device realizes the dynamic adjustment of the flow threshold interval by analyzing the power spectrum characteristics of the vortex signal frequency and the change trend of the Strouhal number.

[0020] Second aspect, an embodiment of the present application provides a monitoring device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, and the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the monitoring device to execute the method described in the first aspect and any possible implementation manner in the first aspect.

[0021] Third aspect, an embodiment of the present application provides a computer program product containing instructions, when the above computer program product runs on the monitoring device, enabling the above monitoring device to execute the method described in the first aspect and any possible implementation manner in the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, including instructions that, when running on a monitoring device, cause the monitoring device to execute the method described in the first aspect and any possible implementation manner in the first aspect.

[0023] It can be understood that the monitoring device provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, and will not be elaborated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0025] 1. Since a segmentation modeling and real-time monitoring method based on deep learning is adopted, the historical data of the vortex flowmeter is preprocessed and the training set is divided, and a dedicated deep learning model set is constructed according to the flow characteristics, it is possible to accurately model and real-time monitor the characteristics of different flow ranges, effectively solve the problem that a single model in the prior art is difficult to adapt to the full working conditions and the monitoring accuracy is insufficient, and then realize the intelligent real-time monitoring of the performance of the vortex flowmeter, improve the monitoring accuracy and timeliness, and reduce the maintenance cost.

[0026] 2. Since a zoning modeling strategy based on fluid mechanics theory is adopted, the working range is divided into laminar flow, transition and turbulent flow intervals, and the flow threshold points are determined according to the corresponding relationship between the Reynolds number and the Strouhal number, and dedicated neural network structures are designed for different intervals, it is possible to fully consider the fluid characteristic differences under different flow states, effectively solve the problem that the model in the prior art is difficult to accurately describe the characteristics of different flow states, and then realize the accurate modeling of the vortex flowmeter within the full working condition range, and improve the adaptability and prediction accuracy of the model.

[0027] 3. Since an adaptive zoning method based on the frequency analysis of the vortex street signal is adopted, the main frequency characteristics are extracted by Fourier transform, and the flow threshold interval is dynamically adjusted in combination with the change trend of the Strouhal number, it is possible to adapt to the working condition changes in real time and optimize the model zoning, effectively solve the problem that fixed zoning in the prior art is difficult to cope with the working condition changes, and then realize the adaptive optimization of the flowmeter monitoring model, and improve the monitoring performance of the model under dynamic working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flowchart of a method for monitoring a vortex flowmeter based on deep learning in an embodiment of the present application;

[0029] Figure 2It is another process schematic diagram of the vortex flowmeter monitoring method based on deep learning in the embodiments of the present application;

[0030] Figure 3 It is a schematic structural diagram of an entity device of the monitoring device in the embodiments of the present application. Detailed implementation manners

[0031] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification of the present application, the singular forms "a", "an", "the above", "the", and "this" are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or indicating relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise stated, the meaning of "a plurality" is two or more.

[0033] For ease of understanding, the application scenarios of the embodiments of the present application are introduced below.

[0034] In an ethylene production line of a large chemical plant, vortex flowmeters are used to monitor the flow rates of raw materials and products. Due to strict process requirements, the accuracy of flow measurement directly affects product quality and production safety. During long-term operation, traditional vortex flowmeters will have problems such as zero drift and range change, and these performance degradations are often gradual and difficult to detect. For example, in the first quarter of a certain year, the plant found that the quality of some products was abnormal. After investigation, it was found that it was caused by the decrease in the accuracy of the flowmeter, resulting in an imbalance in the raw material ratio, causing an economic loss of about one million yuan. In addition, the production stoppage losses and labor costs caused by regular calibration and maintenance are also very large. Therefore, how to achieve real-time monitoring and predictive maintenance of vortex flowmeters has become an urgent problem in the process industry.

[0035] In the related art, the performance monitoring of vortex flowmeters can be achieved by combining periodic calibration and manual inspection. Specifically, regular calibration is carried out through a portable standard device, or the performance of the flowmeter is evaluated by means of on-line comparison. At the same time, the operating parameters of the device are recorded by manual inspection to judge the working state. The scenarios of using the vortex flowmeter monitoring method based on deep learning in the related art are introduced below.

[0036] Currently, the chemical plant monitors the vortex flowmeter by combining regular calibration and manual inspection. Every quarter, metrology personnel are arranged to carry portable standard devices to conduct on-site calibration of key flowmeters, and manual inspections are carried out twice a week to record the output signal characteristics of the flowmeters. However, this method has obvious defects: First, there is a lack of effective monitoring during the calibration interval, and abnormal performance of the flowmeter is often discovered only after it has affected the process quality; Second, manual inspection mainly relies on experience judgment, making it difficult to detect minor abnormalities in a timely manner; Third, the fixed-cycle calibration and maintenance method is neither economical nor flexible enough to optimize the maintenance strategy according to the actual state of the equipment. For example, during a routine calibration in a certain month and year, it was found that the accuracy of a certain flowmeter exceeded the standard, and through tracing production data, it was found that the problem had occurred one month ago.

[0037] By adopting the deep learning-based vortex flowmeter monitoring method in the embodiments of the present application, through constructing a deep learning model set with sub-intervals and a multi-objective loss function, real-time evaluation of the flowmeter performance and fault prediction are achieved, which not only improves the timeliness and accuracy of monitoring, but also can predict potential faults and give maintenance suggestions. The following introduces the scenario where the deep learning-based vortex flowmeter monitoring method in the present application is used.

[0038] After adopting the deep learning monitoring method of the present application, the monitoring device can collect and analyze the operation data of the vortex flowmeter in real time. The system automatically selects the most suitable deep learning model for calculation according to the working range of the flowmeter to accurately evaluate the equipment status. For example, in a certain month and year, the system detected that the correction coefficient of a flowmeter showed abnormal fluctuations. Through analyzing historical data and physical model prediction, it was judged that the sensitive element of the sensor might start to deteriorate. The system promptly sent out a warning message and recommended a maintenance inspection within two weeks. The inspection results confirmed that there was indeed a problem with the wear of the sensitive element, and timely replacement avoided possible measurement inaccuracies. In this way, the chemical plant has significantly improved the reliability of the flowmeter and reduced the quality accidents related to metrology.

[0039] It can be seen that by adopting the deep learning-based vortex flowmeter monitoring method in the embodiments of the present application, while realizing real-time monitoring of the flowmeter, it can also effectively solve problems such as monitoring lag, high false alarm rate, and high maintenance cost in traditional methods, and thus achieve accurate evaluation of the equipment status and predictive maintenance.

[0040] For ease of understanding, the following combines the above scenario to describe the process of the method provided in this embodiment. Please refer to Figure 1 , which is a flow schematic diagram of the deep learning-based vortex flowmeter monitoring method in the embodiments of the present application.

[0041] S101. Obtain the historical time-series data of the vortex flowmeter, and preprocess the historical time-series data to obtain standardized data.

[0042] Among them, historical time-series data represents a set of data continuously recorded by a vortex flowmeter at a fixed sampling interval within a specific past time period, including parameters such as flow rate values, temperature values, pressure values, and signal frequencies; preprocessing refers to the process of performing operations such as cleaning, denoising, and formatting on the original data to improve data quality; standardized data represents a data sequence after normalization processing, enabling unified calculation of parameters with different dimensions.

[0043] Before starting to build a deep learning model, the monitoring device needs to first obtain sufficient historical data as the basis for model training. Specifically, the monitoring device first establishes a connection with the vortex flowmeter through a data acquisition interface, sets an appropriate sampling time interval (such as 1 second, 5 seconds, etc.), and continuously collects operation data for a preset time period (such as 3 months); then performs outlier detection on the collected original data, uses the 3σ criterion to identify and mark outlier data points; then calculates the mean within a local time window for filling in the outliers; then segments the processed data according to the preset time window length (determined based on the inherent frequency characteristics of the vortex flowmeter) to obtain multiple data segments; finally, performs normalization processing on each data segment, including maximum-minimum normalization and Z-score normalization, to make the data distribution more suitable for the training of subsequent deep learning models; the sampling interval refers to the time interval between two data acquisitions, usually determined according to the response characteristics of the flowmeter; the data quality index is used to measure the integrity, accuracy, and consistency of the data.

[0044] In some embodiments, data preprocessing and normalization can be achieved in various ways: Optionally, the monitoring device can adopt a statistics-based method for data processing. First, use the moving average method to smooth the original data to eliminate high-frequency noise; then calculate the mean μ and standard deviation σ of the data, and mark the data points outside the range of [μ - 3σ, μ + 3σ] as outliers; finally, use the local linear regression method to correct the outliers. Optionally, the monitoring device can also adopt a wavelet transform-based method. First, perform multi-scale decomposition on the original signal to separate different frequency components; then perform threshold processing on the coefficients at each scale to remove noise components; finally, obtain the cleaned signal through wavelet reconstruction. It can be understood that other data preprocessing methods can also be adopted, such as signal processing techniques such as principal component analysis and independent component analysis, which are not limited here.

[0045] S102. Based on a preset random number seed, divide the standardized data into a training set and a validation set in a preset ratio using the Bernoulli distribution.

[0046] Among them, the preset random number seed represents a fixed value used to initialize the random number generator, ensuring the repeatability of data partitioning; the Bernoulli distribution refers to a probability distribution with only two possible outcomes, which is used here to randomly determine the allocation of data samples; the training set refers to the data subset used for model training; the validation set refers to an independent data subset used for evaluating the model performance.

[0047] After the monitoring device completes data standardization, it is necessary to divide the data set into two parts, training and validation, to ensure the objectivity of model evaluation. Specifically, the monitoring device first sets a fixed random number seed (such as 9191972) to ensure the repeatability of the experiment; then applies the Bernoulli distribution to each data sample to generate a binary random variable (1 represents the training set, probability p = 0.7; -1 represents the validation set, probability q = 0.3); finally, allocates the data samples to the corresponding sets according to the values of the random variables to obtain a training set and a validation set with a ratio of 7:3.

[0048] In some embodiments, the data set partitioning can be achieved in various ways: Optionally, the monitoring device can use a stratified sampling strategy. First, divide the data into multiple layers according to the flow rate value, and perform sampling within each layer according to the Bernoulli distribution to ensure that both the training set and the validation set after partitioning can contain samples in each flow rate range; then check the data distribution similarity of the two sets and adjust the sampling ratio if necessary; finally, perform data index rearrangement to obtain the final training set and validation set. Optionally, a time series partitioning method can also be used, where the data in the earlier time period is used for training, and the data in the later time period is used for validation, and a buffer period is set between the two time periods to avoid data leakage. It can be understood that other data partitioning methods, such as random K-fold cross-validation, can also be used, which are not limited here.

[0049] S103. According to the flow characteristics of the vortex flowmeter, segment the training set according to the preset flow threshold intervals, and respectively construct deep learning models based on the data within each flow threshold interval to obtain a flow model set.

[0050] Among them, the flow characteristics represent the measurement characteristics shown by the vortex flowmeter in different flow rate ranges; the flow threshold interval refers to the different flow regime ranges divided according to the hydrodynamic characteristics; the deep learning model refers to a mathematical model containing multiple layers of neural networks; the flow model set represents a set of multiple dedicated models constructed for different flow rate intervals.

[0051] The monitoring device needs to construct a dedicated model according to the characteristic differences in different flow rate intervals. Specifically, the monitoring device first divides the working range into a laminar flow interval, a transition interval, and a turbulent flow interval based on fluid mechanics theory; then within each interval, it determines specific flow rate threshold points according to the corresponding relationship between the Reynolds number and the Strouhal number; then segments the training set data according to these threshold points to obtain multiple sub-datasets; finally, constructs a deep neural network model containing a convolutional layer, a recurrent layer, and a fully connected layer for each sub-dataset, where the network structures of each interval are set with different convolutional kernel sizes and the number of recurrent units according to different fluid characteristics.

[0052] In some embodiments, the model construction can be achieved in multiple ways: Optionally, the monitoring device can adopt an adaptive modeling strategy. First, it analyzes the characteristics of the data in each flow rate interval, extracts key features such as signal frequency features, temperature and pressure features, etc.; then automatically adjusts the number of network layers and neurons according to the complexity of the features; finally, determines the optimal network structure through iterative optimization. Optionally, an ensemble learning method can also be used to construct multiple deep learning models with different structures within each flow rate interval and form a final prediction model through weighted combination. It can be understood that other modeling methods can also be used, such as transfer learning, multi-task learning, etc., which are not limited here.

[0053] S104. Validate multiple deep learning models in the flow rate model set using the validation set, and optimize the flow rate model set based on the validation error results.

[0054] Among them, the validation error represents the deviation between the model prediction value and the actual value; model optimization refers to adjusting the model structure and parameters according to the validation results; flow rate model set optimization refers to improving the performance of the entire model set.

[0055] The monitoring device needs to evaluate and improve the performance of the constructed model. Specifically, the monitoring device first constructs a loss function containing a flow rate prediction error term, a signal frequency consistency constraint term, and a physical law constraint term based on the physical characteristics of the vortex flowmeter; then uses the validation set data to evaluate the performance of the model in each flow rate interval, calculates the prediction error and the generalization error; then adjusts the network structure and hyperparameters of the model according to the evaluation results, including the number of layers, the number of neurons, the activation function, etc.; finally, through multiple rounds of iterative optimization, obtains an optimal performance model set.

[0056] It should be noted that in the construction of the loss function, the total loss function L consists of three terms: L = α1L1 + α2L2 + α3L3. Among them, for the flow prediction error term L1, an improved Huber loss is used to calculate the deviation between the predicted flow and the actual flow. The mean square error form is used when the deviation is small to maintain sensitivity, and the absolute error form is used when the deviation is large to reduce the influence of outliers. The signal frequency consistency constraint term L2 ensures the consistency between the prediction result and the actual physical signal by calculating the cross-correlation coefficient between the theoretical frequency corresponding to the predicted flow and the measured vortex street signal frequency. The physical law constraint term L3 is constructed based on the Bernoulli equation and the continuity equation, integrating the basic laws of fluid dynamics into the model training process. The weight coefficients α1, α2, and α3 are determined by the grid search method on the validation set to obtain the optimal combination. For example, in the monitoring of a vortex flowmeter in a chemical plant, the optimized weight ratio is 0.5:0.3:0.2, which effectively improves the physical interpretability of the model while ensuring the prediction accuracy.

[0057] S105. Collect the real-time data of the vortex flowmeter, and according to the flow threshold interval to which the real-time data belongs, call the corresponding deep learning model in the flow model set for calculation to obtain the flow correction value.

[0058] Among them, the real-time data refers to the parameter values such as the flow rate, temperature, and pressure collected by the vortex flowmeter at the current moment; the flow threshold interval judgment refers to determining the working interval to which it belongs according to the real-time flow rate value; the flow correction value represents the flow rate value calibrated by the deep learning model.

[0059] After the monitoring device completes model training and optimization, it needs to perform online real-time calculation. Specifically, the monitoring device first obtains the measurement data of the vortex flowmeter in real time through the data acquisition interface, including the original flow rate value, temperature value, pressure value, and signal frequency; then preprocesses and standardizes the collected data to meet the model input requirements; then determines the threshold interval to which the current flow rate value belongs and calls the corresponding deep learning model; finally, inputs the preprocessed data into the model, calculates the corrected flow rate value, and outputs the result to the monitoring system.

[0060] In some embodiments, real-time calculation can be achieved in various ways: Optionally, the monitoring device can adopt a sliding window strategy to maintain a cache of historical data of a certain length, and perform comprehensive analysis in combination with the current collected value to improve the stability of the calculation result; then optimize the calculation efficiency through model quantization and compression technologies; finally, achieve a response speed of milliseconds. Optionally, a distributed computing framework can also be adopted to deploy models in different flow rate intervals on multiple processing units for parallel calculation. It can be understood that other calculation methods, such as edge computing and stream computing, can also be adopted, which are not limited here.

[0061] S106. When the performance of the flowmeter calculated based on the flow correction value is lower than the preset performance threshold, an abnormal warning message is generated.

[0062] Among them, the performance index of the flowmeter represents the comprehensive evaluation value of the measurement accuracy and reliability of the vortex flowmeter; the preset performance threshold represents the acceptable minimum performance standard; the abnormal warning message includes the evaluation result of the equipment status and the maintenance suggestion.

[0063] The monitoring device needs to evaluate the equipment status in real time and issue warnings in a timely manner. Specifically, the monitoring device first constructs a state evaluation model based on time series, using the flow correction value, temperature value, and pressure value within multiple consecutive time windows as inputs; then calculates various performance indicators, including measurement error, signal stability, response characteristics, etc.; then compares the calculation results with the preset threshold to determine whether there is a trend of performance degradation; finally, when an abnormality is detected, it identifies the fault type according to the historical fault case library and generates a warning message including the fault prediction result, maintenance time, and maintenance suggestion.

[0064] In some embodiments, the state warning determination adopts a multi-index comprehensive evaluation mechanism. The system calculates multiple performance indicators based on the flow correction value, including: root mean square of relative error (RMSE), signal stability index (SSI), and dynamic response characteristic (DRC). After normalizing these indicators, the comprehensive performance index CPI is obtained through weighted summation. The preset performance threshold is divided into three levels: warning level (CPI < 0.85), alarm level (CPI < 0.75), and emergency level (CPI < 0.65). At the same time, the system also tracks the change rate of CPI. When the change rates of three consecutive evaluations are all negative and the cumulative decrease exceeds the preset threshold, even if the CPI has not reached the warning level, a warning will be triggered. For example, the CPI of a certain flowmeter gradually drops from 0.92 to 0.87 within a week. Although it has not reached the warning level, due to the obvious continuous downward trend, the system still generates a warning message to remind the operation and maintenance personnel to pay attention to the equipment status.

[0065] It should be noted that the state evaluation model adopts a multi-level time series feature extraction architecture. First, multi-scale decomposition is performed on the flow correction value, temperature value, and pressure value within multiple consecutive time windows to obtain the change characteristics at different time scales. Then, local pattern features are extracted through a convolutional neural network, and a bidirectional long short-term memory network (Bi-LSTM) is used to capture long-term dependencies. Finally, the equipment state score is obtained through mapping by a fully connected layer. When training the model, the known fault cases in the historical operation data are used as labeled samples, and the model parameters are optimized by minimizing the cross-entropy loss function. The state score is represented by normalization from 0 to 1, where 1 represents the best state of the equipment and 0 represents the complete failure of the equipment. For example, when the state score shows a downward trend and the decline exceeds 20% within three consecutive evaluation periods, the system will trigger a performance degradation warning.

[0066] In the above embodiments, the monitoring device processes data in different flow rate intervals through multiple deep learning models and optimizes them in combination with physical constraints. In practical applications, this way of modeling by intervals can better adapt to the non-linear characteristics of the vortex flowmeter under different working conditions and improve the prediction accuracy of the model. The following supplements the scenarios of this embodiment.

[0067] As the system continues to run and data accumulates, the performance of the monitoring method is further improved. By analyzing a large number of historical cases, the system has established a complete fault feature library and prediction model. For example, in a certain month, based on historical data analysis, the system found that a certain type of flowmeter is prone to zero drift under specific working conditions, and then automatically adjusted the monitoring parameters and warning thresholds under these working conditions. At the same time, the system can automatically optimize the maintenance plan according to the equipment status, achieving true predictive maintenance. After implementing this solution for half a year, the unplanned downtime of the flowmeters in the chemical plant has been reduced by 85%, the maintenance cost has been reduced by 30%, and the product quality stability has been improved by 25%. In addition, the data and experience accumulated by the system have also helped the manufacturer optimize the product design and further improve the reliability of new products.

[0068] After combining the above scenarios, the following further describes the more specific process of the method provided in this embodiment in detail. Please refer to Figure 2 , which is another process schematic diagram of the vortex flowmeter monitoring method based on deep learning in the embodiments of the present application.

[0069] S201. Obtain the historical time series data of the vortex flowmeter, and preprocess the historical time series data to obtain standardized data.

[0070] Referring to step S101, the monitoring device will obtain the historical operation data of the vortex flowmeter through the data acquisition interface, perform outlier processing and standardization on the data to ensure that the data quality meets the modeling requirements.

[0071] In some embodiments, the monitoring device will perform maximum-minimum normalization and Z-score standardization on the data, that is, the monitoring device will collect the original data of the vortex flowmeter within a preset time period, and the original data includes flow rate value, temperature value, pressure value and signal frequency; perform outlier processing on the original data, calculate the local mean value, fill the identified outliers based on the local mean value to obtain the initialized data; segment the initialized data according to the preset time window length to obtain multiple data segments; the preset time window length is determined based on the inherent frequency characteristics of the vortex flowmeter; perform standardization processing on each data segment to obtain standardized data; the standardization processing includes maximum-minimum normalization and Z-score standardization.

[0072] Among them, the original data represents the unprocessed real-time measurement data collected from the vortex flowmeter, including the flow value, fluid temperature value, pressure value generated by fluid flow, and the frequency of the vortex street oscillation signal. The local mean refers to the arithmetic mean of the data within a preset time window. An outlier refers to a data point whose deviation from the local mean exceeds a preset threshold. A data segment represents a continuous data sequence divided by a fixed time window length. The normalization process refers to the mathematical transformation process of converting the data to a unified scale.

[0073] During the normal operation of the vortex flowmeter, the monitoring device needs to continuously collect and process the measurement data to evaluate the device status. Specifically, the monitoring device first continuously collects the original data for a period of time at a preset sampling frequency, then calculates the local mean using a sliding time window, identifies and corrects the outlier data points. Then, according to the inherent oscillation frequency of the vortex flowmeter, the appropriate time window length is determined, and the processed data is segmented into equal-length segments. Finally, maximum-minimum normalization and Z-score normalization are performed on each data segment to obtain the normalized data with a mean of 0 and a variance of 1.

[0074] In some embodiments, data preprocessing can be achieved in the following manner: the monitoring device first uses the median absolute deviation method to detect outliers, calculates the median and deviation distribution of the data, and marks the points that deviate more than 3 standard deviations from the median as outliers. Then, local polynomial regression is used to interpolate and correct the outliers to maintain the temporal continuity of the data. Finally, based on Fourier analysis, the main frequency components of the signal are determined, and a time window not less than 3 vortex street oscillation periods is selected for data segmentation and normalization processing. Optionally, an outlier detection method based on kernel density estimation can also be used, combined with the multi-scale analysis method of wavelet transform for data preprocessing. The specific processing parameters need to be optimized and adjusted according to the actual working characteristics of the vortex flowmeter. It can be understood that other data preprocessing methods can also be used to clean and normalize the original data, which is not limited here.

[0075] S202. Based on a preset random number seed, divide the normalized data into a training set and a validation set with a preset ratio using the Bernoulli distribution.

[0076] Referring to step S102, the monitoring device will initialize the data division process using a fixed random number seed, and divide the normalized data into a training set and a validation set in a ratio of 7:3 through the Bernoulli distribution.

[0077] S203. Based on the fluid mechanics theory, divide the working range of the vortex flowmeter into a laminar flow interval, a transition interval, and a turbulent flow interval.

[0078] Among them, the hydrodynamics theory represents the basic physics principle describing the laws of fluid motion and is used to guide the flow regime analysis; the working range refers to the Reynolds number range within which the vortex flowmeter can effectively measure; the laminar flow interval represents the state range in which the fluid flows in an orderly layered manner; the transition interval refers to the intermediate state range in which the fluid transitions from laminar flow to turbulent flow; the turbulent flow interval represents the state range in which the fluid exhibits irregular mixed motion; the Reynolds number is a dimensionless parameter characterizing the ratio of the inertial force to the viscous force of the fluid.

[0079] Before the monitoring device constructs the model in segments, it is necessary to first scientifically divide the flow regime based on the hydrodynamics theory. Specifically, the monitoring device first calculates the critical Reynolds number according to parameters such as the pipe diameter, fluid viscosity, and density; then it demarcates the range where the Reynolds number is less than 2300 as the laminar flow interval, at this time the fluid motion presents regular layered flow; then it demarcates the range where the Reynolds number is between 2300 and 4000 as the transition interval, at this time the fluid motion state is unstable and has intermittent characteristics; afterwards, it demarcates the range where the Reynolds number is greater than 4000 as the turbulent flow interval, at this time the fluid motion presents strong three-dimensional random fluctuations; finally, it establishes a characteristic parameter database for each interval for subsequent model construction.

[0080] In some embodiments, the flow regime interval division can be achieved in various ways: Optionally, the monitoring device can adopt a multi-parameter comprehensive judgment method. First, it constructs a flow regime discrimination criterion by combining dimensionless parameters such as the Reynolds number, Froude number, and Weber number; then it refines the interval boundaries by analyzing the velocity pulsation characteristics and pressure fluctuation characteristics of the fluid; finally, it establishes fuzzy logic rules to achieve a smooth transition between intervals. Optionally, the monitoring device can also adopt a data-driven method. First, it identifies different characteristic patterns of fluid motion through cluster analysis; then it uses time-frequency analysis methods to extract the main frequency components of each interval; finally, it establishes an adaptive interval division standard based on statistical characteristics. It can be understood that other flow regime division methods can also be adopted, such as phase space reconstruction, entropy value analysis and other technologies, which are not limited here.

[0081] S204. Within each interval, determine multiple flow threshold points according to the corresponding relationship between the Reynolds number and the Strouhal number.

[0082] Among them, the corresponding relationship between the Reynolds number and the Strouhal number represents a mathematical model describing the correlation between the vortex street generation frequency and the fluid motion state; the flow threshold point is the key numerical point for distinguishing different flow measurement characteristics; the Strouhal number represents the dimensionless ratio of the vortex street frequency to the flow velocity; the corresponding relationship curve is used to characterize the change trend of the vortex street generation law under different flow regimes.

[0083] The monitoring device needs to determine specific flow threshold points within the divided flow regime intervals to achieve more accurate piecewise modeling. Specifically, the monitoring device first collects a large amount of experimental data and establishes a database of the corresponding relationship between the Reynolds number and the Strouhal number; then analyzes the change trend of the Strouhal number within each flow regime interval and identifies the inflection points and mutation points of the curve; then calculates the flow values corresponding to these characteristic points in combination with fluid mechanics theory; then considers the measurement accuracy requirements and adds transition threshold points near the characteristic points; finally, sorts all the threshold points and establishes an index for subsequent data segmentation.

[0084] It should be noted that the calculation process of the Strouhal number (St) comprehensively considers the influence of temperature and pressure on fluid characteristics. First, according to the temperature and pressure values of the current working condition, the state equation is used to calculate the actual density and viscosity of the fluid. Then, based on the main frequency component f and the characteristic length d (usually the characteristic size of the flowmeter detection body), combined with the flow velocity v, the initial Strouhal number St_0 = fd / v is calculated. Then, the temperature correction coefficient K_t and the pressure correction coefficient K_p are introduced, and these two coefficients are obtained through polynomial fitting. Finally, the corrected Strouhal number St = St_0·K_t·K_p is obtained. For example, when the temperature rises, since the decrease in fluid viscosity will cause changes in the vortex shedding characteristics, the temperature correction coefficient K_t will be less than 1 at this time, so as to compensate for the temperature influence.

[0085] S205. Segment the data in the training set according to the determined flow threshold points to obtain multiple sub-datasets.

[0086] Among them, the segmentation operation refers to the process of dividing the dataset into multiple continuous intervals according to the flow threshold points; the sub-dataset refers to the set of data samples within each flow interval; the flow interval boundary represents the demarcation point between adjacent sub-datasets; the data distribution characteristics are used to describe the statistical laws of the samples within each sub-dataset; the data density refers to the distribution density of the sample points in each sub-dataset.

[0087] After the monitoring device determines the flow threshold points, it needs to scientifically segment the training data for subsequent modeling. Specifically, the monitoring device first sorts the data in the training set according to the flow value; then divides the data into multiple intervals according to the determined flow threshold points; then checks whether the data volume in each interval meets the minimum sample number requirement, and adjusts the positions of adjacent threshold points if necessary; then analyzes the statistical characteristics of the data in each interval, including the mean, variance, and distribution form; finally, numbers and marks each sub-dataset and establishes a data index for quick access.

[0088] In some embodiments, data segmentation can be achieved in various ways: Optionally, the monitoring device can adopt an adaptive segmentation strategy. First, calculate the gradient change of the data near each threshold point; then dynamically adjust the width of the segmentation interval according to the gradient magnitude; finally, ensure the relative balance of the data volume of each sub-dataset through sample density analysis. Optionally, the monitoring device can also adopt an overlapping segmentation method. First, set an overlapping area between adjacent intervals; then process the data in the overlapping area using a weighted average method; finally, ensure the continuity of the model at the interval boundaries through smooth transition. It can be understood that other data segmentation methods can also be adopted, such as quantile segmentation, clustering segmentation and other techniques, which are not limited here.

[0089] S206. For each sub-dataset, construct a deep neural network model including a convolutional layer, a recurrent layer, and a fully connected layer to obtain a traffic model set.

[0090] Among them, according to the different fluid characteristics, different convolutional kernel sizes and the number of recurrent units are set for the neural network structures of each interval. The deep neural network model represents a machine learning model composed of multiple processing layers; the convolutional layer is used to extract the local features and time-frequency characteristics of the signal; the recurrent layer is used to capture the dynamic change rules of the time-series data; the fully connected layer is used to realize the non-linear mapping of the features; the traffic model set refers to a set of multiple dedicated models constructed for different traffic intervals.

[0091] The monitoring device needs to construct a dedicated deep learning model for each sub-dataset. Specifically, the monitoring device first analyzes the feature dimensions and time-series characteristics of each sub-dataset; then designs the network structure, including determining the convolutional kernel size, the type of recurrent unit, and the number of hidden layer nodes; then initializes the model parameters, using the Xavier or He initialization method; then selects an appropriate loss function and optimization algorithm, such as mean squared error loss and Adam optimizer; finally, iteratively optimizes the model parameters through the backpropagation algorithm until convergence or reaching the preset number of training rounds.

[0092] In some embodiments, model construction can be achieved in various ways: Optionally, the monitoring device can adopt a multi-scale modeling strategy. First, construct a feature extraction network including multiple-scale convolutional layers; then use a bidirectional LSTM layer to process the dependencies of different time scales; finally, enhance the expression ability of the model through residual connections and attention mechanisms. Optionally, the monitoring device can also adopt an ensemble learning method. First, construct multiple basic models with different structures for each sub-dataset; then use the Bagging or Boosting method to combine the prediction results of multiple models; finally, determine the optimal ensemble strategy through cross-validation. It can be understood that other modeling methods can also be adopted, such as transfer learning, multi-task learning and other techniques, which are not limited here.

[0093] In some embodiments, the monitoring device first performs network optimization, that is, the monitoring device extracts the feature matrix of the vortex street signal in each flow threshold interval; performs singular value decomposition on the feature matrix to obtain the main eigenvectors; and determines the network structure parameters of each interval according to the distribution characteristics of the main eigenvectors.

[0094] Among them, the feature matrix represents the multi-dimensional feature representation of the vortex street signal in the time domain and frequency domain. Singular value decomposition is a mathematical method that decomposes a matrix into the product of three matrices. The main eigenvectors refer to the set of eigenvectors corresponding to the largest singular values. The network structure parameters include configuration items such as the number of layers, number of nodes, and activation function of the neural network.

[0095] Before constructing the deep learning model, the monitoring device needs to analyze the data characteristics of different flow intervals to determine the appropriate network structure. Specifically, the monitoring device first extracts the time-domain statistical features (mean, variance, skewness, etc.) and frequency-domain features (power spectrum, phase spectrum, etc.) of the vortex street signal to form a multi-dimensional feature matrix. Then, through singular value decomposition for dimensionality reduction, the main eigenvectors with an energy proportion exceeding 95% are retained. Finally, based on the distribution characteristics of the eigenvectors, the structure parameters of the neural network in each flow interval are determined, including the size of the convolutional kernel, the number of recurrent units, etc.

[0096] In some embodiments, feature extraction and network design can be achieved in the following way: First, the short-time Fourier transform of the vortex street signal is calculated using the sliding window method to obtain the time-frequency spectrogram. Then, singular value decomposition is performed on the time-frequency spectrogram to analyze the variation law of the eigenvectors with frequency. According to the main distribution range of the eigenvectors, the receptive field size and the number of filters of the convolutional layer are determined, and at the same time, the number of memory units of the recurrent layer is determined according to the temporal correlation. Optionally, wavelet transform can also be used to extract multi-scale features, combined with principal component analysis for feature dimensionality reduction and network structure optimization. The optimal network configuration is determined through repeated experimental verification. It can be understood that other feature engineering methods can also be used to guide the structure design of the neural network, which is not limited here.

[0097] S207. Verify multiple deep learning models in the flow model set using the validation set, and optimize the flow model set based on the verification error results.

[0098] Referring to step S104, the monitoring device will use the validation set data to evaluate the prediction performance of each deep learning model, and optimize and adjust the model structure and parameters according to the verification error results to improve the generalization ability of the model.

[0099] In some embodiments, the monitoring device dynamically evaluates and adjusts the flow threshold range before invoking the deep learning model to ensure that the model is applicable to the current working condition. That is, the monitoring device constructs a loss function based on the physical characteristics of the vortex flowmeter. The loss function includes a flow prediction error term, a signal frequency consistency constraint term, and a physical law constraint term. The flow prediction error term characterizes the deviation between the model prediction value and the actual value. The signal frequency consistency constraint term characterizes the correlation between the prediction result and the vortex signal frequency. The physical law constraint term is constructed based on Bernoulli's equation and the continuity equation in fluid mechanics. The cross-validation method is used to evaluate the generalization performance of the flow model set under different data segmentation schemes, and a generalization evaluation result is obtained. Based on the generalization evaluation result, the network structure and hyperparameters of each deep learning model in the flow model set are adjusted and optimized.

[0100] Among them, the loss function represents a mathematical expression for measuring the prediction effect of the model. The flow prediction error term is used to quantify the deviation degree between the prediction value and the actual value. The consistency constraint term reflects the correlation degree between the prediction result and the physical signal. The physical law constraint term ensures that the model prediction result conforms to the basic laws of fluid mechanics. The generalization performance refers to the prediction ability of the model for unseen data.

[0101] In the model training and optimization stage, the monitoring device needs to construct an appropriate loss function and evaluate the model performance. Specifically, the monitoring device first constructs a flow prediction error term based on the mean square error, and then introduces the signal frequency correlation constraint and the fluid mechanics equation constraint to form a multi-objective loss function. Then the K-fold cross-validation method is used to evaluate the generalization ability of the model using different data partitioning schemes. Finally, according to the verification result, the structure parameters and training hyperparameters of the network, such as the learning rate, batch size, etc., are adjusted.

[0102] In some embodiments, model optimization can be achieved in the following way: First, construct a weighted multi-objective loss function. The prediction error term uses the Huber loss to reduce the influence of outliers. The consistency constraint uses the cross-correlation coefficient. The physical constraint uses the dimensionless Bernoulli equation residual. Then use the grid search method to optimize the weight coefficient of the loss function, and at the same time adopt strategies such as learning rate decay and early stopping to avoid overfitting. Based on the performance metrics of the model on the validation set, such as the mean relative error, R-squared value, etc., select the optimal model configuration. It can be understood that other model optimization methods can also be used to improve the prediction accuracy, which is not limited here.

[0103] S208. Obtain the vortex signal frequency sequence and fluid parameter sequence within a preset time window.

[0104] Among them, the preset time window represents the fixed time period length for signal analysis; the vortex street signal frequency sequence refers to the vortex street shedding frequency data continuously collected within a period of time; the fluid parameter sequence includes the time series of physical quantities such as temperature and pressure; the sampling frequency refers to the time interval of signal acquisition; the data synchronization represents the corresponding relationship of different parameter sequences in time.

[0105] The monitoring device needs to obtain the device operation data in real time for online analysis. Specifically, the monitoring device first determines the appropriate time window length according to the dynamic characteristics of the vortex flowmeter; then synchronously collects the vortex street signal and fluid parameters through the high-speed data acquisition module; then performs time alignment and synchronization processing on the collected raw data; then checks the integrity and continuity of the data, and performs interpolation and completion if necessary; finally, organizes and stores the processed data sequence in a predetermined format for subsequent analysis.

[0106] In some embodiments, data acquisition and preprocessing can be achieved in various ways: Optionally, the monitoring device can adopt an adaptive sampling strategy, first monitor the change rate of the signal; then dynamically adjust the sampling frequency according to the change rate; finally, ensure the data quality through digital filtering. Optionally, the monitoring device can also adopt a multi-buffer mechanism, first set up multiple levels of data buffers; then implement pipeline processing of the data; finally, ensure the accuracy of the data through real-time verification. It can be understood that other data acquisition methods can also be adopted, such as distributed acquisition, edge computing and other technologies, which are not limited here.

[0107] S209. Calculate the power spectrum of the vortex street signal frequency sequence by Fourier transform, and extract the main frequency component.

[0108] Among them, the Fourier transform represents a mathematical tool for converting a time-domain signal to a frequency domain; the power spectrum is a representation of the energy distribution of a signal in the frequency domain; the main frequency component represents the frequency component with the strongest energy in the power spectrum; the frequency resolution refers to the smallest frequency interval that can be distinguished; the spectral leakage effect represents the phenomenon of spectral energy diffusion caused by a finite sampling length.

[0109] The monitoring device needs to perform spectral analysis on the vortex street signal to obtain the characteristic frequency. Specifically, the monitoring device first performs windowing processing on the signal sequence, selects an appropriate window function (such as a Hanning window) to reduce spectral leakage; then performs a fast Fourier transform on the windowed signal to obtain a complex spectrum; then calculates the square of the modulus of the spectrum to obtain the power spectral density function; then identifies the significant peaks in the power spectrum through a peak detection algorithm; finally, determines the main frequency component characterizing the vortex street shedding characteristics according to the energy magnitude and frequency position of the peaks.

[0110] In some embodiments, spectrum analysis can be achieved in various ways: Optionally, the monitoring device can adopt the short-time Fourier transform method. First, the signal is divided into multiple overlapping short-time windows; then, the Fourier transform is performed on each window separately; finally, the variation characteristics of the dominant frequency over time are analyzed through a time-frequency diagram. Optionally, the monitoring device can also adopt the wavelet analysis method. First, an appropriate wavelet basis function is selected; then, multi-scale wavelet decomposition is performed; finally, the characteristic frequency is extracted through the wavelet energy distribution. It can be understood that other spectrum analysis methods can also be adopted, such as Hilbert transform, empirical mode decomposition and other techniques, which are not limited here.

[0111] S210. Calculate the Strouhal number under the current working condition based on the correspondence between the dominant frequency component and the fluid parameter sequence.

[0112] Among them, the working condition refers to the actual operating state of the fluid, including parameters such as temperature and pressure; the Strouhal number is a dimensionless parameter characterizing the vortex street frequency characteristics; the dominant frequency-flow velocity relationship refers to the functional relationship between the vortex street frequency and the fluid velocity; the parameter correction coefficient is used to compensate for the influence of temperature and pressure on the measurement; the calculation accuracy represents the measurement accuracy of the Strouhal number.

[0113] The monitoring device needs to calculate the Strouhal number in real time to characterize the vortex street characteristics. Specifically, the monitoring device first obtains the fluid physical property parameters under the current working condition, including density, viscosity, etc.; then calculates the characteristic velocity according to the dominant frequency component and the fluid parameters; then considers the influence of temperature and pressure on the fluid physical properties and performs parameter correction; then calculates the Strouhal number using the corrected parameters and the dominant frequency component; finally, analyzes the uncertainty of the calculation result to evaluate the measurement accuracy.

[0114] In some embodiments, the calculation of the Strouhal number can be achieved in various ways: Optionally, the monitoring device can adopt the physical property compensation method. First, establish the mapping relationship between temperature, pressure and fluid physical properties; then update the correction coefficient in real time; finally, obtain the accurate Strouhal number through iterative calculation. Optionally, the monitoring device can also adopt the neural network method. First, train a prediction model including the influence of physical properties; then input the current working condition parameters; finally, directly output the compensated Strouhal number. It can be understood that other calculation methods can also be adopted, such as look-up table interpolation, polynomial fitting and other techniques, which are not limited here.

[0115] S211. Dynamically adjust the flow threshold interval according to the change trend of the Strouhal number.

[0116] Among them, the change trend represents the evolution law of the Strouhal number over time; dynamic adjustment refers to adaptively updating the threshold interval according to real-time data; the adjustment range refers to the change range of the threshold point position; the stability criterion is used to evaluate the necessity of adjustment; the threshold update period represents the time interval for performing the adjustment operation.

[0117] The monitoring device needs to optimize the flow interval division according to the actual operating status. Specifically, the monitoring device first calculates the change rate of the Strouhal number within a certain time window; then analyzes whether the change trend exceeds the preset fluctuation range; then judges whether the current threshold interval is still applicable; then determines whether it is necessary to move the threshold point position according to the judgment result; finally, when it is confirmed that adjustment is needed, calculate the coordinates of the new threshold point and update the interval division.

[0118] In some embodiments, dynamic adjustment can be achieved in various ways: Optionally, the monitoring device can adopt the sliding window analysis method. First, maintain a historical data window with a fixed length; then calculate the statistical features within the window; finally, determine the adjustment strategy based on the feature changes. Optionally, the monitoring device can also adopt the fuzzy control method. First, construct a fuzzy rule base based on the change rate; then evaluate the necessity and range of adjustment in real time; finally, obtain the specific adjustment amount through defuzzification. It can be understood that other adjustment methods can also be adopted, such as adaptive control, state observer and other technologies, which are not limited here.

[0119] It should be noted that the dynamic adjustment of the flow threshold interval is based on the real-time monitoring results of the Strouhal number. The system maintains a sliding time window (usually 24 hours), calculates the mean μ and standard deviation σ of the Strouhal number within the window. When the change trend of the Strouhal number exceeds the range of μ±2σ, the threshold adjustment mechanism is triggered. The adjustment process uses an adaptive step size algorithm, and the step size is proportional to the deviation degree, but there is a maximum adjustment range limit to ensure the system stability. The new threshold position is determined by the exponential smoothing method to avoid the impact of sudden changes on the model prediction. For example, during the process of the fluid temperature gradually rising in a certain flowmeter, the system detects that the overall Strouhal number shifts, and immediately adjusts the threshold point of the corresponding flow interval upward by 5%, so that the model always works within the most suitable range.

[0120] S212. Collect the real-time data of the vortex flowmeter, and call the corresponding deep learning model in the flow model set according to the flow threshold interval to which the real-time data belongs, and obtain the flow correction value.

[0121] Referring to step S105, the monitoring device will collect the operation data of the vortex flowmeter in real time, judge the flow threshold interval to which the data belongs, and call the corresponding deep learning model for flow calculation and correction.

[0122] In some embodiments, the monitoring device establishes an equipment status evaluation model based on continuously monitored data, combines historical failure cases for fault prediction, and generates corresponding maintenance suggestions. That is, the monitoring device constructs a status evaluation model based on time series, uses the flow correction values, temperature values, and pressure values within multiple consecutive time windows as inputs to determine the equipment status score; probabilistically predicts the potential failure types of the equipment status score based on the historical failure case library to obtain the fault prediction result; when it is determined that the vortex flowmeter has a performance degradation trend based on the equipment status score, generates the maintenance time and maintenance suggestions according to the fault prediction result.

[0123] Among them, the status evaluation model refers to a mathematical model used to evaluate the operating status of the equipment. The status score is a numerical index that quantifies the equipment performance level. The fault prediction result includes the potential failure type and its occurrence probability. The performance degradation trend refers to the changing rule that the equipment performance gradually decreases over time.

[0124] In the real-time monitoring stage, the monitoring device needs to continuously evaluate the equipment status and predict potential faults. Specifically, the monitoring device first constructs a status evaluation model based on time series, using the flow correction values, temperature values, and pressure values within multiple time windows as input features. Then, based on the characteristic patterns of historical failure cases, it uses probability statistical methods to predict the possible failure types in the current state. Finally, according to the changing trend of the status score, combined with the fault prediction result, it generates targeted maintenance suggestions and the optimal maintenance time.

[0125] In some embodiments, the status evaluation and fault prediction can be achieved in the following manner: First, construct a status evaluation model based on the long short-term memory network, extract multi-scale time series features and calculate the status score. Then, use the Gaussian mixture model to perform clustering analysis on historical failure cases to establish a fault feature library. Based on the similarity between the current status score and the fault features, calculate the occurrence probability of various faults and determine the maintenance strategy. Optionally, other machine learning methods such as support vector machines can also be used for status evaluation and fault diagnosis. It can be understood that other status monitoring methods can also be used to achieve equipment health management, which is not limited here.

[0126] S213. When the performance of the flowmeter calculated based on the flow correction value is lower than the preset performance threshold, generate an abnormal warning message.

[0127] Referring to step S106, the monitoring device evaluates the performance status of the flowmeter based on the flow correction value. When the performance index is lower than the preset threshold, it generates an abnormal warning message including the fault type, severity, and handling suggestions.

[0128] In the embodiments of the present application, due to the adoption of a sub-region modeling strategy based on hydrodynamic characteristics, a dedicated deep learning model is constructed according to the different characteristics of the laminar flow, transitional flow, and turbulent flow regions; the design of a multi-objective loss function integrating physical constraints takes the hydrodynamic laws as the constraint conditions for model training; the state evaluation and fault prediction method based on time series analysis realizes the accurate quantification of the equipment state and the early warning of faults. Therefore, it can maintain a high monitoring accuracy under different working conditions, effectively solving problems such as monitoring lag, high false alarm rate, and low maintenance efficiency in traditional methods. Furthermore, four important goals are achieved: improving the monitoring timeliness of the vortex flowmeter, reducing the equipment maintenance cost, enhancing the metering reliability, and realizing predictive maintenance.

[0129] The monitoring device in the embodiments of the present invention application will be described from the perspective of hardware processing. Please refer to Figure 3 , which is a schematic structural diagram of an entity device of the monitoring device in the embodiments of the present application.

[0130] It should be noted that Figure 3 The structure of the monitoring device shown is only an example and should not bring any limitations to the functions and usage scopes of the embodiments of the present invention.

[0131] As Figure 3 shown, the monitoring device includes a CPU 301, which can perform various appropriate actions and processes according to the program stored in the ROM 302 or the program loaded from the storage section 308 into the RAM 303, such as executing the method described in the above embodiments. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other through a bus 304. The I / O interface 305 is also connected to the bus 304.

[0132] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, a button switch, etc.; an output section 307 including a liquid crystal display (LCD), an audio output device, an indicator light, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A driver 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the driver 310 as needed so that the computer program read from it can be installed into the storage section 308 as needed.

[0133] In particular, according to an embodiment of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present invention includes a computer program product that includes a computer program carried on a computer-readable medium, and the computer program contains a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 309 and / or installed from the removable medium 311. When the computer program is executed by the CPU 301, various functions defined in the present invention are executed.

[0134] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. Among them, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings.

[0135] Specifically, the monitoring device of this embodiment includes a processor and a memory. A computer program is stored on the memory. When the computer program is executed by the processor, the monitoring method of the vortex flowmeter based on deep learning provided in the above embodiment is implemented.

[0136] On the other hand, the present invention also provides a computer-readable storage medium, which may be included in the monitoring device described in the above embodiment; or it may exist separately and not be assembled into the monitoring device. The above storage medium carries one or more computer programs. When the above one or more computer programs are executed by a processor of the monitoring device, the monitoring device is enabled to implement the monitoring method of the vortex flowmeter based on deep learning provided in the above embodiment.

[0137] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application 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 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 various embodiments of the present application.

[0138] As used in the foregoing embodiments, depending on the context, the term "when" can be interpreted to mean "if" or "after" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "upon determining" or "if (the stated condition or event) is detected" can be interpreted to mean "if determined" or "in response to determining" or "when (the stated condition or event) is detected" or "in response to detecting (the stated condition or event)".

Claims

1. A monitoring method for vortex flowmeters based on deep learning, characterized in that, Applied to a monitoring device, the method includes: Obtain the historical time-series data of the vortex flowmeter, and preprocess the historical time-series data to obtain standardized data; Based on a preset random number seed, divide the standardized data into a training set and a validation set with a preset ratio using the Bernoulli distribution; Based on the fluid mechanics theory, divide the working range of the vortex flowmeter into a laminar flow interval, a transition interval, and a turbulent flow interval; within each interval, determine multiple flow threshold points according to the correspondence between the Reynolds number and the Strouhal number; segment the data in the training set according to the determined flow threshold points to obtain multiple sub-datasets; Construct a deep neural network model including a convolutional layer, a recurrent layer, and a fully connected layer for each sub-dataset to obtain a flow model set; the neural network structures in each interval are set with different convolutional kernel sizes and the number of recurrent units according to different fluid characteristics; Construct a loss function based on the physical characteristics of the vortex flowmeter; the loss function includes a flow prediction error term, a signal frequency consistency constraint term, and a physical law constraint term; the flow prediction error term represents the deviation between the model prediction value and the actual value, the signal frequency consistency constraint term represents the correlation between the prediction result and the vortex signal frequency, and the physical law constraint term is constructed based on the Bernoulli equation and the continuity equation in fluid mechanics; Adopt a cross-validation method to evaluate the generalization performance of the flow model set under different data segmentation schemes to obtain a generalization evaluation result; adjust and optimize the network structure and hyperparameters of each deep learning model in the flow model set based on the generalization evaluation result; Validate the multiple deep learning models in the flow model set using the validation set, and optimize the flow model set based on the validation error result; Collect the real-time data of the vortex flowmeter, and call the corresponding deep learning model in the flow model set for calculation according to the flow threshold interval to which the real-time data belongs to obtain a flow correction value; When the performance of the flowmeter calculated based on the flow correction value is lower than a preset performance threshold, generate an abnormal warning message.

2. The method according to claim 1, wherein The step of obtaining the historical time-series data of the vortex flowmeter and preprocessing the historical time-series data to obtain standardized data specifically includes: Collect the original data of the vortex flowmeter within a preset time period, where the original data includes flow values, temperature values, pressure values, and signal frequencies; Perform outlier processing on the original data, calculate the local mean, and fill the identified outliers based on the local mean to obtain initialized data; Segment the initialized data according to a preset time window length to obtain multiple data segments; the preset time window length is determined based on the inherent frequency characteristics of the vortex flowmeter; Perform standardized processing on each data segment to obtain the standardized data; the standardized processing includes maximum-minimum normalization and Z-score normalization.

3. The method according to claim 1, characterized in that Before the step of constructing a deep neural network model including a convolutional layer, a recurrent layer, and a fully connected layer for each sub-dataset to obtain a flow model set, the method further includes: Extract the feature matrix of the vortex street signal within each flow threshold interval; Perform singular value decomposition on the feature matrix to obtain the main eigenvectors; Determine the network structure parameters of each interval according to the distribution characteristics of the main eigenvectors.

4. The method according to claim 1, characterized in that After the step of collecting the real-time data of the vortex flowmeter and calling the corresponding deep learning model in the flow model set for calculation according to the flow threshold interval to which the real-time data belongs to obtain the flow correction value, the method further includes: Construct a state evaluation model based on time series, and use the flow correction values, temperature values, and pressure values within a continuous plurality of time windows as inputs to determine the device state score; Perform probability prediction on the potential failure types of the device state score based on the historical failure case library to obtain the failure prediction result; When it is determined that the vortex flowmeter has a performance degradation trend according to the device state score, generate the maintenance time and maintenance suggestions according to the failure prediction result.

5. The method according to claim 1, wherein Before the step of collecting the real-time data of the vortex flowmeter and calling the corresponding deep learning model in the flow model set for calculation according to the flow threshold interval to which the real-time data belongs to obtain the flow correction value, the method further includes: Obtain the vortex street signal frequency sequence and the fluid parameter sequence within a preset time window; Calculate the power spectrum of the vortex street signal frequency sequence by using Fourier transform, and extract the main frequency component; Calculate the Strouhal number under the current working condition based on the corresponding relationship between the main frequency component and the fluid parameter sequence; Dynamically adjust the flow threshold interval according to the change trend of the Strouhal number.

6. A monitoring device, characterized in that, The monitoring device includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the monitoring device to execute the method according to any one of claims 1-5.

7. A computer-readable storage medium, comprising instructions, characterized in that, When the instruction runs on the monitoring device, enable the monitoring device to execute the method according to any one of claims 1-5.

8. A computer program product, characterized in that, When the computer program product runs on the monitoring device, enable the monitoring device to execute the method according to any one of claims 1-5.

Images