Intelligent double-screw flowmeter data management system and method based on Internet of Things

Abstract

The invention relates to the technical field of double-screw flow analysis, and discloses an intelligent double-screw flowmeter data management system and method based on the Internet of Things. The system comprises a data acquisition module, a preprocessing module, an anomaly detection module, an adaptive learning module, a dynamic optimization module, a result feedback module and a fault tracing module. The data acquisition module acquires flow, pressure and temperature data in real time, and performs local caching and missing value processing; the preprocessing module performs wavelet transform and noise reduction on the data to generate standardized time series data; the anomaly detection module outputs an anomaly probability by using an isolated forest algorithm, and the adaptive learning module dynamically updates an anomaly judgment threshold; the dynamic optimization module performs multi-objective optimization on working parameters based on a genetic algorithm; the result feedback module feeds back the optimization instruction to the cloud platform; and the fault traceability module locates a fault source. According to the invention, data quality is improved, accurate anomaly detection and fault tracing are realized, working parameters are optimized, and industrial intelligent development is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of twin-screw flow analysis, and in particular to an intelligent twin-screw flowmeter data management system and method based on the Internet of Things. Background Art

[0002] In industrial production, twin-screw flowmeters, as key devices for accurately measuring fluid flow, are widely used in numerous fields, including petroleum, chemical, and food. Their measurement accuracy and operational stability directly impact production process control precision, product quality, and the company's economic benefits. However, traditional twin-screw flowmeters present numerous challenges in data management and operational control.

[0003] From a data acquisition perspective, traditional flow meters often only collect basic flow data and fail to capture parameters closely related to flow, such as pressure and temperature. Even when multi-parameter data is collected, data accuracy is often poor. For one thing, harsh industrial environments, such as complex electromagnetic interference and mechanical vibration, can easily introduce noise into sensor data, disrupting the true data signal. Furthermore, sensors themselves have limited precision and can age after long-term use, further reducing data acquisition accuracy.

[0004] The data processing and analysis links also have prominent problems. Traditional methods make it difficult to effectively process and deeply analyze large amounts of collected data. Due to the lack of advanced data preprocessing methods, noise data is not properly handled, which directly affects the reliability of subsequent analysis results. In terms of anomaly detection, it mainly relies on manual experience judgment or simple threshold setting, which cannot adapt to complex and changing industrial conditions. The parameter range for normal operation of the twin-screw flowmeter will change at different production stages. Fixed thresholds are difficult to accurately identify anomalies, and are prone to misjudgment or omission. Potential fault hazards cannot be discovered in time, affecting the continuity and safety of production.

[0005] Traditional twin-screw flowmeters rely on manual adjustment of operating parameters for optimal control, a method that is inefficient and difficult to achieve precise control. Manual adjustment often fails to balance multiple performance indicators, making it difficult to find the optimal balance between pressure fluctuations, flow stability, and temperature control. Furthermore, manual operation is susceptible to subjective factors, and different operators use different adjustment methods and standards, making it impossible to guarantee optimal results every time. This leads to significant energy waste and increased production costs.

[0006] The rapid development of emerging technologies such as the Internet of Things, big data, and artificial intelligence has provided new ideas and methods for solving the difficult problems of twin-screw flowmeter data management and operation control. However, the research and practice of effectively integrating these advanced technologies into the field of twin-screw flowmeters is still in the exploratory stage. The market urgently needs a twin-screw flowmeter system and method that can fully utilize the advantages of new technologies to achieve intelligent data management and efficient operation control to meet the growing high-precision and intelligent production needs of modern industry. Summary of the Invention

[0007] The purpose of the present invention is to provide an intelligent twin-screw flowmeter data management system and method based on the Internet of Things to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an intelligent twin-screw flowmeter data management system based on the Internet of Things, the system comprising:

[0009] A data acquisition module, which is used to collect operating data of the twin-screw flowmeter in real time through sensors, including flow data, pressure data and temperature data;

[0010] A data preprocessing module, configured to perform wavelet transform noise reduction processing on the operating data to generate standardized time series data;

[0011] An anomaly detection module, the anomaly detection module is used to input the standardized time series data into an isolation forest algorithm model and output a probability of abnormal operation of the flow meter;

[0012] An adaptive learning module, configured to dynamically update an abnormality determination threshold of the isolation forest algorithm model through a long short-term memory network based on historical operating data and real-time operating data;

[0013] A dynamic optimization module, configured to perform multi-objective optimization on the operating parameters of the twin-screw flowmeter based on a genetic algorithm to generate optimized pressure-flow control instructions;

[0014] A result feedback module is used to feed back the optimized pressure-flow control instruction to the cloud control platform through a RESTful API.

[0015] Preferably, the execution steps of the data preprocessing module include:

[0016] Dividing the flow data, pressure data and temperature data into multiple data segments according to time series;

[0017] Select the wavelet basis function for each data segment and calculate the wavelet coefficients of each scale;

[0018] The noise coefficient is filtered out by hard threshold method, and the normalized time series data after noise reduction is reconstructed;

[0019] The wavelet coefficient calculation satisfies:

[0020]

[0021] The hard threshold processing formula is:

[0022]

[0023] Where: W j,k is the wavelet coefficient of scale j and position k, ψ j,k (t) is the wavelet basis function, T is the noise threshold, is the coefficient after processing.

[0024] Preferably, the execution step of the anomaly detection module further includes:

[0025] The training process for building the isolation forest algorithm model includes:

[0026] Generating a sub-dataset by randomly sampling from the standardized time series data;

[0027] Generate multiple isolation trees through recursive partitioning, and the node splitting of each tree is based on randomly selected features and splitting thresholds;

[0028] The path length of each data point in the isolation tree is counted and the anomaly score is calculated.

[0029] Preferably, the execution steps of the adaptive learning module include:

[0030] Dividing the historical operation data into a training set and a validation set according to a time window;

[0031] Extracting time series features through the long short-term memory network to predict the abnormal probability distribution of the next time window;

[0032] According to the difference between the predicted distribution and the real-time abnormal probability, the determination threshold of the isolation forest algorithm model is dynamically adjusted.

[0033] Preferably, the execution steps of the dynamic optimization module include:

[0034] Define multi-objective functions, including minimizing the pressure fluctuation variance, maximizing the flow stability index, and minimizing the temperature deviation coefficient;

[0035] Iteratively generate a set of candidate parameters through crossover, mutation, and selection operations of the genetic algorithm;

[0036] Screening optimal pressure-flow control instructions based on Pareto front;

[0037] The multi-objective function expression is:

[0038]

[0039] Where: is the pressure fluctuation variance, η q is the flow stability index, δ T is the temperature deviation coefficient, and X is the set of decision variables.

[0040] Preferably, the execution steps of the data acquisition module further include:

[0041] Locally caching the operating data through edge computing nodes and enabling offline data synchronization protocols when the network is interrupted;

[0042] Perform timestamp alignment and missing value interpolation on the raw data collected by the sensor.

[0043] Preferably, the missing value interpolation adopts a dynamic filling method based on the K-nearest neighbor algorithm, including:

[0044] Calculate the Euclidean distance between missing data points and adjacent data points;

[0045] Select the K nearest neighbor data points and take weighted average to generate the interpolated value.

[0046] Preferably, the execution steps of the result feedback module further include:

[0047] Encrypting the optimized pressure-flow control instruction to generate a digital signature;

[0048] Push encryption instructions to designated terminal devices via the message queue telemetry transmission protocol.

[0049] Preferably, the system further comprises:

[0050] The fault tracing module is used to locate the fault root node based on the anomaly detection results through a pre-built causal reasoning graph model and generate a fault tracing report.

[0051] Preferably, the present invention also includes an intelligent twin-screw flowmeter data management method based on the Internet of Things, the method comprising:

[0052] Step 1: Using a data acquisition module to collect operating data of the twin-screw flowmeter in real time through a sensor, the operating data includes flow data, pressure data and temperature data;

[0053] Step 2: Use the data preprocessing module to perform wavelet transform noise reduction on the collected operating data to generate standardized time series data;

[0054] Step 3: Input the standardized time series data into the isolation forest algorithm model with the help of the anomaly detection module, and output the probability of flow meter operation abnormality;

[0055] Step 4: Use the adaptive learning module to dynamically update the abnormality judgment threshold of the isolation forest algorithm model through the long short-term memory network based on historical operation data and real-time operation data;

[0056] Step 5: Use the dynamic optimization module to perform multi-objective optimization on the working parameters of the twin-screw flowmeter based on the genetic algorithm to generate optimized pressure-flow control instructions;

[0057] Step 6: Feedback the optimized pressure-flow control instructions to the cloud control platform through the result feedback module using the RESTful API.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The data acquisition module of this invention not only collects flow data but also simultaneously acquires pressure and temperature data, providing rich information for comprehensive analysis of the flow meter's operating status. Local caching at the edge computing node and an offline data synchronization protocol ensure data integrity during network anomalies, preventing data loss. A dynamic filling method based on the K-nearest neighbor algorithm is used to handle missing values, along with timestamp alignment technology. This significantly improves data accuracy and consistency, laying a solid foundation for subsequent precise analysis and reliable decision-making.

[0060] The anomaly detection module utilizes an isolation forest algorithm model to effectively identify the probability of abnormal flow meter operation. Compared to traditional threshold judgment methods, it is more adaptable and can accurately respond to complex and changing industrial conditions, significantly reducing the rates of false positives and false negatives. The adaptive learning module leverages a long-short-term memory network to dynamically update anomaly judgment thresholds based on historical and real-time data, further improving the accuracy of anomaly detection. The fault tracing module, based on a causal reasoning graph model, can quickly locate the root cause of a fault when an anomaly occurs and generate a detailed fault tracing report, significantly shortening troubleshooting time, improving system maintenance efficiency, ensuring production continuity and stability, and reducing economic losses caused by failures.

[0061] The dynamic optimization module uses a genetic algorithm to perform multi-objective optimization of twin-screw flowmeter operating parameters, comprehensively considering metrics such as minimizing pressure fluctuation variance, maximizing the flow stability index, and minimizing the temperature deviation coefficient. This multi-objective optimization approach identifies the optimal operating parameter combination under different operating conditions, achieving precise control of pressure, flow, and temperature, and improving flowmeter performance. This not only helps improve control accuracy during the production process and ensure product quality, but also reduces energy consumption, saves production costs, and enhances the company's market competitiveness.

[0062] The result feedback module transmits optimized pressure-flow control instructions to the cloud control platform via a RESTful API. The instructions are encrypted and digitally signed, and then pushed to designated terminal devices using a message queue telemetry transmission protocol, ensuring the security, accuracy, and efficiency of instruction transmission. In complex IoT environments, this effectively prevents instruction theft and tampering, ensuring the reliability and stability of system operation.

[0063] This invention deeply integrates advanced technologies such as the Internet of Things, big data analysis, and artificial intelligence into a twin-screw flowmeter data management system, providing an innovative solution for industrial flow measurement and control. This will help advance industrial production toward intelligent and automated processes, improve production efficiency and management across the entire industrial sector, and meet the urgent need for high-precision, intelligent production in modern industry. This is of great significance for promoting industrial upgrading and sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a working principle diagram of the intelligent twin-screw flowmeter data management system of the present invention;

[0065] Figure 2 This is the workflow diagram of the data preprocessing module;

[0066] Figure 3 Workflow diagram for training the isolation forest algorithm model;

[0067] Figure 4 This is the workflow diagram of the dynamic optimization module. DETAILED DESCRIPTION

[0068] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0069] See also Figure 1-4 The present invention relates to an intelligent twin-screw flowmeter data management system and method based on the Internet of Things. The technical solution of the present invention will be described in detail below in conjunction with specific implementation methods.

[0070] The data acquisition module uses sensors to collect real-time operating data from the twin-screw flowmeter, including flow, pressure, and temperature. In practical applications, sensors are installed at key locations within the twin-screw flowmeter to accurately capture this operating data. These sensors possess high-precision measurement capabilities. For example, the flow sensor utilizes electromagnetic induction to quickly and accurately measure changes in fluid flow. The pressure sensor, based on the piezoelectric effect, monitors pressure within the pipeline in real time. The temperature sensor utilizes a thermistor, achieving temperature measurement accuracy that meets industrial production requirements.

[0071] The data preprocessing module performs wavelet transform noise reduction on the collected operational data to generate standardized time series data. This module's processing is crucial for ensuring the accuracy of subsequent data processing, effectively removing noise interference from the data and making it more regular and usable.

[0072] The anomaly detection module feeds standardized time series data into the Isolation Forest algorithm model, which outputs the probability of abnormal flow meter operation. By learning from normal data, the Isolation Forest algorithm model can identify data that deviates significantly from normal patterns, thereby determining whether the flow meter is operating abnormally.

[0073] The adaptive learning module dynamically updates the anomaly threshold of the isolation forest algorithm model using a long-short-term memory network based on historical and real-time operating data. The long-short-term memory network learns the time series characteristics of the data and continuously adjusts the anomaly threshold over time and with the accumulation of data to adapt to anomaly detection needs under different operating conditions.

[0074] The dynamic optimization module uses a genetic algorithm to perform multi-objective optimization of the twin-screw flowmeter's operating parameters and generate optimized pressure-flow control instructions. The genetic algorithm simulates the natural evolutionary process, using operations such as crossover, mutation, and selection to find the optimal solution among numerous possible parameter combinations to optimize the twin-screw flowmeter's operating parameters.

[0075] The result feedback module transmits optimized pressure-flow control instructions to the cloud-based control platform via a RESTful API. The cloud-based control platform can further process and analyze these instructions, enabling remote monitoring and control of the twin-screw flowmeter. A RESTful API is an application programming interface (API) based on the REST (Representational State Transfer) architectural style.

[0076] In addition, the system can also include a fault tracing module, which uses a causal reasoning graph model to locate the root cause of the fault based on the anomaly detection results and generate a fault tracing report. This helps to quickly and accurately identify the problem after an anomaly is discovered, improving system maintenance efficiency and stability.

[0077] The technical solution of the present invention is further described in detail below with reference to specific embodiments.

[0078] Example 1:

[0079] This embodiment mainly describes the specific execution steps of the data preprocessing module. The data preprocessing module plays a vital role in the entire data management system. It can clean and organize the original collected data and provide a high-quality data foundation for subsequent data analysis and processing.

[0080] The data preprocessing module first divides flow, pressure, and temperature data into multiple data segments based on time series. This facilitates segmented data processing and improves processing efficiency. In practice, the length of each data segment can be appropriately set based on the data acquisition frequency and data volume. For example, if the data acquisition frequency is 10 times per second, the data can be segmented into 100 data points, or 10 seconds per segment.

[0081] Next, a wavelet basis function is selected for each data segment, and the wavelet coefficients of each scale are calculated. The choice of wavelet basis function will affect the decomposition effect of the data. Different wavelet basis functions are suitable for different types of data. In this embodiment, based on the characteristics of the twin-screw flowmeter operating data, after multiple tests and comparisons, the Daubechies wavelet basis function with good localization characteristics was selected. When calculating the wavelet coefficients, the formula is used:

[0082]

[0083] Among them, W j,k is the wavelet coefficient of scale j and position k, ψ j,k (t) is the wavelet basis function. Through this formula, wavelet transform is performed on each data segment to obtain wavelet coefficients at different scales.

[0084] Finally, the noise coefficient is filtered out by the hard threshold method, and the normalized time series data after noise reduction is reconstructed. The processing formula of the hard threshold method is:

[0085]

[0086] Where T is the noise threshold, is the processed coefficient. In practical applications, determining the noise threshold T is crucial. It can typically be set based on the data's noise level and empirical values, or it can be adjusted using adaptive methods. For example, through statistical analysis of large amounts of historical data, an appropriate noise threshold range can be determined. Then, during actual processing, the value of T can be dynamically adjusted within this range based on the characteristics of the current data segment. After hard thresholding, the noise coefficient is removed. A reconstruction algorithm is then used to reconstruct the processed wavelet coefficients into de-noised, standardized time series data, completing the entire data preprocessing process.

[0087] Example 2:

[0088] This embodiment focuses on the execution steps of the anomaly detection module and the adaptive learning module. The anomaly detection module can promptly detect abnormal operating conditions of the twin-screw flowmeter, while the adaptive learning module can continuously optimize the anomaly detection criteria to improve detection accuracy and adaptability.

[0089] The training process for the Isolation Forest algorithm model in the anomaly detection module involves randomly sampling from the standardized time series data to generate sub-datasets. In practice, to ensure the representativeness and diversity of the sub-datasets, a random sampling method can be used. For example, a certain percentage of data from the entire standardized time series data set can be randomly selected as the sub-dataset. This percentage can be adjusted based on the data volume and computing resources, and is generally set to 30%-50%.

[0090] Multiple isolation trees are generated through recursive partitioning. Each tree node is split based on randomly selected features and a splitting threshold. During the partitioning process, a feature dimension of the data is randomly selected, and then a splitting threshold is randomly selected on that dimension to partition the data into two child nodes, the left and right. This process is repeated until the amount of data at each node falls below the minimum node data size or the tree depth reaches a preset value. For example, if the minimum node data size is set to 5 and the tree depth is set to 10, the tree growth stops when the data size of a node falls below 5 or the tree depth reaches 10.

[0091] The path length of each data point in the isolation tree is counted to calculate the anomaly score. The higher the anomaly score, the more likely the data point is an anomaly. By calculating the anomaly scores of all data points, the probability of flow meter operation anomaly can be determined.

[0092] The adaptive learning module executes by dividing historical data into training and validation sets based on time windows. The time window size can be set based on data variation patterns and actual needs, for example, 1 hour or 2 hours. The purpose of dividing the training and validation sets is to evaluate model performance during training and prevent overfitting.

[0093] Long Short-Term Memory (LSTM) networks extract time series features and predict the probability distribution of anomalies in the next time window. Long Short-Term Memory (LSTM) networks have the ability to memorize long-term information and effectively learn the time series characteristics of data. When training the LSTM model, the training set data is used and the model parameters are adjusted to accurately predict the probability distribution of anomalies in the next time window.

[0094] The Isolation Forest algorithm model's decision threshold is dynamically adjusted based on the discrepancy between the predicted distribution and the real-time anomaly probability. A significant discrepancy between the predicted distribution and the real-time anomaly probability indicates that the current anomaly decision threshold may need adjustment. For example, if the predicted anomaly probability distribution indicates a low anomaly probability for a certain time period, but the actual real-time anomaly probability is higher, the Isolation Forest algorithm model's decision threshold should be appropriately lowered to more accurately detect anomalies. By continuously and dynamically adjusting the decision threshold, the Isolation Forest algorithm model can better adapt to anomaly detection needs under different operating conditions.

[0095] Example 3:

[0096] The dynamic optimization module can improve the operating performance of the twin-screw flowmeter, reduce energy consumption and improve production efficiency through multi-objective optimization of the working parameters of the twin-screw flowmeter.

[0097] The dynamic optimization module first defines a multi-objective function, including minimizing the pressure fluctuation variance, maximizing the flow stability index, and minimizing the temperature deviation coefficient. The multi-objective function expression is:

[0098]

[0099] Where, is the pressure fluctuation variance, η q is the flow stability index, δ T is the temperature deviation coefficient, and X is the set of decision variables. The pressure fluctuation variance reflects the degree of pressure fluctuation; the smaller the variance, the more stable the pressure; the larger the flow stability index, the better the flow stability; and the smaller the temperature deviation coefficient, the smaller the deviation between the actual temperature and the ideal temperature.

[0100] Candidate parameter sets are iteratively generated through the genetic algorithm's crossover, mutation, and selection operations. A genetic algorithm is an optimization algorithm that simulates the natural evolutionary process. In the crossover operation, two parent individuals are randomly selected and their genes are partially exchanged according to a certain crossover probability to generate new offspring individuals. The mutation operation randomly alters the genes of individuals with a certain mutation probability, increasing the diversity of the population. The selection operation selects individuals with higher fitness to advance to the next generation based on their fitness values, thus evolving the population towards a more optimal direction. In this embodiment, the crossover probability can be set to 0.8 and the mutation probability to 0.01. Through multiple iterations, a large number of candidate parameter sets are generated.

[0101] Optimal pressure-flow control instructions are screened based on the Pareto front. The Pareto front is the set of solutions in a multi-objective optimization problem where one objective cannot be further improved by sacrificing other objectives. From the numerous candidate parameter sets generated, solutions on the Pareto front are selected. These solutions are the optimal pressure-flow control instructions that achieve a balance between multiple objectives. This approach allows for the development of optimized solutions that meet diverse requirements, allowing users to select the most appropriate control instructions based on their specific needs.

[0102] Example 4:

[0103] The data acquisition module uses edge computing nodes to locally cache operational data and activates an offline data synchronization protocol in the event of a network interruption. Edge computing nodes have local storage capabilities and can cache data promptly after collection. For example, embedded storage devices such as SD cards or solid-state drives are used to store collected flow, pressure, and temperature data locally. When the network is normal, the data is uploaded to the cloud server at a set frequency. When the network is interrupted, the offline data synchronization protocol is activated, and the edge computing node temporarily stores the locally cached data. Once the network is restored, the data is uploaded to the cloud according to a specific strategy to ensure that the data is not lost.

[0104] Perform timestamp alignment and missing value interpolation on the raw data collected by sensors. Timestamp alignment ensures temporal consistency of data collected by different sensors, facilitating subsequent data analysis and processing. In practical applications, each sensor records a timestamp when collecting data. By calibrating and synchronizing timestamps, all data is aligned on the timeline.

[0105] Missing value interpolation uses a dynamic filling method based on the K-nearest neighbor algorithm. First, the Euclidean distance between the missing data point and its neighboring data points is calculated. Euclidean distance measures the similarity between data points; closer distances indicate greater similarity. The K nearest neighboring data points are then selected and weighted averaged to generate the interpolated value. The K value should be adjusted based on the characteristics and distribution of the data. For example, for data with a relatively uniform distribution, a moderate K value, such as 5-10, can be selected. For data with a relatively sparse distribution, the K value can be increased to ensure sufficient neighboring data points are obtained. When performing the weighted average, closer neighboring data points receive higher weights, which more accurately reflects the changing trends of the data and results in more reasonable interpolated values.

[0106] Example 5:

[0107] The result feedback module encrypts the optimized pressure-flow control instructions and generates a digital signature. This encryption uses a secure algorithm, such as the Advanced Encryption Standard (AES), to protect the instructions from being stolen or tampered with during transmission. The digital signature is generated to ensure the integrity and authenticity of the instructions. The instructions are signed with a private key, and the recipient can verify them using the corresponding public key.

[0108] Encrypted commands are pushed to designated end devices via the Message Queuing Telemetry Transport (MQTT) protocol. The MQTT protocol is lightweight, low-power, and supports multiple network environments, making it suitable for data transmission in IoT environments. In practical applications, an MQTT server is set up, and the result feedback module sends the encrypted commands to the MQTT server. The server then pushes the commands to designated end devices, such as the controller of a twin-screw flowmeter or related devices on a cloud-based control platform.

[0109] Based on the anomaly detection results, the fault tracing module locates the root cause of the fault using a causal reasoning graph model and generates a fault tracing report. The causal reasoning graph model is a causal relationship-based model that analyzes the causal relationship between abnormal data and various system components to identify the root cause of the anomaly. When building a causal reasoning graph model, it is necessary to collect a large amount of historical fault data and system operation data to analyze the relationship between different faults and various factors. When the anomaly detection module detects an anomaly, the fault tracing module uses the causal reasoning graph model to perform reasoning, locate the root cause of the fault, and generate a detailed fault tracing report. The report includes information such as the time of the fault, possible causes, and the scope of impact, providing an important basis for system maintenance and repair.

[0110] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0111] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent twin-screw flowmeter data management system based on the Internet of Things, characterized in that: include: A data acquisition module, which is used to collect operating data of the twin-screw flowmeter in real time through sensors, including flow data, pressure data and temperature data; A data preprocessing module, configured to perform wavelet transform noise reduction processing on the operating data to generate standardized time series data; An anomaly detection module, the anomaly detection module is used to input the standardized time series data into an isolation forest algorithm model and output a probability of abnormal operation of the flow meter; An adaptive learning module, configured to dynamically update an abnormality determination threshold of the isolation forest algorithm model through a long short-term memory network based on historical operating data and real-time operating data; A dynamic optimization module, configured to perform multi-objective optimization on the operating parameters of the twin-screw flowmeter based on a genetic algorithm to generate optimized pressure-flow control instructions; A result feedback module is used to feed back the optimized pressure-flow control instruction to the cloud control platform through a RESTful API.

2. The system according to claim 1, wherein The execution steps of the data preprocessing module include: Dividing the flow data, pressure data and temperature data into multiple data segments according to time series; Select the wavelet basis function for each data segment and calculate the wavelet coefficients of each scale; The noise coefficient is filtered out by hard threshold method, and the normalized time series data after noise reduction is reconstructed; The wavelet coefficient calculation satisfies: The hard threshold processing formula is: Where: W j,k is the wavelet coefficient of scale j and position k, ψ j,k (t) is the wavelet basis function, T is the noise threshold, is the coefficient after processing.

3. The system according to claim 2, wherein: The execution steps of the anomaly detection module also include: The training process for building the isolation forest algorithm model includes: Generating a sub-dataset by randomly sampling from the standardized time series data; Generate multiple isolation trees through recursive partitioning, and the node splitting of each tree is based on randomly selected features and splitting thresholds; The path length of each data point in the isolation tree is counted and the anomaly score is calculated.

4. The system according to claim 2, wherein: The execution steps of the adaptive learning module include: Dividing the historical operation data into a training set and a validation set according to a time window; Extracting time series features through the long short-term memory network to predict the abnormal probability distribution of the next time window; According to the difference between the predicted distribution and the real-time abnormal probability, the determination threshold of the isolation forest algorithm model is dynamically adjusted.

5. The system according to claim 2, wherein: The execution steps of the dynamic optimization module include: Define multi-objective functions, including minimizing the pressure fluctuation variance, maximizing the flow stability index, and minimizing the temperature deviation coefficient; Iteratively generate a set of candidate parameters through crossover, mutation, and selection operations of the genetic algorithm; Screening optimal pressure-flow control instructions based on Pareto front; The multi-objective function expression is: Where: is the pressure fluctuation variance, η q is the flow stability index, δ T is the temperature deviation coefficient, and X is the set of decision variables.

6. The system according to claim 2, wherein: The execution steps of the data acquisition module also include: Locally caching the operating data through edge computing nodes and enabling offline data synchronization protocols when the network is interrupted; Perform timestamp alignment and missing value interpolation on the raw data collected by the sensor.

7. The system according to claim 6, wherein: The missing value interpolation adopts a dynamic filling method based on the K-nearest neighbor algorithm, including: Calculate the Euclidean distance between missing data points and adjacent data points; Select the K nearest neighbor data points and take weighted average to generate the interpolated value.

8. The system according to claim 1, wherein: The execution steps of the result feedback module also include: Encrypting the optimized pressure-flow control instruction to generate a digital signature; Push encryption instructions to designated terminal devices via the message queue telemetry transmission protocol.

9. The system according to claim 1, wherein: The system further comprises: The fault tracing module is used to locate the fault root node based on the anomaly detection results through a pre-built causal reasoning graph model and generate a fault tracing report.

10. An intelligent twin-screw flowmeter data management method based on the Internet of Things, characterized in that: The method comprises: Step 1: Using a data acquisition module to collect operating data of the twin-screw flowmeter in real time through a sensor, the operating data includes flow data, pressure data and temperature data; Step 2: Use the data preprocessing module to perform wavelet transform noise reduction on the collected operating data to generate standardized time series data; Step 3: Input the standardized time series data into the isolation forest algorithm model with the help of the anomaly detection module, and output the probability of flow meter operation abnormality; Step 4: Use the adaptive learning module to dynamically update the abnormality judgment threshold of the isolation forest algorithm model through the long short-term memory network based on historical operation data and real-time operation data; Step 5: Use the dynamic optimization module to perform multi-objective optimization on the working parameters of the twin-screw flowmeter based on the genetic algorithm to generate optimized pressure-flow control instructions; Step 6: Feedback the optimized pressure-flow control instructions to the cloud control platform through the result feedback module using the RESTful API.

Images