Reverse flow valve supporting IoT remote monitoring

Abstract

The present invention provides a reverse flow valve supporting remote monitoring of the Internet of Things, comprising a valve body with a fluid inlet, an outlet and an internal flow channel, and an electric actuator is configured to control the valve opening; the control unit has a built-in microprocessor and a memory, and processes sensor data to calculate a fluid density correction coefficient and a pressure correction coefficient, corrects the original flow data collected by the flow sensor, and generates an actual flow value; the control unit further identifies the current fluid type based on the relationship between the flow and pressure data and adaptively adjusts the reverse flow detection threshold of the upstream and downstream pressure difference; when the pressure difference exceeds the adjusted threshold, the electric actuator performs an emergency shutdown operation; at the same time, it is equipped with a wireless communication unit to support real-time data transmission and receive control parameters issued by external monitoring equipment, so as to realize intelligent control and remote management; the present invention can effectively improve the intelligence and operation efficiency of the reverse flow valve, and is suitable for safety protection and operation optimization under complex fluid working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of reverse flow valves, and in particular to a reverse flow valve supporting remote monitoring of the Internet of Things. Background Art

[0002] The reverse flow valve is an important device widely used in fluid delivery systems. It is mainly used to prevent the backflow of fluid in the pipeline system, thereby protecting the safety of pipelines and related equipment. Traditional reverse flow valves usually rely on mechanical structures to achieve basic reverse flow protection functions. Some advanced reverse flow valves have added pressure and flow sensors to monitor the operating status. These devices achieve a certain degree of automation through wired or local control. However, with the rise of Internet of Things technology, the demand for remote monitoring and intelligent control in more and more fields has gradually increased, making the traditional reverse flow valve gradually show its technical limitations.

[0003] In the prior art, although some reverse flow valves are equipped with sensors or basic control units, these systems are difficult to achieve accurate reverse flow detection and valve control under complex working conditions. Specifically, existing equipment usually relies on a fixed pressure difference threshold to detect reverse flow, and lacks the ability to dynamically identify fluid characteristics and adjust thresholds. In addition, the existing reverse flow valves have limited interaction capabilities with external monitoring equipment, and the efficiency of data transmission and control parameter updates is low, making it difficult to meet the needs of modern industry for remote management and intelligent optimization.

[0004] Based on the above situation, it is necessary to develop a reverse flow valve that supports remote monitoring of the Internet of Things. Summary of the invention

[0005] The present application provides a reverse flow valve that supports remote monitoring of the Internet of Things to improve the operating efficiency of the reverse flow valve.

[0006] The present application provides a reverse flow valve supporting remote monitoring of the Internet of Things, comprising:

[0007] a valve body having a fluid inlet, a fluid outlet, and an internal flow passage;

[0008] An electric actuator, disposed on the valve body, for controlling the valve opening;

[0009] A multimodal sensing unit, electrically connected to the control unit, comprising an inlet pressure sensor, an outlet pressure sensor, a flow sensor and a temperature sensor, for real-time acquisition of valve operation parameters; wherein the inlet pressure sensor is arranged at the fluid inlet of the valve body, for measuring the fluid inlet pressure; the outlet pressure sensor is arranged at the fluid outlet of the valve body, for measuring the fluid outlet pressure; the flow sensor is arranged in the internal flow channel of the valve body, for measuring the fluid flow; the temperature sensor is arranged in the internal flow channel of the valve body, for measuring the fluid temperature;

[0010] A control unit is arranged on the valve body, the control unit includes a microprocessor and a memory, and is used to receive and process the pressure, flow and temperature data collected by the multimodal sensing unit; calculate the fluid density correction coefficient according to the fluid temperature data collected by the temperature sensor; calculate the pressure correction coefficient according to the pressure data collected by the inlet pressure sensor and the outlet pressure sensor; correct the original flow data collected by the flow sensor according to the fluid density correction coefficient and the pressure correction coefficient to obtain a corrected actual flow value; identify the current fluid type based on the relationship between the corrected actual flow value and the pressure data; adaptively adjust the backflow detection threshold according to the identified fluid type, wherein the backflow detection threshold is a threshold of the upstream and downstream pressure difference; monitor the pressure difference between the inlet pressure and the outlet pressure, and when the pressure difference exceeds the adjusted backflow detection threshold, control the electric actuator to perform an emergency shutdown operation;

[0011] A wireless communication unit is arranged at the top of the valve body. The wireless communication unit is electrically connected to the control unit and supports wireless protocol communication. It is used to transmit real-time data of the control unit to an external monitoring device and receive control parameters issued by the external monitoring device. The control parameters include a pressure difference threshold for backflow detection, a valve response time, a valve closing speed, and an opening limit value.

[0012] This application has the following beneficial technical effects:

[0013] (1) The present invention can accurately correct the flow data by collecting pressure, flow and temperature data through a multimodal sensing unit, combined with the fluid density and pressure correction coefficient calculated by the control unit, and dynamically identify the current fluid type according to the relationship between the flow and pressure data, and adaptively adjust the pressure difference threshold of the backflow detection, thereby improving the accuracy of the backflow detection. (2) When the difference between the inlet pressure and the outlet pressure is detected to exceed the adjusted backflow detection threshold, the control unit can automatically control the electric actuator to perform an emergency shutdown operation to quickly prevent the backflow from occurring, significantly improving the safety and response efficiency of the system, and is particularly suitable for emergency processing needs under complex working conditions. (3) Through the interaction between the wireless communication unit and the external monitoring device, the present invention can realize real-time data transmission and receive control parameters sent from the outside, such as the pressure difference threshold of the backflow detection and the valve closing speed. In addition, the external device can also optimize the control strategy based on historical data to improve the operating efficiency and management flexibility of the system. (4) By identifying different fluid types and adjusting the threshold and control logic, the present application can adapt to a variety of fluid characteristics and complex working conditions, avoiding the misjudgment or control failure that may be caused by the traditional backflow valve relying on fixed parameters, and significantly improving the applicability and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a schematic diagram of a reverse flow valve supporting IoT remote monitoring provided in the first embodiment of the present application. DETAILED DESCRIPTION

[0015] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0016] The first embodiment of the present application provides a reverse flow valve that supports remote monitoring of the Internet of Things. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 The first embodiment of the present application provides a reverse flow valve that supports remote monitoring of the Internet of Things.

[0017] The reverse flow valve supporting remote monitoring via the Internet of Things includes a valve body 101 , an electric actuator 102 , a multimodal sensing unit 103 , a control unit 104 and a wireless communication unit 105 .

[0018] A valve body 101 is provided with a fluid inlet, a fluid outlet and an internal flow passage.

[0019] The valve body 101 is the core component of the reverse flow valve provided in this embodiment, and its design and structure directly affect the functional realization and performance of the device. The valve body 101 has a fluid inlet, a fluid outlet, and an internal flow channel running through the two, which is used to accommodate and guide the flow of fluid. The fluid inlet and outlet are respectively located at both ends of the valve body, and are usually designed in the form of pipe connectors, which can be easily connected to the industrial pipeline system to ensure that the fluid can smoothly enter and discharge the valve body. The internal flow channel is located inside the valve body, and its cross-sectional shape and size are precisely designed to minimize the resistance during the fluid flow process, while ensuring the stability and measurability of the fluid parameters.

[0020] The valve body 101 is usually made of high-strength corrosion-resistant materials, such as stainless steel, brass or high-performance polymers, which can adapt to the flow environment of various media (such as water, oil, gas, etc.) in the industry, and has the characteristics of high pressure and high temperature resistance. In order to ensure that the multimodal sensing unit 103 can accurately collect fluid operating parameters, the valve body 101 is provided with sensor installation interfaces at specific positions of the internal flow channel, such as the pressure sensor interfaces at the inlet and outlet are located on the inner walls of the fluid inlet and outlet, and the interfaces of the flow sensor and the temperature sensor are located in the center of the flow channel. The shapes and sizes of these interfaces are standardized, which can realize the rapid installation and sealing connection of the sensors to prevent fluid leakage.

[0021] The outer surface of the valve body 101 is designed to be a flat structure to facilitate the installation of other functional units, such as the electric actuator 102 and the control unit 104. The electric actuator is installed on the top or side of the valve body through a fixing bracket, and the control unit is fixedly connected to the outer wall of the valve body by bolts or welding. A special interface for installing the wireless communication unit 105 is also provided on the top of the valve body to facilitate the electrical connection between the wireless communication unit and the control unit.

[0022] In addition, the valve body 101 is designed with fluid spoilers or guide grooves in the internal flow channel. These structures can optimize the fluid flow path and reduce fluid turbulence, thereby improving the measurement accuracy of the sensor and the control sensitivity of the device. The overall structural design of the valve body also needs to consider the installation environment and pipeline layout. For example, in industrial scenarios, a variety of valve body sizes can be provided according to actual needs to meet the connection requirements of different pipe diameters.

[0023] In summary, the valve body 101 lays a solid foundation for accurate fluid monitoring and intelligent backflow control through reasonable material selection, precise structural design and reliable integration with other components.

[0024] The electric actuator 102 is disposed on the valve body and is used to control the valve opening.

[0025] The electric actuator 102 is an important component for realizing valve opening control. Its function is to adjust the valve opening according to the command of the control unit 104, so as to realize precise control of fluid flow, including opening adjustment during normal operation and rapid shutdown in emergency. The electric actuator generally includes a motor, a reducer, a valve stem, a transmission mechanism and a limit device, and is firmly connected to the valve body 101 through a fixed structure.

[0026] The motor in the actuator can be a DC motor, an AC motor or a stepper motor, which can provide sufficient torque to overcome the influence of fluid pressure according to the actual working conditions. The motor is connected to the control unit through an electrical interface to receive signals for opening adjustment or shut-off operation. The motor output shaft is connected to the valve stem through a reducer. The function of the reducer is to reduce the speed and increase the torque to ensure that the valve stem can drive the valve smoothly and accurately.

[0027] The valve stem is connected to the valve core inside the valve body, and changes the valve opening through rotation or linear motion, thereby adjusting the amount of fluid passing through. The transmission mechanism usually adopts gear transmission, worm gear transmission or other efficient transmission methods to achieve reliable transmission of the motor power to the valve action. In order to ensure the accuracy of control, an opening feedback device, such as a potentiometer or photoelectric encoder, is set in the transmission mechanism to feed back the real-time opening information to the control unit to ensure that the valve action is consistent with the instruction.

[0028] The limit device is used to limit the maximum and minimum opening of the valve to prevent the valve from being over-opened or over-closed due to misoperation or system abnormality. The limit device usually consists of a mechanical limiter and an electrical limit switch. The mechanical limiter is responsible for physical limitation, and the electrical limit switch is responsible for transmitting the limit signal to the control unit to stop the motor action.

[0029] Electric actuators also include protection devices, such as overload protection and emergency manual operation devices. When the motor load exceeds the rated value or the power supply is abnormal, the overload protection device can cut off the power supply to prevent damage to the actuator; the emergency manual operation device allows the valve opening to be adjusted manually when the electrical system fails, ensuring the safety and operability of the system.

[0030] In terms of installation, the electric actuator is fixed to the top or side of the valve body by bolts or other mechanical connections. The choice of installation position must ensure that the actuator and the valve stem are aligned to avoid transmission loss or wear due to eccentricity. The actuator housing adopts a dustproof and waterproof design, which usually meets industrial standards to adapt to various complex working environments.

[0031] The electric actuator is connected to the control unit through a standardized electrical interface, which supports signal transmission and power supply. The signal transmission part usually uses analog signals (such as 4-20mA) or digital communication protocols (such as Modbus or CAN) to ensure high-precision and low-latency control. The power supply of the actuator is selected from DC or AC power supply according to the motor type, and the common supply voltage is 24V DC or 220V AC.

[0032] In summary, the electric actuator 102 achieves high-precision adjustment of the valve opening and emergency shut-off operation through precise mechanical and electrical structure design, providing reliable guarantee for the intelligence and safety of the reverse flow valve.

[0033] The multimodal sensing unit 103 is electrically connected to the control unit, and includes an inlet pressure sensor, an outlet pressure sensor, a flow sensor and a temperature sensor, and is used to collect valve operating parameters in real time; wherein the inlet pressure sensor is arranged at the fluid inlet of the valve body to measure the fluid inlet pressure; the outlet pressure sensor is arranged at the fluid outlet of the valve body to measure the fluid outlet pressure; the flow sensor is arranged in the internal flow channel of the valve body to measure the fluid flow rate; the temperature sensor is arranged in the internal flow channel of the valve body to measure the fluid temperature.

[0034] The multimodal sensing unit 103 is a key component for real-time acquisition of valve operating parameters, which includes an inlet pressure sensor, an outlet pressure sensor, a flow sensor, and a temperature sensor. These sensors are installed at different positions of the valve body 101 in a reasonable layout, which can comprehensively monitor the operating status of the fluid and provide accurate input data for the control unit 104, thereby achieving accurate judgment of the backflow risk and intelligent control of the valve.

[0035] The inlet pressure sensor is installed at the fluid inlet of the valve body. Its sensing element is in direct contact with the fluid and is used to measure the pressure value of the inlet fluid in real time. This sensor is usually made of high-pressure resistant and corrosion-resistant materials, such as ceramic or stainless steel diaphragms, to ensure its long-term stability under complex working conditions. The output signal of the sensor is transmitted to the control unit through a cable or integrated circuit, and is used in the control unit to calculate the pressure difference and correct the flow data.

[0036] The outlet pressure sensor has similar functions to the inlet pressure sensor, but is installed at the fluid outlet of the valve body to measure the pressure value of the outlet fluid. The two pressure sensors work together to provide pressure difference data of the fluid before and after the valve, providing a basis for the dynamic adjustment of the backflow detection threshold. The installation position should be selected to ensure that the sensor can directly contact the fluid and avoid measurement errors caused by eddy currents or fluid turbulence.

[0037] The flow sensor is set in the middle of the flow channel inside the valve body to measure the flow value of the fluid in real time. Its principle can be a turbine flowmeter, electromagnetic flowmeter or ultrasonic flowmeter. The specific selection needs to be optimized according to the fluid medium characteristics (such as conductivity, temperature range, flow rate, etc.). The flow sensor provides raw flow data through its high-precision measurement capability, and combines it with the output data of the pressure and temperature sensors, and finally corrects it to the actual flow value by the control unit.

[0038] The temperature sensor is installed in the flow channel near the flow sensor to monitor the temperature of the fluid in real time. This is because temperature changes can significantly affect the density and flow characteristics of the fluid, so temperature data is a key input for calculating the fluid density correction factor. The temperature sensor can be a thermocouple, thermistor or semiconductor temperature sensor, and the appropriate type should be selected according to the actual temperature range of the application.

[0039] Each sensor in the multimodal sensing unit is integrated with the valve body through a standardized interface. Its installation method must ensure effective contact with the fluid and avoid measurement deviations caused by leakage or installation errors. In addition, in order to improve anti-interference ability, the signal output of each sensor is usually processed by shielded cable or filtering to ensure the accuracy of the data transmitted to the control unit.

[0040] In summary, the multimodal sensing unit 103 can comprehensively and accurately obtain the core parameters of the fluid operation state through reasonable sensor layout and high-performance measurement capabilities, providing a solid data foundation for the intelligent control of the reverse flow valve.

[0041] A control unit 104 is arranged on the valve body, and the control unit includes a microprocessor and a memory, and is used to receive and process the pressure, flow and temperature data collected by the multimodal sensing unit; calculate the fluid density correction coefficient according to the fluid temperature data collected by the temperature sensor; calculate the pressure correction coefficient according to the pressure data collected by the inlet pressure sensor and the outlet pressure sensor; correct the original flow data collected by the flow sensor according to the fluid density correction coefficient and the pressure correction coefficient to obtain a corrected actual flow value; identify the current fluid type based on the relationship between the corrected actual flow value and the pressure data; adaptively adjust the backflow detection threshold according to the identified fluid type, wherein the backflow detection threshold is a threshold of the upstream and downstream pressure difference; monitor the pressure difference between the inlet pressure and the outlet pressure, and when the pressure difference exceeds the adjusted backflow detection threshold, control the electric actuator to perform an emergency shutdown operation.

[0042] The control unit 104 is the core component for realizing the intelligent control of the reverse flow valve. It is responsible for processing the pressure, flow and temperature data collected by the multimodal sensing unit 103, performing real-time analysis and calculation of the data, generating control signals, and coordinating with other functional units. The control unit is installed on the outside of the valve body 101 and is integrated with the valve body through a firm mechanical fixation. It includes a microprocessor and a memory, as well as necessary interface circuits, for realizing the reception, processing, storage and output of data.

[0043] The microprocessor in the control unit is responsible for performing core logic operations, including signal processing, data calculation, judgment analysis and control output. After receiving the signal from the multimodal sensing unit, the microprocessor first performs preliminary processing on the sensor data, such as denoising, filtering and linearization of the pressure, flow and temperature signals to ensure the accuracy and consistency of the data. The processed data is used in a series of calculation steps, including calculating the density correction factor of the fluid based on the temperature data, and calculating the pressure correction factor based on the inlet and outlet pressure data. These correction coefficients compensate for the influence of measurement errors under actual working conditions through mathematical models, providing higher accuracy for subsequent flow correction and judgment.

[0044] When calculating the density correction factor, the control unit uses the fluid property table or empirical formula pre-stored in the memory, combined with the data from the temperature sensor for real-time interpolation or calculation. Similarly, the pressure correction factor is calculated based on the input of the pressure sensor through a mathematical model. Then, these two correction factors are combined with the raw data of the flow sensor to perform complex calculations to generate the corrected actual flow value. This actual flow value is crucial in fluid dynamic monitoring and can accurately reflect the true state of the fluid.

[0045] By analyzing the relationship between the corrected actual flow value and the pressure data, the control unit can identify the type of the current fluid. For example, for different media such as liquid and gas, the control unit uses different algorithms or empirical models to make judgments.

[0046] The identified fluid type is further used to adaptively adjust the threshold parameters of backflow detection. Specifically, the upstream and downstream pressure difference threshold is dynamically optimized according to the fluid characteristics to ensure accurate detection of backflow risks. The control unit monitors the pressure difference between the inlet and outlet in real time. When the pressure difference exceeds the adjusted backflow detection threshold, an emergency shutdown signal is immediately generated and transmitted to the electric actuator 102 through the electrical interface to perform an emergency shutdown operation.

[0047] In addition, the control unit is connected to the wireless communication unit 105 through a built-in communication interface to realize the transmission and reception of real-time data. The wireless communication unit sends the data of the control unit to the external monitoring device, and the external device performs in-depth analysis and optimizes the control strategy based on the received data. The control unit can receive control parameters sent by the external device, such as the pressure difference threshold of backflow detection, valve response time, closing speed and opening limit value, and store these parameters in the memory to participate in the subsequent control logic.

[0048] The function of the memory is not only to store real-time control parameters, but also to save historical operation data and model parameters. These data can support the control unit to perform long-term trend analysis and improve the accuracy and adaptability of the control logic. The hardware design of the control unit fully considers the requirements of the industrial environment, and has dustproof, waterproof and anti-electromagnetic interference capabilities. Its shell design meets industrial standards and can operate stably in complex environments such as high temperature, high humidity, and strong vibration.

[0049] In summary, the control unit 104 realizes intelligent and precise control of the reverse flow valve through powerful data processing capabilities and flexible control logic.

[0050] Furthermore, the control unit is used to identify the current fluid type based on the relationship between the corrected actual flow value and the pressure data, specifically including:

[0051] Receive corrected actual flow value and upstream and downstream pressure data;

[0052] Perform multi-point matching analysis on the corrected actual flow value and the corresponding pressure data, and compare these data with characteristic curves of various fluid types pre-stored in the memory;

[0053] In the comparison process, the curve pattern that meets the characteristics of a specific fluid type is determined by matching the trend slope of the flow value with pressure, the position of the inflection point, and the data distribution range;

[0054] If the deviation between the variation trend of the current data point and a certain fluid characteristic curve is within a preset allowable range, the current fluid is determined to be this type of fluid;

[0055] If the current data cannot completely match the pre-stored fluid characteristic curve, the possible type of the fluid is determined based on the closest matching result and combined with the priority rules set by the user, and the current data is stored as a reference for subsequent optimization of the characteristic curve.

[0056] The control unit can effectively identify the current fluid type by analyzing the relationship between the corrected actual flow value and the upstream and downstream pressure data, and adopts a multi-point matching analysis method, and provides basic support for the intelligent control of the reverse flow valve. The identification process covers steps such as data reception, feature matching, type determination, and adaptive optimization to ensure that it can accurately adapt to a variety of complex working conditions.

[0057] The control unit first receives the corrected actual flow value and upstream and downstream pressure data transmitted in real time. These data are collected by the multimodal sensing unit and processed by density and pressure correction calculation to ensure the accuracy and consistency of the input data. The flow value reflects the dynamic change characteristics of the fluid passing through the valve body, while the upstream and downstream pressure data provide the resistance characteristics of the fluid flow process. These data together constitute the basic parameters of fluid behavior.

[0058] During the data processing phase, the control unit compares the received flow value and pressure data with the characteristic curves of various fluid types pre-stored in the memory. The characteristic curves are generated based on the historical data of different fluid types and contain typical relationships between flow values ​​and pressure data, such as linear or nonlinear trends, slope changes, key inflection points, and data distribution ranges. During the comparison process, the control unit uses a multi-point matching analysis method, that is, it matches multiple feature points in the current input data with the corresponding points on the characteristic curve one by one, and evaluates the similarity of their trend slopes, the degree of overlap of inflection point positions, and the data distribution range.

[0059] If the deviation between the trend of the current data point and a specific fluid characteristic curve is within the preset allowable range, the current fluid is determined to be of that specific type. The setting of the allowable range takes into account measurement errors, environmental fluctuations, and natural differences between different fluid types, thereby ensuring that the identification process is both sensitive and reliable. If the current data cannot completely match any characteristic curve in the memory, the control unit will infer the possible type of fluid based on the closest match result and the priority rules set by the user. For example, for fluids with similar flow and pressure characteristics, the priority rules can be adjusted according to the operating conditions or the importance weight of historical data.

[0060] In addition, while identifying the fluid type, the control unit stores the currently unmatched or partially matched data as new characteristic curve candidate data for subsequent optimization of the characteristic curve library. Through this mechanism, the system can continuously learn and improve during operation, improving its adaptability to new or special fluid types.

[0061] Overall, the control unit builds an efficient and dynamic fluid identification system through multi-point matching analysis and adaptive optimization.

[0062] Furthermore, the control unit is used to adaptively adjust the backflow detection threshold according to the identified fluid type, specifically including:

[0063] After identifying the current fluid type, historical operating data and characteristic parameters related to the fluid type are retrieved from the memory, including typical pressure difference range, flow stability index and dynamic response characteristics;

[0064] According to the characteristic parameters of the fluid type, set the initial backflow detection threshold, and monitor the upstream and downstream pressure difference change trend in real time during operation;

[0065] When the operating pressure difference is close to the set initial threshold, the current pressure difference fluctuation amplitude, frequency and flow rate change rate are analyzed, and the numerical range of the backflow detection threshold is dynamically adjusted according to the analysis results to expand or narrow the allowable pressure difference range to adapt to the fluid behavior characteristics under specific working conditions.

[0066] The control unit provides an intelligent control method that dynamically responds to complex working conditions by identifying the current fluid type and adaptively adjusting the backflow detection threshold. This process is based on historical data and real-time data, combined with the characteristic parameters of the fluid and the operating status, to achieve dynamic optimization of the backflow detection threshold.

[0067] After identifying the current fluid type, the control unit first retrieves the historical operating data and characteristic parameters related to the fluid type from the memory. These data include the pressure difference range, flow stability index and dynamic response characteristics of the fluid under typical working conditions. The pressure difference range reflects the upstream and downstream pressure difference fluctuation characteristics of the fluid under normal flow conditions; the flow stability index is used to quantify the smoothness of flow and pressure changes; the dynamic response characteristics describe the flow change trend and response speed of the fluid under different pressure difference conditions. These characteristic parameters provide a key basis for the subsequent adjustment of the backflow detection threshold.

[0068] Based on the retrieved fluid characteristic parameters, the control unit sets the initial backflow detection threshold. This initial threshold is generally within the pressure difference range of historical operation data, which can avoid false alarms of backflow and respond to potential abnormal flows in a timely manner. At the same time, the control unit monitors the change trend of the upstream and downstream pressure difference in real time, including the real-time value, change amplitude and fluctuation frequency of the pressure difference.

[0069] When the operating pressure difference approaches the set initial threshold, the control unit will conduct a detailed analysis of the current pressure difference fluctuation amplitude, frequency, and flow rate change speed. For example, if the pressure difference fluctuation amplitude is small and the frequency is stable, it means that the fluid flow is relatively stable. The control unit may appropriately expand the allowable pressure difference range to reduce unnecessary operations caused by overly sensitive backflow detection; on the contrary, if the pressure difference fluctuation amplitude is large or the frequency is unstable, accompanied by abnormal flow rate change speed, the control unit will appropriately narrow the allowable pressure difference range to improve the sensitivity and accuracy of backflow detection.

[0070] By analyzing real-time data during operation and combining it with historical characteristic parameters, the control unit can dynamically adjust the numerical range of the backflow detection threshold. This adjustment mechanism allows the system to flexibly adapt to the behavior characteristics of the fluid under different working conditions. For example, in industrial pipelines, when the fluid is a low-viscosity liquid and the flow rate is high, the threshold may need to be narrowed to cope with rapid changes; in scenarios with high-viscosity fluids or low flow rates, the threshold may need to be expanded to avoid false alarms.

[0071] The control unit can not only optimize the initial threshold according to the current fluid type, but also improve the accuracy and reliability of backflow detection through real-time monitoring and dynamic adjustment.

[0072] Furthermore, the control unit uses the following formula 1, according to the temperature ,pressure And the density correction factor calculated from the fluid type characteristic parameters :

[0073] ;

[0074] in, is the reference temperature and reference pressure density of a fluid under conditions of ; usually determined by standard experimental methods. For common fluids, such as water, At standard atmospheric pressure (and 25°C) this is approximately 1000 kg / m³.

[0075] It is the real-time pressure value of the fluid, in Pa, collected in real time by the pressure sensor. The measured data of inlet pressure and outlet pressure are usually stored separately, and the relevant pressure data can be selected according to the specific location where the fluid flows.

[0076] is with The corresponding reference pressure value is usually standard atmospheric pressure or a reference value that matches the specific working conditions.

[0077] It is the fluid temperature value collected in real time by the temperature sensor. The temperature sensor is usually installed in the flow channel inside the valve body to monitor the temperature changes of the fluid in real time.

[0078] is with The corresponding reference temperature value is usually the standard temperature of the physical property data.

[0079] is the expansion coefficient of the temperature effect on density, The value of is usually determined experimentally and is related to the thermal expansion properties of the fluid. For common liquids, such as water, Usually in arrive Recommended value: For water, .

[0080] is the compressibility factor of the effect of pressure on density, The value of is related to the compressibility of the fluid and is usually determined experimentally. For example, for an incompressible fluid, Smaller (recommended value ; For compressed gases, Larger value (recommended value ).

[0081] Furthermore, the control unit uses the following formula 2, according to the inlet pressure of the fluid , outlet pressure , flow rate and viscosity coefficient Calculating the pressure correction factor :

[0082] ;

[0083] Inlet pressure and outlet pressure It is the fluid pressure value collected in real time by the pressure sensor of the multi-modal sensing unit, in Pascal. The difference between the two represents the pressure drop caused by the flow resistance of the fluid in the flow channel, and The ratio reflects the proportion of pressure drop to total pressure and is the basis for pressure correction.

[0084] is the viscosity coefficient that describes the internal resistance of the fluid, usually obtained through experimental measurement or by consulting a fluid property table. For example, water at 25°C Pa·s, air at the same temperature Pa·s.

[0085] It is the average flow velocity of the fluid in the flow channel. The flow value is collected by the flow sensor and calculated in combination with the cross-sectional area of ​​the flow channel. The formula is: ,in For flow, is the cross-sectional area of ​​the flow channel. For example, for a typical industrial pipeline, Usually between 0.1 m / s and 5 m / s.

[0086] in, and is the empirical correction factor related to the fluid type;

[0087] It is used to correct the sensitivity of the pressure correction factor to the effect of fluid viscosity and flow rate. It is dimensionless and the value range is determined according to the physical properties of the fluid. For example, the pressure of low viscosity liquids such as water is The recommended value is 0.5 to 1.5; for highly viscous liquids such as oil, it may be 2.0 to 4.0. For gases, Can be set from 0.8 to 1.2.

[0088] It is the effect of the modified Reynolds number on the fluid flow characteristics. It is dimensionless. Its value is usually used to avoid calculation instability caused by zero or too small Reynolds number. The recommended value is 1 to 10, depending on the fluid characteristics and application scenarios.

[0089] is the Reynolds number, calculated according to the following formula 3:

[0090] ;

[0091] in, is the flow channel diameter, which is usually determined by the pipeline design parameters. is the density correction factor calculated by formula (1).

[0092] Furthermore, the control unit uses the following formula 4, according to the density correction coefficient , pressure correction factor And the raw flow value of the flow sensor Calculate the corrected actual flow value :

[0093] ;

[0094] Among them, the original flow value It is the original flow value of the fluid directly measured by the flow sensor. The flow sensor is usually installed in the internal flow channel of the valve body to monitor the flow of the fluid in real time. Since the measurement process may be affected by environmental changes or sensor accuracy, Further corrections are usually required to improve accuracy.

[0095] It is the density correction factor calculated by formula (1), which reflects the combined effect of temperature and pressure on fluid density. Its sources include real-time temperature and pressure data and the reference density and characteristic parameters of the fluid. For different fluids, Can accurately compensate for density differences caused by changing environmental conditions.

[0096] Pressure correction factor It is the pressure correction coefficient calculated by formula (2) and is used to describe the dynamic effect of the inlet and outlet pressure difference on the fluid flow. It combines the characteristics of fluid viscosity, flow velocity and Reynolds number and is dimensionless. This coefficient can effectively adjust the flow deviation caused by pressure changes.

[0097] and It is the weight index of density and pressure in the correction calculation, and the dimension is dimensionless. These two indices are used to balance the weight of the influence of density correction factor and pressure correction factor on flow calculation.

[0098] For low pressure fluids (such as gases), the recommended value is , , emphasizing the main effect of density on flow.

[0099] For high pressure fluids (such as liquids), the recommended value is , , highlighting the importance of pressure correction.

[0100] It is a dynamic correction factor related to fluid inertia, reflecting the effect of fluid inertia on flow fluctuation. Its value is usually determined by experiment:

[0101] For low inertia fluids (such as gases), the recommended value is to ;

[0102] For high inertia fluids (such as liquids), the recommended value is to .

[0103] Indicates the time rate of change of flow velocity, calculated from the continuous sampling data of the flow sensor.

[0104] The wireless communication unit 105 is arranged at the top of the valve body. The wireless communication unit is electrically connected to the control unit and supports wireless protocol communication. It is used to transmit the real-time data of the control unit to an external monitoring device and receive the control parameters issued by the external monitoring device. The control parameters include the pressure difference threshold of backflow detection, valve response time, valve closing speed and opening limit value.

[0105] The wireless communication unit 105 is a core component for realizing the remote monitoring function of the Internet of Things. It is installed on the top of the valve body 101 and is electrically connected to the control unit 104 through a standardized interface. Its main function is to realize wireless transmission and remote interaction of data, including sending the data processed in real time by the control unit to the external monitoring device, and receiving the control parameters issued by the external monitoring device, thereby supporting the intelligent remote management of the reverse flow valve.

[0106] The hardware structure of the wireless communication unit includes a communication module, a power management module, a signal processing circuit, and an antenna. The communication module is its core part and supports a variety of wireless communication protocols, such as Wi-Fi, LoRa, Zigbee, NB-IoT, or cellular networks (such as 4G / 5G). The specific protocol to be selected depends on the coverage, power consumption requirements, and communication bandwidth requirements of the application scenario. For example, for long-distance, low-power scenarios, LoRa or NB-IoT can be selected, while for high-bandwidth data transmission requirements, Wi-Fi or cellular networks can be selected.

[0107] The design of the antenna takes into account signal strength, directivity and interference suppression capabilities. Omnidirectional antennas or directional antennas are usually used, and they are optimized according to the installation environment. The antenna is connected to the communication module through a coaxial cable or PCB integration to ensure efficient transmission and reception of signals. In order to ensure the stability of the equipment, the shell material of the antenna is usually corrosion-resistant and waterproof, and it is designed with lightning protection to adapt to the complex conditions of the industrial environment.

[0108] The wireless communication unit converts the digital signal of the control unit into a standard signal that complies with the wireless protocol for transmission through the signal processing circuit, and decodes the received wireless signal and restores it to a digital signal that can be recognized by the control unit. This two-way signal conversion ensures the accurate transmission of data. In terms of power management, the wireless communication unit usually obtains DC power from the control unit or an external power supply device. The power management module can adjust the input voltage, provide a stable operating voltage, and has an overcurrent protection function.

[0109] During the data transmission process, the wireless communication unit sends the real-time data generated by the control unit 104, including the fluid inlet and outlet pressure, flow, temperature, and backflow detection status, to the external monitoring device through the wireless network. The external monitoring device can be a cloud platform, an industrial control system, or a local computer, which generates a more advanced control strategy by further processing and analyzing these data. The external device sends optimized control parameters based on the data analysis results, such as the pressure difference threshold for backflow detection, valve response time, closing speed, and opening limit value. After receiving these parameters, the wireless communication unit immediately passes them to the control unit to update the control logic and guide the operation of the valve.

[0110] The wireless communication unit also has data encryption and identity authentication functions to ensure the security and reliability of data transmission. The encryption protocol can use AES, TLS or other common encryption algorithms to prevent data from being illegally intercepted or tampered with during transmission. The identity authentication function verifies the legitimacy of the device through a digital certificate or encryption key to prevent unauthorized devices from accessing the system.

[0111] In addition, the design of the wireless communication unit fully considers the needs of the industrial environment. Its housing meets the protection standards of IP67 or higher, and is dustproof, waterproof, and has the ability to resist vibration and electromagnetic interference. When installed, the wireless communication unit can be bolted to the top of the valve body to ensure that its position is convenient for signal transmission while avoiding the influence of pipeline vibration or other mechanical stress.

[0112] Through the wireless communication unit 105, the real-time data transmission and remote control functions of the reverse flow valve are realized, providing strong support for the application of industrial Internet of Things.

[0113] Furthermore, the external monitoring device includes a pre-trained neural network model for generating control parameters, the neural network model including a feature extraction module, a fluid behavior prediction module and a control parameter generation module;

[0114] The feature extraction module is used to receive real-time data transmitted by the control unit, including the corrected actual flow value, inlet pressure, outlet pressure, temperature and current fluid type; extract high-dimensional features of the input data through a multi-layer convolutional neural network, capture the trend pattern of flow, pressure and temperature changes over time, and generate a set of high-dimensional feature vectors describing the dynamic characteristics of the fluid;

[0115] The fluid behavior prediction module is used to receive the high-dimensional feature vector, and combine the historical operation data and the stored fluid characteristic parameters, and use the long short-term memory network to predict the pressure difference, flow rate and temperature change trend of the fluid in the future specified time window; the output includes the predicted future pressure difference range, flow rate change rate and temperature stability index;

[0116] The control parameter generation module is used for the output of the fluid behavior prediction module and the current fluid type information to generate control parameters through a fully connected neural network and an optimization layer based on reinforcement learning; a reinforcement learning algorithm is used to evaluate the safety and efficiency of different control strategies under the current working conditions, and the weights of the neural network are dynamically adjusted to generate the optimal backflow detection pressure difference threshold, valve response time, valve closing speed and opening limit value.

[0117] The external monitoring device generates control parameters through a pre-trained neural network model to support the intelligent and efficient operation of the reverse flow valve. The design of the neural network model includes a feature extraction module, a fluid behavior prediction module, and a control parameter generation module. The functions and implementations of each part complement each other to jointly complete the optimal control under complex fluid conditions.

[0118] The function of the feature extraction module is to extract high-dimensional features from the real-time data transmitted by the control unit. The real-time data includes parameters such as the corrected actual flow value, inlet pressure, outlet pressure, temperature, and current fluid type. These data are processed by a multi-layer convolutional neural network. The design of the convolutional network can capture the changing trends of the input data in time and space, such as flow fluctuations, dynamic behavior of pressure changes, and stability characteristics of temperature. The convolution kernel of each layer of the convolutional network is adaptive and can adjust the weights according to the distribution characteristics of the input data to more accurately extract the fluid behavior characteristics. Finally, the feature extraction module outputs a set of high-dimensional feature vectors, which not only describe the current fluid dynamic characteristics, but also provide a sufficient basis for subsequent prediction and control.

[0119] The fluid behavior prediction module receives the high-dimensional feature vector generated by the feature extraction module and makes predictions based on the historical operation data and fluid characteristic parameters. The historical operation data includes the pressure, flow rate and temperature change trends over the past period of time, while the fluid characteristic parameters include the physical properties of the fluid such as density, viscosity and thermal conductivity. The prediction module processes the input data through the long short-term memory network (LSTM). The multi-layer structure of the LSTM network can capture short-term changes and long-term trends, such as the cumulative effect of rapid pressure fluctuations or long-term temperature changes. Through the analysis of time series data, the module can predict the pressure difference range, flow rate change rate and temperature stability index within the specified time window in the future. These prediction results are output in the form of multi-dimensional vectors, providing a key reference for the generation of control parameters.

[0120] The function of the control parameter generation module is to generate optimized control parameters based on the output of the fluid behavior prediction module and the current fluid type information. This module performs preliminary processing on the input data through a fully connected neural network, converting complex prediction results into actionable parameter data. At the same time, the module further combines the optimization layer of reinforcement learning to dynamically adjust the generated control parameters. The reinforcement learning optimization layer defines a reward function to evaluate the safety, efficiency and response speed of different control strategies under the current working conditions, and adjusts the weights of the neural network in real time to ensure that the generated control parameters can achieve the best effect. The control parameters finally generated include the pressure difference threshold, valve response time, valve closing speed and opening limit value for backflow detection. These parameters can be directly applied to the control unit to realize the intelligent operation of the backflow valve.

[0121] The reward function is designed as follows to combine multi-objective optimization strategies such as safety, efficiency, and response speed. The specific formula is:

[0122] ;

[0123] The following is the meaning of each term in the formula: Security term: : The actual measured upstream and downstream pressure difference (unit: Pa), collected in real time by the pressure sensor. : The backflow detection pressure difference value set by the current control strategy, provided by the control parameter generation module. This item reflects the relative deviation between the actual pressure difference and the detection threshold. The smaller the deviation, the higher the safety. Use the square form to amplify the deviation to avoid high-risk status. Weight Indicates the importance of safety in the overall reward. The recommended value range is 0.4 to 0.6. Efficiency term: : System efficiency index, defined as the ratio of actual energy consumption to theoretical minimum energy consumption during valve operation, with a value range of . Energy consumption calculation through real-time monitoring of valve action. : Valve response time (unit: s), which indicates the time required from the detection of an abnormality to the start of valve action, and is directly recorded by the system operating status. : Factor that adjusts the balance between efficiency and response time. The recommended value is 0.1 to 1. The specific value depends on the fluid type and working conditions. This item encourages control strategies with higher efficiency and shorter response time. Increased sensitivity to response time. Indicates the importance of efficiency in the overall reward. Recommended values ​​range from 0.3 to 0.5.

[0124] Response speed and penalty terms: ;

[0125] : Current valve closing speed (unit: ), which is monitored by the actual movement speed of the electric actuator. : Maximum allowable closing speed of the valve (unit: ), which is set by the physical parameters of the device. : Reflects the relative efficiency of valve response speed, with a range of values . :Potential penalty indicators triggered by the strategy, such as pipeline impact, backflow risk, etc., are calculated by real-time analysis of upstream and downstream pressure fluctuations and flow rate changes.

[0126] Penalty Indicator It quantifies the negative impact of a strategy by analyzing the dynamic behavior of the system in real time. It takes into account the following key factors:

[0127] Impact on pipelines: Based on the pressure fluctuation amplitude and frequency detected by upstream and downstream pressure sensors, analyze whether the pressure change exceeds the safe tolerance range of the pipeline. If the pressure fluctuation is too large or the change is too drastic, it may cause mechanical stress accumulation in the pipeline structure and increase the risk of pipeline damage.

[0128] Backflow risk: By monitoring the upstream and downstream pressure difference and flow direction, determine whether there is a possibility of backflow. For example, when the outlet pressure suddenly increases and exceeds the inlet pressure, the system may mark this state as a potential backflow risk and quantify its impact as a penalty value.

[0129] System stability: The rate and frequency of flow rate changes are combined to assess whether the flow of the fluid is stable. If the flow rate changes too quickly or too frequently, it may cause turbulence or unstable flow, which will increase the uncertainty of system operation.

[0130] Equipment wear: Consider the impact of valve closing or opening adjustments too quickly on mechanical components. If the control strategy causes the equipment to move too violently, it may increase long-term wear and maintenance requirements.

[0131] By converting the above factors into quantifiable indicators, the control unit assigns weights according to their severity and comprehensively calculates an overall penalty value. The larger the penalty value, the more significant the potential negative impact of the current strategy on the system. This penalty indicator will be used as a constraint in the reinforcement learning process to ensure that the generated control strategy minimizes damage to system stability and equipment life while optimizing safety and efficiency.

[0132] : Penalty term sensitivity coefficient, the recommended value range is 0.5 to 2.0.

[0133] Meaning: This option encourages valves to respond quickly, while penalizing strategies that have a greater impact on system stability. Limit excessive penalty values.

[0134] Weight : Indicates the importance of response speed in the overall reward. The recommended value range is 0.2 to 0.4.

[0135] The control parameter generation module calculates each part of the above reward function based on the real-time input fluid behavior prediction data and the current valve state. The reinforcement learning optimization layer maximizes the reward function The control strategy is evaluated and updated in real time. The formula is combined with multi-objective optimization design to ensure the dynamic balance of safety, efficiency and response speed, and is suitable for intelligent operation under different working conditions.

[0136] Overall, the neural network model has built a complete control chain from data extraction to prediction analysis to parameter generation through the organic combination of three modules. The model is optimized through a large amount of fluid data during the training phase to ensure its accuracy and robustness in actual working conditions; during the operation phase, it can be dynamically adjusted in combination with real-time data to ensure its adaptability to complex working conditions.

[0137] Furthermore, the feature extraction module includes a sliding window mechanism for time series data processing, which is used to divide the real-time data transmitted by the control unit into time windows of fixed length to capture short-term and long-term change patterns; within each time window, the multi-layer convolutional neural network extracts the spatial and temporal features of flow, pressure and temperature data through adaptive convolution kernels; the extracted high-dimensional feature vector includes indicators reflecting the pressure change rate, flow fluctuation amplitude and temperature stability.

[0138] The feature extraction module uses a sliding window mechanism for time series data processing to accurately capture the dynamic change characteristics of real-time data and extract high-dimensional features. The implementation of this mechanism relies on the segmented processing of real-time data transmitted by the control unit, dividing the continuous data stream into time windows of fixed length, so that the data in each window can fully reflect the change pattern within a specific time period.

[0139] The core of the sliding window mechanism is the dynamic data slicing and window update strategy. During operation, real-time data is sampled and stored at fixed intervals, and the sliding window extracts a fixed-length subset from the data stream. For example, for the corrected actual flow value, inlet pressure, outlet pressure and temperature data, the sliding window can capture data from the past few seconds to minutes. These windows have a certain degree of overlap, so that rapid changes in the short term can be continuously captured during the window movement while preserving the integrity of the data. The window length and overlap can be configured according to the specific application requirements to adapt to the dynamic behavior characteristics under different fluid conditions.

[0140] In each time window, a multi-layer convolutional neural network (CNN) processes the segmented data. The convolution layer of the CNN scans the input data through the convolution kernel to extract the spatial and temporal features of the data. For example, for pressure and flow data, the convolution kernel can identify the slope, fluctuation frequency and peak position of pressure changes, as well as the amplitude and periodicity of flow fluctuations. The size and number of convolution kernels are dynamically adjusted by the network structure to adapt to the distribution characteristics of the input data. This adaptability ensures that the network can optimize the feature extraction process for different fluid properties.

[0141] The high-dimensional feature vector output by the feature extraction module is an efficient encoding of the original data, which can reflect the core dynamic indicators within the time window. For example, the high-dimensional feature vector may include an indicator reflecting the pressure change rate, which is used to describe the rapid change characteristics of pressure; an indicator of flow fluctuation amplitude, which is used to evaluate the stability of fluid flow; and an indicator of temperature stability, which is used to monitor the change trend of fluid temperature. These indicators can comprehensively reflect the dynamic behavior of the fluid and provide an important basis for subsequent prediction analysis and control parameter generation.

[0142] By introducing the sliding window mechanism, the feature extraction module of the present invention can capture short-term and long-term change patterns in real-time data processing, so that the neural network can still accurately extract core features under complex working conditions.

[0143] Furthermore, the fluid behavior prediction module includes a multi-layer long short-term memory network structure, wherein each layer is used to capture fluid behavior patterns at different time scales; the first layer is used for short-term prediction, focusing on rapidly changing pressure and flow characteristics, and the second layer is used for long-term prediction, identifying the cumulative effects of temperature stability and fluid flow trends;

[0144] The fluid behavior prediction module combines the stored fluid characteristic parameters and assigns different historical data weights through the attention mechanism to improve the prediction accuracy of fluid dynamics. The output future pressure difference range, flow rate change rate and temperature stability index are multidimensional data structures.

[0145] The fluid behavior prediction module uses a multi-layer long short-term memory network (LSTM) structure to accurately predict the dynamic behavior of complex fluids. The LSTM network has the ability to capture short-term and long-term correlations in time series data. Through the design of a multi-layer structure, the module can analyze and model fluid behavior patterns at different time scales.

[0146] In the first layer, the network focuses on short-term predictions, mainly analyzing the fast-changing characteristics of pressure and flow data. The real-time input fluid data is divided into small time slices, and the LSTM unit uses its memory gating mechanism to identify patterns with significant time dependence in the input data, such as sudden changes in pressure and sharp fluctuations in flow. The output of this layer can reflect the rapid response characteristics of the fluid system in a short period of time, providing support for higher-level long-term predictions.

[0147] The second layer of the network is used for long-term prediction, focusing on identifying the cumulative effects of temperature stability and fluid flow trends. By modeling data within a larger time window, this layer is able to capture slowly changing but highly indicative dynamic patterns, such as a gradual increase or decrease in fluid temperature, and smooth changes in long-term pressure trends. This information is of great value for fluid control under complex conditions because it directly affects the reliability of backflow detection and the rationality of thresholds.

[0148] To further improve the accuracy of the prediction, the fluid behavior prediction module combines stored fluid characteristic parameters, such as the density, viscosity, compressibility, and thermal conductivity of the fluid. By combining real-time data with these parameters, the module is able to more accurately model and predict the behavior of specific fluid types. In addition, the module uses an attention mechanism to assign weights to different historical data. The attention mechanism assigns higher weights to data that has a greater impact on the prediction results by calculating the correlation between historical data and current data. This mechanism can effectively avoid the interference of redundant information and highlight the contribution of key data points, thereby further optimizing the prediction performance.

[0149] The output of the fluid behavior prediction module is a multidimensional data structure, including the predicted future pressure difference range, flow rate change rate and temperature stability index. These output data comprehensively reflect the dynamic change characteristics of the fluid system in the future specified time window, providing a reliable basis for the subsequent control parameter generation.

[0150] Furthermore, the control parameter generation module includes a reinforcement learning optimization layer, which evaluates the control effect under the current working conditions by constructing a multi-objective optimization strategy; the strategy takes into account the accuracy of backflow detection, the agility of valve response time and the energy efficiency of fluid control; the reinforcement learning algorithm dynamically adjusts the weights of the fully connected neural network by defining a reward function, and the generated control parameters include two configurations: safety priority mode and efficiency priority mode, which are used for real-time adaptation of different working conditions.

[0151] The control parameter generation module realizes dynamic generation and adaptive adjustment of control parameters under different working conditions through the reinforcement learning optimization layer. This module aims to generate a set of optimal control parameters based on the current fluid state, system operating conditions and user needs to ensure the safety and efficiency of the reverse flow valve under complex working conditions.

[0152] The reinforcement learning optimization layer comprehensively evaluates the control effect under the current working conditions by constructing a multi-objective optimization strategy. The multi-objective optimization strategy covers three core evaluation dimensions, including the accuracy of backflow detection, the agility of valve response time, and the energy efficiency of fluid control. The accuracy of backflow detection is related to the ability to identify and respond to potential backflow risks; the agility of valve response time reflects the system's ability to adapt to rapidly changing working conditions; and energy efficiency focuses on the rational use of resources during valve operation and control. By weighing these three goals, the module can generate optimized control parameters suitable for the current working conditions.

[0153] The reinforcement learning algorithm introduces a reward function in the optimization process to quantify the comprehensive performance of different control strategies on the above objectives. The reward function is dynamically adjusted based on the real-time collected data and feedback information from the system operation. For example, when the fluid flow stability is detected to be high, the reward function can give priority to energy efficiency to reduce resource waste; in the case of drastic fluid flow fluctuations or potential backflow risks, the reward function will increase the weight of backflow detection accuracy and response time to ensure the safe operation of the system. The reinforcement learning optimization layer learns the best parameter configuration from experience by constantly trying and adjusting strategies, thereby dynamically optimizing the weights of the fully connected neural network.

[0154] In terms of specific implementation, the control parameters generated by this module include two configuration modes: safety priority mode and efficiency priority mode. Safety priority mode is suitable for working conditions with high backflow risk, such as environments with complex fluid types or large pressure difference fluctuations; efficiency priority mode is suitable for scenarios with stable fluid state and relatively stable operating conditions, and maximizes operating efficiency by optimizing the valve response time and controlling energy consumption. The module can dynamically switch between these two modes according to real-time working conditions to ensure the flexible adaptability of the system.

[0155] By introducing the reinforcement learning optimization layer, the control parameter generation module can respond to complex fluid control requirements under different working conditions in an intelligent manner.

[0156] The following is the reference implementation code of the neural network model:

[0157] import torch

[0158] import torch.nn as nn

[0159] import torch.optim as optim

[0160] import numpy as np

[0161] # Define feature extraction module

[0162] class FeatureExtractionModule(nn.Module):

[0163] """

[0164] Feature extraction module: extracts high-dimensional features from real-time input through a multi-layer convolutional neural network.

[0165] Input: Corrected flow value, inlet pressure, outlet pressure, temperature, current fluid type

[0166] Output: high-dimensional feature vector

[0167] """

[0168] def __init__(self, input_channels, output_dim):

[0169] super(FeatureExtractionModule, self).__init__()

[0170] # First layer of convolution + activation + pooling

[0171] self.conv1 = nn.Conv1d(input_channels, 64, kernel_size=3, padding=1)

[0172] self.relu1 = nn.ReLU()

[0173] self.pool1 = nn.MaxPool1d(kernel_size=2)

[0174] # Second layer of convolution + activation + pooling

[0175] self.conv2 = nn.Conv1d(64, 128, kernel_size=3, padding=1)

[0176] self.relu2 = nn.ReLU()

[0177] self.pool2 = nn.MaxPool1d(kernel_size=2)

[0178] # The fully connected layer outputs a high-dimensional feature vector

[0179] self.fc = nn.Linear(128, output_dim)

[0180] def forward(self, x):

[0181] """

[0182] x: input time series data (batch_size, input_channels, time_steps)

[0183] """

[0184] x = self.pool1(self.relu1(self.conv1(x)))

[0185] x = self.pool2(self.relu2(self.conv2(x)))

[0186] x = torch.mean(x, dim=2) # Average across time dimensions

[0187] x = self.fc(x)

[0188] return x

[0189] # Define the fluid behavior prediction module

[0190] class FluidBehaviorPredictionModule(nn.Module):

[0191] """

[0192] Fluid behavior prediction module: Predict future behavior by combining historical data with multi-layer LSTM.

[0193] Input: high-dimensional feature vector

[0194] Output: Future pressure difference range, flow rate change rate, temperature stability index

[0195] """

[0196] def __init__(self, input_dim, hidden_dim, num_layers, output_dim):

[0197] super(FluidBehaviorPredictionModule, self).__init__()

[0198] self.lstm = nn.LSTM(input_dim, hidden_dim, num_layers, batch_first=True)

[0199] self.fc = nn.Linear(hidden_dim, output_dim)

[0200] def forward(self, x, hidden_state=None):

[0201] """

[0202] x: high-dimensional feature vector sequence (batch_size, seq_length, input_dim)

[0203] hidden_state: LSTM state of the previous time step (h_0, c_0)

[0204] """

[0205] lstm_out, hidden_state = self.lstm(x, hidden_state)

[0206] pred = self.fc(lstm_out[:, -1, :]) # Get the output of the last time step

[0207] return pred, hidden_state

[0208] # Define the control parameter generation module

[0209] class ControlParameterGenerationModule(nn.Module):

[0210] """

[0211] Control parameter generation module: Dynamically generate optimization parameters in combination with reinforcement learning.

[0212] Input: Fluid behavior prediction output + current fluid type information

[0213] Output: Optimized control parameters

[0214] """

[0215] def __init__(self, input_dim, output_dim):

[0216] super(ControlParameterGenerationModule, self).__init__()

[0217] self.fc = nn.Linear(input_dim, output_dim) # Preliminary parameter generation

[0218] self.relu = nn.ReLU()

[0219] def forward(self, x):

[0220] """

[0221] x: fluid behavior prediction output and fluid type information (batch_size, input_dim)

[0222] """

[0223] x = self.relu(self.fc(x))

[0224] return x

[0225] # Reinforcement Learning Optimization Layer (Reward Function Example)

[0226] def reward_function(delta_p_actual, delta_p_threshold, efficiency, response_time, close_speed, max_speed, penalty):

[0227] """

[0228] Define the reward function: multi-objective optimization that integrates safety, efficiency, and responsiveness.

[0229] """

[0230] w1, w2, w3 = 0.5, 0.3, 0.2 # Weights of each target

[0231] reward = (

[0232] w1 (1 - (delta_p_actual / delta_p_threshold)) 2 +

[0233] w2 efficiency / (1 + 0.1 response_time) +

[0234] w3 (close_speed / max_speed) np.exp(-1.0 penalty) )

[0236] return reward

[0237] # Overall neural network model

[0238] class FlowControlNetwork(nn.Module):

[0239] """

[0240] Overall model: includes feature extraction module, fluid behavior prediction module and control parameter generation module.

[0241] """

[0242] def __init__(self, input_channels, feature_dim, lstm_hidden_dim, lstm_layers, output_dim):

[0243] super(FlowControlNetwork, self).__init__()

[0244] self.feature_extraction = FeatureExtractionModule(input_channels,feature_dim)

[0245] self.behavior_prediction = FluidBehaviorPredictionModule(feature_dim, lstm_hidden_dim, lstm_layers, output_dim)

[0246] self.control_generation = ControlParameterGenerationModule(output_dim + 1, 4) # Contains fluid type information

[0247] def forward(self, real_time_data, fluid_type_info, hidden_state=None):

[0248] """

[0249] real_time_data: real-time data input

[0250] fluid_type_info: Current fluid type information

[0251] hidden_state: LSTM state

[0252] """

[0253] features = self.feature_extraction(real_time_data)

[0254] predictions, hidden_state = self.behavior_prediction(features.unsqueeze(1), hidden_state)

[0255] combined_input = torch.cat([predictions, fluid_type_info], dim=1)

[0256] control_params = self.control_generation(combined_input)

[0257] return control_params, hidden_state

[0258] The steps to train this neural network model include: first, prepare a data set, including corrected real-time flow values, inlet pressure, outlet pressure, temperature, current fluid type, and historical operation data. All data must be standardized. Secondly, divide the data into training and validation sets, and generate time series inputs with a sliding window mechanism for feature extraction modules and time-dependent modeling of LSTM. Then, define a loss function, such as mean square error (MSE) for the behavior prediction module, and the reinforcement learning part uses a reward function to evaluate the optimization effect of the generated parameters. Through backpropagation and optimizers (such as Adam), the model weights are gradually adjusted to minimize the loss function. Finally, monitor the performance of the validation set during model training to avoid overfitting, and save the final model for real-time application after performance convergence.

[0259] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

Claims

1. A reverse flow valve supporting remote monitoring of the Internet of Things, characterized in that: include: a valve body having a fluid inlet, a fluid outlet, and an internal flow passage; An electric actuator, disposed on the valve body, for controlling the valve opening; A multimodal sensing unit, electrically connected to the control unit, comprising an inlet pressure sensor, an outlet pressure sensor, a flow sensor and a temperature sensor, for real-time acquisition of valve operation parameters; wherein the inlet pressure sensor is arranged at the fluid inlet of the valve body, for measuring the fluid inlet pressure; the outlet pressure sensor is arranged at the fluid outlet of the valve body, for measuring the fluid outlet pressure; the flow sensor is arranged in the internal flow channel of the valve body, for measuring the fluid flow; the temperature sensor is arranged in the internal flow channel of the valve body, for measuring the fluid temperature; A control unit is arranged on the valve body, the control unit includes a microprocessor and a memory, and is used to receive and process the pressure, flow and temperature data collected by the multimodal sensing unit; calculate the fluid density correction coefficient according to the fluid temperature data collected by the temperature sensor; calculate the pressure correction coefficient according to the pressure data collected by the inlet pressure sensor and the outlet pressure sensor; correct the original flow data collected by the flow sensor according to the fluid density correction coefficient and the pressure correction coefficient to obtain a corrected actual flow value; identify the current fluid type based on the relationship between the corrected actual flow value and the pressure data; adaptively adjust the backflow detection threshold according to the identified fluid type, wherein the backflow detection threshold is a threshold of the upstream and downstream pressure difference; monitor the pressure difference between the inlet pressure and the outlet pressure, and when the pressure difference exceeds the adjusted backflow detection threshold, control the electric actuator to perform an emergency shutdown operation; A wireless communication unit is arranged at the top of the valve body. The wireless communication unit is electrically connected to the control unit and supports wireless protocol communication. It is used to transmit real-time data of the control unit to an external monitoring device and receive control parameters issued by the external monitoring device. The control parameters include a pressure difference threshold for backflow detection, a valve response time, a valve closing speed, and an opening limit value.

2. The reverse flow valve supporting remote monitoring of the Internet of Things according to claim 1 is characterized in that: The external monitoring device includes a pre-trained neural network model for generating control parameters, the neural network model includes a feature extraction module, a fluid behavior prediction module and a control parameter generation module; The feature extraction module is used to receive real-time data transmitted by the control unit, including the corrected actual flow value, inlet pressure, outlet pressure, temperature and current fluid type; extract high-dimensional features of the input data through a multi-layer convolutional neural network, capture the trend pattern of flow, pressure and temperature changes over time, and generate a set of high-dimensional feature vectors describing the dynamic characteristics of the fluid; The fluid behavior prediction module is used to receive the high-dimensional feature vector, and combine the historical operation data and the stored fluid characteristic parameters, and use the long short-term memory network to predict the pressure difference, flow rate and temperature change trend of the fluid in the future specified time window; the output includes the predicted future pressure difference range, flow rate change rate and temperature stability index; The control parameter generation module is used for the output of the fluid behavior prediction module and the current fluid type information to generate control parameters through a fully connected neural network and an optimization layer based on reinforcement learning; a reinforcement learning algorithm is used to evaluate the safety and efficiency of different control strategies under the current working conditions, and the weights of the neural network are dynamically adjusted to generate the optimal backflow detection pressure difference threshold, valve response time, valve closing speed and opening limit value.

3. The reverse flow valve supporting remote monitoring of the Internet of Things according to claim 2 is characterized in that: The feature extraction module includes a sliding window mechanism for time series data processing, which is used to divide the real-time data transmitted by the control unit into time windows of fixed length to capture short-term and long-term change patterns; within each time window, a multi-layer convolutional neural network extracts spatial and temporal features of flow, pressure and temperature data through an adaptive convolution kernel; the extracted high-dimensional feature vector includes indicators reflecting the pressure change rate, flow fluctuation amplitude and temperature stability.

4. The reverse flow valve supporting remote monitoring of the Internet of Things according to claim 2, characterized in that: The fluid behavior prediction module includes a multi-layer long short-term memory network structure, in which each layer is used to capture fluid behavior patterns at different time scales; the first layer is used for short-term prediction, focusing on rapidly changing pressure and flow characteristics, and the second layer is used for long-term prediction, identifying the cumulative effects of temperature stability and fluid flow trends; The fluid behavior prediction module combines the stored fluid characteristic parameters and assigns different historical data weights through the attention mechanism to improve the prediction accuracy of fluid dynamics. The output future pressure difference range, flow rate change rate and temperature stability index are multidimensional data structures.

5. The reverse flow valve supporting remote monitoring of the Internet of Things according to claim 2, characterized in that: The control parameter generation module includes a reinforcement learning optimization layer, which evaluates the control effect under the current working conditions by constructing a multi-objective optimization strategy; the strategy takes into account the accuracy of backflow detection, the agility of valve response time and the energy efficiency of fluid control; the reinforcement learning algorithm dynamically adjusts the weights of the fully connected neural network by defining a reward function, and the generated control parameters include two configurations: safety priority mode and efficiency priority mode, which are used for real-time adaptation of different working conditions.

6. The reverse flow valve supporting remote monitoring of the Internet of Things according to claim 1, characterized in that: The control unit is used to identify the current fluid type based on the relationship between the corrected actual flow value and the pressure data, specifically including: Receive corrected actual flow value and upstream and downstream pressure data; Perform multi-point matching analysis on the corrected actual flow value and the corresponding pressure data, and compare these data with characteristic curves of various fluid types pre-stored in the memory; In the comparison process, the curve pattern that meets the characteristics of a specific fluid type is determined by matching the trend slope of the flow value with pressure, the position of the inflection point, and the data distribution range; If the deviation between the variation trend of the current data point and a certain fluid characteristic curve is within a preset allowable range, the current fluid is determined to be this type of fluid; If the current data cannot completely match the pre-stored fluid characteristic curve, the possible type of the fluid is determined based on the closest matching result and combined with the priority rules set by the user, and the current data is stored as a reference for subsequent optimization of the characteristic curve.

7. The reverse flow valve supporting remote monitoring of the Internet of Things according to claim 1, characterized in that: The control unit is used to adaptively adjust the backflow detection threshold according to the identified fluid type, specifically including: After identifying the current fluid type, historical operating data and characteristic parameters related to the fluid type are retrieved from the memory, including typical pressure difference range, flow stability index and dynamic response characteristics; According to the characteristic parameters of the fluid type, set the initial backflow detection threshold, and monitor the upstream and downstream pressure difference change trend in real time during operation; When the operating pressure difference is close to the set initial threshold, the current pressure difference fluctuation amplitude, frequency and flow rate change rate are analyzed, and the numerical range of the backflow detection threshold is dynamically adjusted according to the analysis results to expand or narrow the allowable pressure difference range to adapt to the fluid behavior characteristics under specific working conditions.

Images