An industrial regulating valve remote monitoring and self-diagnosis method based on an internet of things control system

Abstract

The application discloses an industrial regulating valve remote monitoring and self-diagnosis method based on an Internet of Things control system and belongs to the technical field of industrial control systems, which comprises the following steps: collecting low-frequency operation data by an edge computing gateway; issuing a bipolar micro-pulse diagnosis instruction by a cloud diagnosis platform when it is determined that a steady-state condition is met; pre-waking up a high-frequency sensor by the edge computing gateway to establish a high-frequency data annular buffer area and controlling a positioner to execute a pulse instruction to superimpose positive and negative displacements, and the holding time of a single pulse is less than the volume delay time constant of a process fluid; uploading the acoustic emission high-frequency mutation characteristics of the critical point of the static friction to kinetic friction conversion of a valve rod to the cloud by the gateway as a reference point; diagnosing the mechanical degradation type by the cloud; and updating the dead zone compensation parameters by the gateway according to the mechanical degradation type. The application realizes zero net disturbance online testing by using fluid low-pass filtering effect, eliminates the power-on delay of hardware, and realizes lossless interception of weak fault characteristics and closed-loop repair.

Description

Technical Field

[0001] This invention relates to the field of industrial control system technology, and in particular to a method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things (IoT) control system. Background Technology

[0002] With the continuous evolution of industrial automation technology, traditional distributed control systems are rapidly transforming into IoT control systems based on edge computing and cloud collaboration. In modern industrial production processes, industrial control valves, as the core underlying physical execution nodes in fluid pipeline networks, directly determine the safety boundaries and product quality of the entire industrial control system through their operational stability and regulation accuracy. Therefore, achieving remote monitoring and early fault diagnosis of industrial control valves under an IoT control system architecture has become a key digital technology that urgently needs to be mastered in the field of industrial control systems.

[0003] Currently, existing IoT control systems generally adopt a passive data acquisition mode when monitoring industrial control valves. Edge computing devices can typically only collect low-frequency process operation data such as valve inlet pressure, valve outlet pressure, and valve opening at fixed intervals, and upload them to the cloud for simple threshold comparison. This passive monitoring method has serious lag, and often only issues an alarm when the valve has already experienced severe jamming or internal leakage and caused process abnormalities. It cannot detect hidden faults such as increased static friction of valve stem or slight wear of packing in the early stages of mechanical degradation. In order to achieve early diagnosis, theoretically, it is necessary to issue active probing commands to industrial control valves to obtain their dynamic friction characteristics.

[0004] Traditional offline pressure testing, while accurately acquiring valve mechanical properties, requires interrupting industrial production processes, which is absolutely unacceptable in actual chemical plants or refineries. If conventional step test signals are sent directly to the control valve via the IoT control system without shutting down the system, the sudden valve action will inevitably cause drastic transient fluctuations in flow and pressure within the fluid pipeline network, thereby disrupting the overall control stability of the process fluid. Furthermore, to accurately capture the weak signals of mechanical degradation, multimodal sensors such as acoustic emission and high-frequency vibration sensors are often required. However, keeping these high-frequency sensors constantly open in the IoT control system and uploading massive amounts of waveform data in real time would instantly exhaust the communication bandwidth of the industrial IoT and place an unbearable computing load on the edge computing gateway.

[0005] Therefore, the industry urgently needs a new control method that can efficiently and accurately perform online active micro-motion diagnosis of industrial control valves without interfering with the normal production of industrial fluids, using an Internet of Things control system. Summary of the Invention

[0006] This invention overcomes the shortcomings of the prior art and provides a method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things (IoT) control system.

[0007] To achieve the above objectives, the technical solution adopted by this invention is: a method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things (IoT) control system, wherein the IoT control system includes an edge computing gateway, a cloud diagnostic platform, and sensor nodes, and the method includes:

[0008] S1. The edge computing gateway collects low-frequency operating data at a first frequency and uploads it to the cloud diagnostic platform. The low-frequency operating data includes valve inlet pressure, valve outlet pressure, set opening degree, feedback opening degree, and set flow rate.

[0009] S2. When the cloud-based diagnostic platform determines that the health status is abnormal and, based on the low-frequency operating data, determines that the pressure fluctuation and flow deviation meet the steady-state conditions, it issues a bipolar micro-pulse diagnostic command.

[0010] S3. After parsing the bipolar micropulse diagnostic command, the edge computing gateway first wakes up the acoustic emission sensor and vibration sensor in the sensor node to a second frequency higher than the first frequency, and establishes a high-frequency data ring buffer locally.

[0011] S4. After the high-frequency data ring buffer stabilizes, the edge computing gateway controls the positioner of the industrial regulating valve to execute the bipolar micropulse diagnostic command, superimposing positive and negative displacements based on the current steady-state opening, and the duration of a single pulse is less than the volumetric delay time constant of the process fluid.

[0012] S5. The edge computing gateway uses the instruction execution time as a reference point to backtrack and extract the acoustic emission high-frequency mutation characteristics and vibration envelope characteristics of the valve stem of the industrial control valve at the critical point of transition from static friction to dynamic friction from the high-frequency data ring buffer, and uploads them together with the forward and reverse dead zone parameters extracted during micropulse execution to the cloud diagnostic platform.

[0013] S6. The cloud-based diagnostic platform diagnoses the type of mechanical degradation and issues a dead zone compensation update command accordingly.

[0014] S7. The edge computing gateway adaptively updates the dead zone compensation parameters of the locator according to the dead zone compensation update instruction.

[0015] In a preferred embodiment of the present invention, when the cloud-based diagnostic platform determines that the health status is abnormal and, based on the low-frequency operating data, determines that the pressure fluctuation and flow deviation meet the steady-state conditions, it issues a bipolar micro-pulse diagnostic command, including:

[0016] The low-frequency operating data is input into the dynamic health assessment model to calculate the current dynamic health index;

[0017] When the current dynamic health index is lower than a preset threshold, the time derivative of the valve pressure difference is calculated in real time. The valve pressure difference is the difference between the pressure before the valve and the pressure after the valve. The current estimated flow rate is calculated based on the feedback opening, the pressure before the valve, and the pressure after the valve.

[0018] When the absolute value of the time derivative of the valve pressure difference is less than the pressure fluctuation threshold within multiple consecutive control cycles, and the integral of the deviation between the set flow rate and the current estimated flow rate is less than the process allowable error, the steady-state condition is determined to be met and the bipolar micropulse diagnostic command is issued.

[0019] In a preferred embodiment of the present invention, the step of inputting the low-frequency operating data into the dynamic health assessment model to calculate the current dynamic health index includes:

[0020] The low-frequency operating data is subjected to time series alignment and normalization.

[0021] The processed data is input into the dynamic health assessment model, which includes a long short-term memory network layer and a spatial attention mechanism layer.

[0022] The nonlinear coupling weight between the valve differential pressure and the feedback opening is extracted through the spatial attention mechanism layer, and the current dynamic health index, ranging from 0% to 100%, is output.

[0023] In a preferred embodiment of the present invention, establishing a high-frequency data circular buffer locally includes:

[0024] Calculate the hardware power-on initialization delay of the acoustic emission sensor and the vibration sensor;

[0025] After the hardware power-on initialization delay ends, the second frequency is started for analog-to-digital conversion;

[0026] The converted high-frequency digital signal is continuously written into the high-frequency data ring buffer according to the first-in-first-out principle. When the buffer is full, the new sampled data automatically overwrites the oldest data to ensure that the buffer always stores high-frequency historical data for a set period of time before the trigger time.

[0027] In a preferred embodiment of the present invention, the amplitude range of the positive displacement in the bipolar micropulse diagnostic command is 0.5% to 1.5% of the current steady-state opening, and the amplitude of the reverse displacement is equal to the amplitude of the positive displacement but opposite in direction.

[0028] The duration of a single pulse ranges from 0.1 to 0.5 seconds, and the volumetric delay time constant of the process fluid ranges from 2 to 10 seconds.

[0029] In a preferred embodiment of the present invention, the step of backtracking and extracting the acoustic emission high-frequency abrupt change characteristics and vibration envelope characteristics of the valve stem at the critical point of transition from static friction to dynamic friction from the high-frequency data annular buffer includes:

[0030] The trigger reference origin is the moment when the execution command is sent to the locator;

[0031] The data sequence from a first predetermined time period before the trigger reference origin to a second predetermined time period after the trigger reference origin is extracted from the high-frequency data ring buffer as high-frequency multimodal data;

[0032] The moment when the feedback opening first changes displacement in the high-frequency multimodal data is located as the critical point of the transition from static friction to dynamic friction, and the peak value of the acoustic emission signal energy at the critical point of the transition is taken as the high-frequency abrupt change feature of the acoustic emission.

[0033] In a preferred embodiment of the present invention, the cloud-based diagnostic platform diagnoses the type of mechanical degradation and issues a dead zone compensation update command accordingly, and the edge computing gateway adaptively updates the dead zone compensation parameters of the locator according to the dead zone compensation update command, including:

[0034] The cloud-based diagnostic platform integrates various features and parameters into a multi-dimensional fault vector and calculates the cosine similarity between it and the pre-stored packing wear feature nodes, valve core cavitation feature nodes, and insufficient air source pressure feature nodes in the fault map.

[0035] The cloud-based diagnostic platform identifies the feature node with the highest cosine similarity and greater than the preset matching threshold as the confirmed mechanical degradation type, and generates the dead zone compensation update instruction containing the increase in dead zone parameters.

[0036] The edge computing gateway dynamically adjusts the dead zone compensation threshold of the locator according to the increase in the dead zone parameter, and superimposes a high-frequency micro-amplitude jitter signal to overcome valve stem jamming.

[0037] A remote monitoring and self-diagnostic system for industrial control valves based on an Internet of Things (IoT) control system includes:

[0038] IoT sensor nodes are deployed on the side of industrial control valves to collect low-frequency operating data and high-frequency multimodal data;

[0039] An edge computing gateway is communicatively connected to the IoT sensor node and is used to collect the low-frequency operating data and perform high-frequency data ring caching, feature backtracking and interception, locator control, and update dead zone compensation parameters.

[0040] The cloud-based diagnostic platform communicates with the edge computing gateway and is used to assess health status, determine steady-state conditions, issue bipolar micropulse diagnostic commands, diagnose mechanical degradation types, and issue dead zone compensation update commands.

[0041] In a preferred embodiment of the present invention, the Internet of Things sensor node includes a low-frequency process sensor and a high-frequency sensor containing an acoustic emission sensor and a vibration sensor;

[0042] The acoustic emission sensor is rigidly fixed to the outer wall of the valve body of the industrial control valve by a waveguide rod;

[0043] The vibration sensor is installed at the connection between the valve stem of the industrial control valve and the mechanical feedback rod of the positioner.

[0044] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things (IoT) control system.

[0045] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0046] This invention, through a cloud-based diagnostic platform, issues bipolar micro-pulse diagnostic commands after the system enters the steady-state diagnostic window. An edge computing gateway controls the positioner to superimpose forward and reverse displacements based on the current steady-state opening, while strictly limiting the duration of a single pulse to less than the volumetric delay time constant of the process fluid. It cleverly utilizes the physical inertia and volumetric effect of the fluid pipeline network as a natural low-pass filter, ensuring that transient flow fluctuations caused by minute forward and reverse displacements are completely and smoothly canceled out within the macroscopic pipeline network. This achieves non-destructive online active micro-motion testing of industrial control valves during continuous production. Compared to the inherent defects of conventional step tests in existing technologies, which easily lead to drastic pressure fluctuations in pipelines and even shutdowns, this invention not only ensures the absolute stability and production safety of industrial fluid processes but also breaks through the time and space limitations of traditional offline pressure testing, greatly improving the routine monitoring capabilities of the IoT control system for underlying execution nodes.

[0047] This invention, through an edge computing gateway, awakens the high-frequency sensor and establishes a local high-frequency data ring buffer before the positioner operates, after parsing the micropulse command. Then, it performs time window backtracking based on the command execution time. This hardware and software collaborative timing pre-triggering mechanism completely eliminates the hardware power-on initialization delay blind spot of the sensor analog-to-digital converter, ensuring that the ring buffer always completely encapsulates the high-frequency historical data before and after the valve operation. This allows for lossless capture of the acoustic emission high-frequency abrupt change characteristics and vibration envelope characteristics within the extremely short few milliseconds of the valve stem transitioning from static friction to dynamic friction. This overcomes the technical bottleneck of existing technologies that only trigger sampling when the operation occurs, thus perfectly missing the most critical static friction critical characteristics. It significantly improves the system's diagnostic accuracy for hidden mechanical degradation such as minor packing wear and early cavitation, providing the most basic and realistic physical data support for predictive maintenance of industrial equipment.

[0048] This invention relies on an architecture that integrates an edge computing gateway and a cloud-based diagnostic platform. Normally, it only uploads low-frequency operational data at a primary frequency. High-frequency sensors are only activated within the steady-state diagnostic window. After high-frequency feature extraction and capture at the edge, lightweight feature vectors and dead-zone parameters are uploaded to the cloud. This mechanism offloads the processing load of massive amounts of raw high-frequency waveform data to the industrial edge, transmitting only high-value feature results across networks via IoT protocols. This significantly frees up communication bandwidth for industrial IoT and reduces the concurrent computing power consumption of cloud servers. It effectively avoids the network congestion and system paralysis risks caused by real-time full cloud uploads of multimodal high-frequency sensors in pursuit of high-precision diagnostics, as is common in existing technologies. This allows for large-scale engineering deployment of IoT control systems in large chemical plants or refineries with hundreds or thousands of valve nodes, with extremely low hardware deployment costs and network overhead.

[0049] This invention utilizes a cloud-based diagnostic platform to diagnose mechanical degradation types based on cosine similarity matching between multidimensional fault vectors and fault maps. An edge computing gateway then dynamically adjusts the dead-zone compensation threshold of the positioner and superimposes high-frequency micro-amplitude jitter signals accordingly. This deeply integrates the global AI diagnostic computing power of the cloud with the underlying proportional-integral-derivative (PID) control algorithm at the edge. For specific diagnosed physical faults, parameter-level corrections are performed directly within the actuator, achieving a fully automated closed loop from precise fault perception to adaptive control strategy repair. This completely changes the outdated situation of traditional IoT monitoring systems that only monitor but do not control, heavily relying on manual on-site inspections and adjustments. It effectively extends the fault-free operating cycle of industrial control valves under harsh conditions, significantly reduces unplanned downtime and overall maintenance costs for enterprises, and truly endows industrial IoT control systems with intelligent attributes for autonomous health management. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of a remote monitoring and self-diagnosis method for industrial control valves based on an Internet of Things control system according to the present invention.

[0052] Figure 2 This is a schematic diagram of the bipolar micropulse timing and high-frequency feature extraction logic of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0055] Application Overview:

[0056] In the continuous production networks of modern process industries, industrial control valves, as the lowest-level physical execution node of fluid control, directly determine the safety boundary and product yield of the entire industrial control system due to their mechanical health status. Under long-term, harsh operating conditions, industrial control valves inevitably undergo mechanical degradation. The two most typical and insidious degradation mechanisms are the nonlinear abrupt change in static friction force caused by valve stem packing wear, and the physical peeling damage to the valve core surface caused by cavitation resulting from fluid throttling.

[0057] To address the aforementioned degradation mechanism, existing IoT control systems face irreconcilable core technological contradictions when conducting online monitoring. On the one hand, purely passive IoT monitoring relies heavily on threshold comparisons of low-frequency operating data. This approach is not only severely lagging but also highly susceptible to interference from macroscopic turbulent noise in the field fluid, making it impossible to determine the dead zone and actual frictional changes of industrial control valves. On the other hand, while traditional offline pressure testing or conventional step signal testing can acquire dynamic frictional characteristics, in continuous fluid processes, sending step test signals to industrial control valves can cause severe transient fluctuations in flow and pressure within the fluid pipeline network, posing a high risk of disrupting normal production or even causing shutdown accidents, thus preventing them from being routinely executed without shutting down the system.

[0058] To completely overcome the dual technical biases of uninterrupted continuous production and the easy loss of weak mechanical transient characteristics, this application proposes a groundbreaking method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things (IoT) control system. The core technical logic of this invention lies in cleverly utilizing the physical inertia of the fluid pipeline network itself as a natural low-pass filter. By issuing extremely short-duration bipolar micro-pulse diagnostic commands, transient flow fluctuations caused by minute forward and reverse displacements are completely and smoothly canceled out in the macroscopic pipeline network. Simultaneously, it innovatively introduces a sensor pre-wake-up and high-frequency data ring buffer mechanism at the edge computing gateway, completely eliminating the time delay blind spot of traditional analog-to-digital converter hardware power-on initialization. This enables lossless interception of microsecond-level high-frequency abrupt changes in acoustic emission characteristics at the critical point of valve stem transition from static friction to dynamic friction. Without increasing the on-site communication bandwidth load, it achieves a fully automated closed-loop diagnosis from underlying physical sensing and edge high-frequency interception to cloud-based intelligent diagnosis.

[0059] Example 1:

[0060] like Figure 1 As shown, a method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things (IoT) control system is disclosed. The IoT control system includes an edge computing gateway, a cloud diagnostic platform, and sensor nodes. The method includes:

[0061] S1. The edge computing gateway collects low-frequency operating data at a first frequency and uploads it to the cloud diagnostic platform. The low-frequency operating data includes valve inlet pressure, valve outlet pressure, set opening degree, feedback opening degree, and set flow rate.

[0062] S2. When the cloud-based diagnostic platform determines that the health status is abnormal and, based on the low-frequency operating data, determines that the pressure fluctuation and flow deviation meet the steady-state conditions, it issues a bipolar micro-pulse diagnostic command.

[0063] S3. After parsing the bipolar micropulse diagnostic command, the edge computing gateway first wakes up the acoustic emission sensor and vibration sensor in the sensor node to a second frequency higher than the first frequency, and establishes a high-frequency data ring buffer locally.

[0064] S4. After the high-frequency data ring buffer stabilizes, the edge computing gateway controls the positioner of the industrial regulating valve to execute the bipolar micropulse diagnostic command, superimposing positive and negative displacements based on the current steady-state opening, and the duration of a single pulse is less than the volumetric delay time constant of the process fluid.

[0065] S5. The edge computing gateway uses the instruction execution time as a reference point to backtrack and extract the acoustic emission high-frequency mutation characteristics and vibration envelope characteristics of the valve stem of the industrial control valve at the critical point of transition from static friction to dynamic friction from the high-frequency data ring buffer, and uploads them together with the forward and reverse dead zone parameters extracted during micropulse execution to the cloud diagnostic platform.

[0066] S6. The cloud-based diagnostic platform diagnoses the type of mechanical degradation and issues a dead zone compensation update command accordingly.

[0067] S7. The edge computing gateway adaptively updates the dead zone compensation parameters of the locator according to the dead zone compensation update instruction.

[0068] In translating the above theoretical methods into practical engineering applications, two major technological hurdles are encountered: first, obtaining valve stem displacement sufficient to overcome static friction without causing macroscopic process flow fluctuations; and second, capturing extremely short-duration acoustic emission abrupt signals. This embodiment overcomes these hurdles through the following collaborative execution logic:

[0069] First, there is system-level time synchronization and stringent steady-state determination: In industrial settings, there is a huge difference in sampling rate between low-frequency operating data and high-frequency multimodal data. In this embodiment, the edge computing gateway has a built-in high-precision hardware real-time clock and maintains microsecond-level time synchronization with the cloud diagnostic platform through a precise time protocol.

[0070] In establishing steady-state conditions, the cloud-based diagnostic platform does not use a simple static threshold. Instead, it performs sliding window filtering calculations on the valve pressure difference ΔP over multiple consecutive control cycles using the time derivative d(ΔP) / dt. Only when the absolute value of the time derivative d(ΔP) / dt is continuously less than the pressure fluctuation threshold within a preset time window, preferably 0.5% to 2% of the full scale of the upstream pressure sensor, and when the integral of the deviation between the set flow rate and the current estimated flow rate approaches zero within the preset time window, does the system strictly determine that the current fluid pipeline has entered a macroscopic fluid dynamic steady state, thereby unlocking and issuing a bipolar micropulse diagnostic command.

[0071] The length of the preset time window is set based on the dynamic response characteristics of the fluid pipeline network. Its value must be greater than the volumetric delay time constant of the process fluid. In this embodiment, it is preferably 3 to 10 consecutive control cycles. This setting is based on ensuring that the macroscopic transient turbulence in the fluid pipeline network has been completely smoothed, thereby providing an absolutely stable fluid dynamic background for subsequent micro-motion diagnosis.

[0072] During the execution of the bipolar micropulse diagnostic command, the micropulse command takes the current steady-state opening as a reference, first superimposes a positive displacement with a value of +A, then superimposes a reverse displacement with a value of -A in a very short time without interruption, and finally returns to the current steady-state opening.

[0073] This embodiment strictly limits the duration Δt of a single pulse to be less than the volumetric delay time constant T of the current process fluid. i Therefore, due to T i Much greater than Δt, the entire fluid pipeline network and downstream storage tanks physically function as a giant low-pass filter. The forward and reverse mechanical micro-movements of the positioner of the industrial control valve in a very short time, although sufficient to expose the mechanical dead zone of the positioner and the resistance characteristics of the valve stem to overcome static friction at the micro level, the resulting increase in forward flow and decrease in reverse flow are completely and smoothly offset by the physical compressibility of the fluid and the volumetric inertia of the pipeline in the macroscopic pipeline network, achieving zero net disturbance to the quality of downstream process products.

[0074] Given the extremely rapid and brief release of acoustic emission energy at the critical point of the valve stem's transition from static friction to dynamic friction, if the analog-to-digital converter (ADC) is triggered to sample at the same time as the action command is issued, the ADC's hardware power-on initialization delay will inevitably miss this crucial energy peak.

[0075] To address this issue, this embodiment introduces a pre-wake-up and high-frequency data ring buffer mechanism. Before the edge computing gateway parses the bipolar micropulse diagnostic command but before driving the locator to execute, it first calculates the hardware power-on initialization delay of the acoustic emission sensor and the vibration sensor. This delay is obtained by reading the factory configuration parameters of the sensor's internal firmware or the system power-on self-test time.

[0076] After the hardware power-on initialization delay ends, the second frequency is started to perform analog-to-digital conversion. The converted high-frequency digital signal is continuously written to the high-frequency data ring buffer of the edge computing gateway according to the first-in-first-out principle. When the buffer is full, the new sampled data automatically overwrites the oldest data to ensure that the buffer always stores high-frequency historical data for a set period of time before the trigger time.

[0077] When the locator actually starts executing the bipolar micro-pulse diagnostic command, the edge computing gateway marks this moment as the trigger reference origin. At this time, the edge computing gateway does not start recording from this trigger reference origin backward, but uses the characteristics of the high-frequency data ring buffer to perform negative delay backtracking interception, that is, extracts the data sequence from the first set time before the trigger reference origin to the second set time after the trigger reference origin as high-frequency multimodal data.

[0078] In the high-frequency multimodal data, the system calculates the first derivative of the feedback opening sequence to locate the moment when the feedback opening first changes displacement as the critical point of the transition from static friction to dynamic friction, and extracts the peak energy of the acoustic emission signal at the critical point of the transition as the high-frequency abrupt change feature of the acoustic emission.

[0079] Meanwhile, for the high-frequency vibration data sequence in the extracted high-frequency multimodal data, the edge computing gateway first extracts its vibration envelope time-domain signal through Hilbert transform, then uses fast Fourier transform to convert the time-domain envelope signal to the frequency domain, and extracts the extreme point with the largest amplitude in the frequency domain spectrum as the frequency domain main peak amplitude F of the vibration envelope feature. peak .

[0080] Furthermore, the logic for extracting the forward and reverse dead-time parameters during micro-pulse execution is as follows:

[0081] The edge computing gateway records the time difference between the issuance of the forward displacement command and the first positive change in the feedback opening, and the time difference between the issuance of the reverse displacement command and the first reverse change in the feedback opening.

[0082] Based on the pressure parameters of the current process fluid, the time difference is converted into the positive dead zone displacement deviation and the reverse dead zone displacement deviation caused by static friction and mechanical clearance of the valve through the dynamic equation of the pneumatic amplifier of the positioner, and used as the positive and reverse dead zone parameters.

[0083] The specific conversion formula is as follows:

[0084] ;

[0085] Where ΔD is the dead zone displacement deviation, Δt is the time difference, and k act P is the equivalent mechanical gain coefficient of the pneumatic actuator. s For the system gas source pressure, P c This refers to the real-time chamber pressure output from the pneumatic amplifier to the cylinder.

[0086] This combination of memory-level time-axis translation, differentiation algorithms, and multimodal digital signal processing fundamentally enables the lossless capture of weak high-frequency friction features.

[0087] Example 2:

[0088] Based on Example 1, this embodiment further expands and refines the mathematical calculation process of the artificial intelligence algorithm model of the cloud diagnostic platform and the closed-loop control logic of the edge computing gateway.

[0089] When the cloud-based diagnostic platform determines that the health status is abnormal, its first step is to input the low-frequency operating data into the dynamic health assessment model to calculate the current dynamic health index.

[0090] During the offline training phase of the model, the system pre-collects a sample set containing historical maintenance records and multi-source heterogeneous sensor data;

[0091] Extract the packing thickness wear and valve core mass loss rate from the historical maintenance records of industrial control valves during each major overhaul, and map them to continuous values ​​between 0 and 100 as the true labels for model training (where 100 represents brand new with no wear, and 0 represents reaching the limit of scrapping).

[0092] Then, the backpropagation algorithm is used to minimize the mean square error between the predicted health index and the true label, thereby solidifying the node weights within the model.

[0093] During the online inference phase, the system first performs time series alignment and normalization on the low-frequency operating data. The time series alignment is based on the sampling timestamp of the feedback opening, and the pressure before the valve, the pressure after the valve, and the set flow rate are resampled and aligned using a zero-order hold or linear interpolation.

[0094] The numerical differences between different units are eliminated, and then the processed data is input into the dynamic health assessment model that includes a long short-term memory network layer and a spatial attention mechanism layer;

[0095] Specifically, the preferred network topology and hyperparameters of the dynamic health assessment model are as follows:

[0096] The model input layer receives low-frequency running data sequences with a fixed time step.

[0097] The Long Short-Term Memory (LSTM) network layer is preferably a two-layer network structure, with each hidden layer containing 64 to 128 neuron nodes for deep extraction of temporal hidden states;

[0098] The dimension of the fully connected mapping layer used to generate query vectors, key vectors, and value vectors in the spatial attention mechanism layer is set to 64 dimensions.

[0099] During the offline training phase of the model, the Adam optimizer is used for gradient updates, with the initial learning rate set to 0.001 and the batch size set to 32. An early stopping mechanism is used to prevent overfitting of the model until the mean squared error loss function converges.

[0100] Long Short-Term Memory (LSTM) network layers calculate the long-term dependency sequence features of low-frequency running data over time through internal forget gates, input gates, and output gates;

[0101] The spatial attention mechanism layer maps the time series of the valve pressure difference to a query vector and the time series of the feedback opening to a key vector and a value vector, respectively. By calculating the dot product of the query vector and the key vector, the influence of small fluctuations in the pressure difference on the actual displacement of the valve is dynamically evaluated, thereby generating a nonlinear coupling weight matrix at that moment. This nonlinear coupling weight matrix enables the model to automatically focus on the moment when there is a small mismatch between the pressure difference and the feedback opening.

[0102] The nonlinear coupling weight matrix is ​​multiplied by the value vector to obtain a weighted feature representation, which is then input into a global average pooling layer for dimensionality reduction. Finally, the dimensionality-reduced features are mapped to probability values ​​between 0 and 1 through a fully connected layer with a Sigmoid activation function, and multiplied by 100 to output the current dynamic health index, which is in the range of 0% to 100%.

[0103] When the current dynamic health index is lower than a preset threshold, in order to provide double protection for steady-state determination, the cloud-based diagnostic platform calculates the time derivative of the valve differential pressure in real time, and calculates the current estimated flow rate Q based on the feedback opening, the upstream pressure P1, and the downstream pressure P2, combined with the industry-standard fluid characteristic equation. The specific calculation formula is as follows:

[0104] ;

[0105] Where C v ρ is the inherent flow capacity coefficient of the valve body, and ρ is the fluid density.

[0106] During the stage of diagnosing the type of mechanical degradation, the cloud-based diagnostic platform fuses various features and parameters received from the edge computing gateway into a multi-dimensional fault vector V. The specific dimensions of this multi-dimensional fault vector strictly include the energy peak E of the high-frequency abrupt change characteristics of acoustic emission. ae The amplitude F of the main peak in the frequency domain of the vibration envelope characteristics peak And the difference ΔD between the forward and reverse dead zone parameters, i.e. .

[0107] The cloud-based diagnostic platform calculates the cosine similarity between the multidimensional fault vector and the pre-stored feature node vector U in the fault map. This feature node vector includes packing wear feature nodes, valve core cavitation feature nodes, and insufficient air source pressure feature nodes. The specific process for establishing the pre-stored feature node vector U in the fault map is as follows:

[0108] 1) During the system initialization or offline deployment phase, extract a large amount of historical high-frequency multimodal data and low-frequency control parameters of industrial control valves with known typical fault types;

[0109] 2) Following the aforementioned feature extraction logic, convert it into a historical multidimensional fault vector sample set containing real physical labels;

[0110] 3) The K-Means clustering algorithm is used to perform spatial clustering on the sample sets of various faults, and the statistical mean vector of each type of fault sample is calculated and extracted as its cluster center;

[0111] 4) The cluster center vectors of each fault type are stored in the cloud database as the pre-stored feature node vector U.

[0112] The calculation logic for cosine similarity is as follows:

[0113] The ratio of the inner product of two multidimensional space vectors to the product of their magnitudes is given by the following formula: ;

[0114] The closer the ratio is to a certain value, the smaller the angle between the vectors, and the more consistent the physical features. The system determines the feature node with the highest cosine similarity and greater than the preset matching threshold as the confirmed mechanical degradation type, and generates the dead zone compensation update instruction containing the increase in the dead zone parameter.

[0115] After the diagnosis is completed, the edge computing gateway dynamically adjusts the dead zone compensation threshold of the locator according to the increase in the dead zone parameter. The specific adjustment logic is to add the product of the increase in the dead zone parameter and the preset proportional coefficient to the old dead zone compensation threshold to form a new dead zone compensation threshold. The preset proportional coefficient is preferably set between 0.1 and 0.3. The physical basis is to adopt an underdamped iterative compensation strategy to prevent the closed-loop control system from oscillating due to excessive dead zone compensation in a single operation.

[0116] An edge computing gateway superimposes a high-frequency micro-amplitude jitter signal to overcome valve stem jamming. The amplitude of this high-frequency micro-amplitude jitter signal is strictly limited within the mechanical dead zone of the positioner. At the same time, the high-frequency micro-amplitude jitter signal adopts a continuous sine waveform, and its frequency setting range is between 10 Hz and 30 Hz, which avoids the first-order natural resonance frequency of the valve's mechanical structure. Its physical principle is to keep the valve stem in a state of micro-friction through tiny high-frequency oscillations, thereby completely eliminating control overshoot caused by sudden changes in static friction.

[0117] Example 3:

[0118] This embodiment provides a remote monitoring and self-diagnostic system for industrial control valves based on an Internet of Things (IoT) control system, applicable to the methods described in Embodiments 1 and 2 above. The system's hardware topology includes IoT sensor nodes, an edge computing gateway, and a cloud-based diagnostic platform.

[0119] The IoT sensor nodes include low-frequency process sensors and high-frequency sensors containing acoustic emission sensors and vibration sensors. In order to ensure that the acquired physical signals have a high signal-to-noise ratio, this embodiment has carried out a strict mechanical conduction design for the installation position of the high-frequency sensors.

[0120] Preferably, the acoustic emission sensor is rigidly fixed to the outer wall of the industrial control valve body via a waveguide rod. The waveguide rod is made of a metal alloy with acoustic impedance matching. Its physical function is to transmit high-frequency stress waves inside the valve to the external sensor without damage, while using the inherent resonant frequency characteristics of the waveguide rod to naturally filter out interference from low-frequency mechanical vibrations in the industrial environment.

[0121] Preferably, the vibration sensor is installed at the connection between the valve stem of the industrial control valve and the mechanical feedback rod of the positioner. This position is the direct force point where the positioner senses the actual physical displacement of the valve stem, and can most sensitively capture the nonlinear mechanical vibration impact caused by the mechanical dead zone.

[0122] The edge computing gateway is responsible for performing high-frequency data circular caching, feature backtracking and interception, locator control, and updating dead zone compensation parameters. Its hardware computing power offloading logic is as follows:

[0123] The edge computing gateway operates in low-power mode most of the time, processing only low-frequency process sensor data. It only runs the high-frequency analog-to-digital converter at full power when steady-state conditions are met and a bipolar micro-pulse diagnostic command is received.

[0124] After extracting the high-frequency abrupt change features and vibration envelope features of acoustic emission, the edge computing gateway transmits only a very small amount of multidimensional fault vectors across the network through the Internet of Things protocol, fundamentally avoiding the risk of industrial IoT paralysis caused by continuously uploading massive amounts of high-frequency raw waveforms to the cloud.

[0125] Example 4:

[0126] To further demonstrate the full disclosure and industrial applicability of the present invention, this embodiment takes the liquid level control fluid pipeline network of a large petrochemical refinery as an example to conduct a specific numerical scenario simulation of the entire monitoring and self-diagnosis method.

[0127] In this scenario, assume the volumetric delay time constant T of the process fluid in the fluid pipeline network. iIt was determined that the edge computing gateway continuously uploads the valve inlet pressure, valve outlet pressure, and feedback opening to the cloud diagnostic platform at a first frequency of 1 Hz every 5 seconds. One day, the cloud diagnostic platform calculated through the dynamic health assessment model that the current dynamic health index had fallen below the preset threshold of 80%, indicating an abnormal health status.

[0128] The system then initiates steady-state determination logic. If the absolute value of the time derivative of the valve pressure difference d(ΔP) / dt is continuously less than the pressure fluctuation threshold within ten consecutive control cycles, and the integral of the deviation between the set flow rate and the current estimated flow rate approaches zero, the system is determined to have entered steady-state conditions. The current steady-state opening is 50%.

[0129] The cloud-based diagnostic platform issues bipolar micropulse diagnostic commands. The positive displacement amplitude in the command is set to one-hundredth of the current steady-state opening, and the reverse displacement amplitude is equal to the positive displacement but opposite in direction.

[0130] After parsing the instructions, the edge computing gateway calculates that the hardware power-on initialization delay is 10 milliseconds. After the delay ends, it wakes up the acoustic emission sensor and vibration sensor in advance, and writes the high-frequency digital signal into the high-frequency data ring buffer at a second frequency of 50,000 Hz.

[0131] Subsequently, the edge computing gateway controls the positioner of the industrial control valve to perform an action, based on the current steady-state opening of 50%, first adding a positive displacement to reach 51%, and then adding a reverse displacement back to 50%.

[0132] The duration Δt of a single pulse was set to 0.2 seconds. Since 0.2 seconds is much smaller than the volumetric delay time constant of 5 seconds, the small flow fluctuations caused by forward and reverse displacements are physically smoothed in the pipeline. No abnormal fluctuations were detected by the flow meters in the macro network, and the zero net disturbance test was perfectly achieved.

[0133] During the micro-pulse execution, the edge computing gateway takes the moment when it sends the execution command to the locator as the trigger reference origin, backtracks forward by 50 milliseconds for a first set duration and forward by 150 milliseconds for a second set duration as high-frequency multimodal data.

[0134] The system successfully located the moment when the feedback opening first changed displacement by differentiating the data, and extracted the high-frequency abrupt change characteristic energy peak of acoustic emission that lasted only 3 milliseconds at that moment. The edge computing gateway uploaded this energy peak along with the extracted 0.8 percent of the forward and reverse dead zone parameters to the cloud.

[0135] The cloud-based diagnostic platform integrates the above data into a multi-dimensional fault vector. The cosine similarity calculation using the cosine similarity formula reveals that the cosine similarity between the vector and the pre-stored packing wear feature nodes is as high as 0.95, which is greater than the preset matching threshold of 0.85. Thus, the mechanical degradation type is diagnosed as packing wear.

[0136] Finally, the cloud-based diagnostic platform issued a dead zone compensation update command containing the increase in dead zone parameters. Based on this, the edge computing gateway adaptively updated the dead zone compensation threshold of the positioner and superimposed a high-frequency micro-amplitude jitter signal, successfully eliminating the potential jamming risk caused by the increase in valve stem static friction and ensuring the absolute safe operation of the refinery's continuous fluid process.

[0137] In summary, the present invention provides a remote monitoring and self-diagnosis method and system for industrial control valves based on an IoT control system. It cleverly integrates the fluid dynamics inertial filtering effect with an edge-cloud collaborative architecture. It not only achieves zero-net-disturbance online active detection of continuous fluid processes by strictly limiting the duration of a single pulse of the bipolar micropulse diagnostic command to be less than the volumetric delay time constant of the process fluid, but also completely overcomes the technical gap of hardware power-on initialization delay through an innovative sensor pre-wake-up and high-frequency data ring buffer backtracking interception mechanism, achieving lossless capture of microsecond-level acoustic emission high-frequency mutation characteristics. Combined with the AI-powered multi-dimensional fault diagnosis of the cloud-based diagnostic platform and the dead-zone compensation adaptive closed-loop control of the edge computing gateway, this invention truly endows the underlying physical execution nodes of industrial fluid pipelines with autonomous health management capabilities with extremely low network communication bandwidth overhead. This comprehensively improves the operational stability and economic benefits of modern large-scale industrial control systems and possesses extremely high engineering promotion value.

[0138] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things (IoT) control system, characterized in that, The IoT control system includes an edge computing gateway, a cloud diagnostic platform, and sensor nodes; the method includes: S1. The edge computing gateway collects low-frequency operating data at a first frequency and uploads it to the cloud diagnostic platform. The low-frequency operating data includes valve inlet pressure, valve outlet pressure, set opening degree, feedback opening degree, and set flow rate. S2. When the cloud-based diagnostic platform determines that the health status is abnormal and, based on the low-frequency operating data, determines that the pressure fluctuation and flow deviation meet the steady-state conditions, it issues a bipolar micro-pulse diagnostic command. S3. After parsing the bipolar micropulse diagnostic command, the edge computing gateway first wakes up the acoustic emission sensor and vibration sensor in the sensor node to a second frequency higher than the first frequency, and establishes a high-frequency data ring buffer locally. S4. After the high-frequency data ring buffer stabilizes, the edge computing gateway controls the positioner of the industrial regulating valve to execute the bipolar micropulse diagnostic command, superimposing positive and negative displacements based on the current steady-state opening, and the duration of a single pulse is less than the volumetric delay time constant of the process fluid. S5. The edge computing gateway uses the instruction execution time as a reference point to backtrack and extract the acoustic emission high-frequency mutation characteristics and vibration envelope characteristics of the valve stem of the industrial control valve at the critical point of transition from static friction to dynamic friction from the high-frequency data ring buffer, and uploads them together with the forward and reverse dead zone parameters extracted during micropulse execution to the cloud diagnostic platform. S6. The cloud-based diagnostic platform diagnoses the type of mechanical degradation and issues a dead zone compensation update command accordingly. S7. The edge computing gateway adaptively updates the dead zone compensation parameters of the locator according to the dead zone compensation update instruction.

2. The method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things control system according to claim 1, characterized in that, When the cloud-based diagnostic platform determines that the health status is abnormal and, based on the low-frequency operating data, determines that the pressure fluctuations and flow deviations meet the steady-state conditions, it issues a bipolar micro-pulse diagnostic command, including: The low-frequency operating data is input into the dynamic health assessment model to calculate the current dynamic health index; When the current dynamic health index is lower than a preset threshold, the time derivative of the valve pressure difference is calculated in real time. The valve pressure difference is the difference between the pressure before the valve and the pressure after the valve. The current estimated flow rate is calculated based on the feedback opening, the pressure before the valve, and the pressure after the valve. When the absolute value of the time derivative of the valve pressure difference is less than the pressure fluctuation threshold within multiple consecutive control cycles, and the integral of the deviation between the set flow rate and the current estimated flow rate is less than the process allowable error, the steady-state condition is determined to be met and the bipolar micropulse diagnostic command is issued.

3. The method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things control system according to claim 2, characterized in that, The step of inputting the low-frequency operating data into the dynamic health assessment model to calculate the current dynamic health index includes: The low-frequency operating data is subjected to time series alignment and normalization. The processed data is input into the dynamic health assessment model, which includes a long short-term memory network layer and a spatial attention mechanism layer. The nonlinear coupling weight between the valve differential pressure and the feedback opening is extracted through the spatial attention mechanism layer, and the current dynamic health index, ranging from 0% to 100%, is output.

4. The method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things control system according to claim 1, characterized in that, The establishment of a local high-frequency data ring buffer includes: Calculate the hardware power-on initialization delay of the acoustic emission sensor and the vibration sensor; After the hardware power-on initialization delay ends, the second frequency is started for analog-to-digital conversion; The converted high-frequency digital signal is continuously written into the high-frequency data ring buffer according to the first-in-first-out principle. When the buffer is full, the new sampled data automatically overwrites the oldest data to ensure that the buffer always stores high-frequency historical data for a set period of time before the trigger time.

5. The method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things control system according to claim 1, characterized in that, The amplitude range of the positive displacement in the bipolar micropulse diagnostic command is 0.5% to 1.5% of the current steady-state opening, and the amplitude of the reverse displacement is equal to the amplitude of the positive displacement but opposite in direction. The duration of a single pulse ranges from 0.1 to 0.5 seconds, and the volumetric delay time constant of the process fluid ranges from 2 to 10 seconds.

6. The method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things control system according to claim 1, characterized in that, The acoustic emission high-frequency abrupt change characteristics and vibration envelope characteristics of the valve stem transitioning from static friction to dynamic friction by backtracking from the high-frequency data annular buffer include: The trigger reference origin is the moment when the execution command is sent to the locator; The data sequence from a first predetermined time period before the trigger reference origin to a second predetermined time period after the trigger reference origin is extracted from the high-frequency data ring buffer as high-frequency multimodal data; The moment when the feedback opening first changes displacement in the high-frequency multimodal data is located as the critical point of the transition from static friction to dynamic friction, and the peak value of the acoustic emission signal energy at the critical point of the transition is taken as the high-frequency abrupt change feature of the acoustic emission.

7. The method for remote monitoring and self-diagnosis of industrial control valves based on an Internet of Things control system according to claim 1, characterized in that, The cloud-based diagnostic platform diagnoses the type of mechanical degradation and issues a dead zone compensation update command accordingly. The edge computing gateway adaptively updates the dead zone compensation parameters of the locator based on the dead zone compensation update command, including: The cloud-based diagnostic platform integrates various features and parameters into a multi-dimensional fault vector and calculates the cosine similarity between it and the pre-stored packing wear feature nodes, valve core cavitation feature nodes, and insufficient air source pressure feature nodes in the fault map. The cloud-based diagnostic platform identifies the feature node with the highest cosine similarity and greater than the preset matching threshold as the confirmed mechanical degradation type, and generates the dead zone compensation update instruction containing the increase in dead zone parameters. The edge computing gateway dynamically adjusts the dead zone compensation threshold of the locator according to the increase in the dead zone parameter, and superimposes a high-frequency micro-amplitude jitter signal to overcome valve stem jamming.

8. A remote monitoring and self-diagnostic system for industrial control valves based on an Internet of Things (IoT) control system, applicable to the method described in any one of claims 1 to 7, characterized in that, include: IoT sensor nodes are deployed on the side of industrial control valves to collect low-frequency operating data and high-frequency multimodal data; An edge computing gateway is communicatively connected to the IoT sensor node and is used to collect the low-frequency operating data and perform high-frequency data ring caching, feature backtracking and interception, locator control, and update dead zone compensation parameters. The cloud-based diagnostic platform communicates with the edge computing gateway and is used to assess health status, determine steady-state conditions, issue bipolar micropulse diagnostic commands, diagnose mechanical degradation types, and issue dead zone compensation update commands.

9. A remote monitoring and self-diagnosis system for industrial control valves based on an Internet of Things control system as described in claim 8, characterized in that, The IoT sensor nodes include low-frequency process sensors and high-frequency sensors containing acoustic emission sensors and vibration sensors. The acoustic emission sensor is rigidly fixed to the outer wall of the valve body of the industrial control valve by a waveguide rod; The vibration sensor is installed at the connection between the valve stem of the industrial control valve and the mechanical feedback rod of the positioner.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for remote monitoring and self-diagnosis of industrial control valves based on the Internet of Things control system as described in any one of claims 1 to 7.

Images