A flow meter instrument fault detection system and method based on smart sensors

Abstract

The application belongs to the technical field of intelligent sensors, and discloses a flow meter instrument fault detection system and method based on an intelligent sensor, which comprises a sensor deployment module, which is used for setting first, second and third intelligent sensors on the signal output end, structural connection part and shell surface of the flow meter instrument respectively, and outputting corresponding multi-source sensing information; an abnormality marking screening module, which is used for generating electrical abnormality marking, vibration space abnormality marking and overheating area marking respectively according to the three types of intelligent sensors, and jointly determining whether to generate a sensing structure, and determining the priority of the sensing structure and the corresponding time sequence alignment range; and an instrument state analysis module, which is used for distributing the multi-source sensing information to the corresponding link of the work causal chain of each functional unit of the flow meter instrument according to the priority of the sensing structure, taking the work causal chain as a framework. The application improves the accuracy, interpretability and engineering application value of flow meter instrument fault detection.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensor technology, and more specifically, to a flow meter fault detection system and method based on intelligent sensors. Background Technology

[0002] Flow meters are widely used in industrial settings such as petroleum, chemical, power, metallurgy, and water treatment for real-time measurement of pipeline media flow. Because flow meters operate under complex conditions including vibration, temperature fluctuations, electromagnetic interference, and media erosion, their internal sensing elements, signal conditioning circuits, signal conversion circuits, and transmitter output circuits are prone to malfunctions such as structural loosening, zero-point drift, and abnormal output. Failure to promptly identify and locate the cause of these malfunctions can not only affect the accuracy of flow measurement but may also lead to deviations in production process control and even safety risks.

[0003] With the development of intelligent sensor technology, existing flow meter fault detection solutions typically deploy electrical sensors, vibration sensors, and temperature sensors on the instrument to monitor its operating status in multiple dimensions. By collecting information such as output signals, electrical parameters, structural vibration, and surface temperature, the ability to detect fault conditions can be improved to a certain extent.

[0004] However, existing technologies still have some problems in practical applications. Most solutions perform a uniform data alignment and fault analysis process on all sensor data in each acquisition cycle, regardless of whether any sensor detects obvious anomalies, and construct analysis data and initiate subsequent diagnostics in a fixed manner. The problem with this is that a large amount of normal data or data with only slight fluctuations will also enter the subsequent processing, resulting in a large system computational load and low processing efficiency.

[0005] Furthermore, existing technologies typically employ independent threshold triggering for each sensor to determine anomalies. When the data from a single sensor slightly exceeds the threshold, a complete fault analysis process is triggered. This approach does not adequately consider the correlation between electrical anomalies, vibration anomalies, and temperature anomalies, and is prone to false triggering due to occasional noise or short-term fluctuations, thus affecting the accuracy of fault detection results.

[0006] Upon detecting an anomaly, existing solutions typically default to using data from all sensors within a fixed time window for full time-series alignment, without distinguishing between the source and type of the anomaly. In reality, different types of faults require different ranges of analytical information. For example, electrical anomalies usually require comprehensive analysis of electrical, vibration, and temperature information, while some mechanical and thermal coupling anomalies can be diagnosed primarily based on vibration and temperature data. If all data is still used uniformly, it not only increases the overhead of data reading and time-series synchronization but also affects the system's real-time response capability.

[0007] In view of this, the present invention proposes a flow meter fault detection system and method based on intelligent sensors to solve the above problems. Summary of the Invention

[0008] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a flow meter fault detection system based on intelligent sensors, comprising: The sensor deployment module is used to install first-class, second-class, and third-class smart sensors at the signal output end, structural connection parts, and outer surface of the flow meter, respectively, and output corresponding multi-source sensing information. The anomaly labeling and filtering module is used to generate electrical anomaly labels, vibration space anomaly labels and overheating area labels based on three types of smart sensors, and to perform joint judgment to determine whether to generate a sensing structure, and to determine the priority of the sensing structure and the corresponding timing alignment range. The instrument status analysis module is used to allocate multi-source sensing information to the corresponding links of the working causal chain based on the priority of the sensing structure and the working causal chain of each functional unit of the flow meter, and to calculate the violation vector corresponding to each link. The fault criterion generation module is used to map the violation vector to a pre-built fault logic space and generate corresponding fault criteria and matching scores. The fault tracing and location module is used to construct a fault cause-effect graph based on the matching degree score, with each functional unit of the flow meter as a node and the signal transmission relationship of each functional unit as a directed edge. It performs reverse tracing and consistency verification in the fault cause-effect graph and outputs the fault root cause location result.

[0009] Preferably, the method for setting up the first, second, and third types of smart sensors respectively includes: After the flow meter is installed, the intelligent sensors are deployed according to the structure of the instrument. A first type of intelligent sensor is set at the signal output end of the flow meter. The first type of intelligent sensor is electrically connected to the output signal loop of the flow meter to collect output voltage, current and signal phase information, and output the corresponding electrical characteristic information. A second type of intelligent sensor is installed at the structural connection points of the flow meter, including the meter head, pipe connection flange, and transmitter housing. The second type of intelligent sensor is used to collect vibration information of each structural connection point and output the corresponding vibration status information. A third type of intelligent sensor is installed on the surface of the flow meter housing. The third type of intelligent sensor consists of an array structure composed of several temperature detection units, which collects temperature information at different locations on the surface of the flow meter housing and outputs the corresponding temperature field distribution information.

[0010] Preferably, the method for outputting the corresponding multi-source sensing information includes: The electrical characteristic information output by the first type of intelligent sensor, the vibration state information output by the second type of intelligent sensor, and the temperature field distribution information output by the third type of intelligent sensor are timestamped according to a unified time base. The various types of information after the timestamping are combined and encapsulated according to a preset data format to output the corresponding multi-source sensing information.

[0011] Preferably, the method for determining whether to generate a perceptual structure includes: The deviation between the current electrical characteristic information and the corresponding reference value is calculated based on the electrical characteristic information output by the first type of intelligent sensor. When the deviation is greater than the preset deviation threshold, an electrical anomaly mark is generated. The vibration spatial consistency index between the structural connection parts is calculated based on the vibration state information output by the second type of intelligent sensor. When the vibration spatial consistency index corresponding to any structural connection part is less than the preset consistency index threshold, a vibration spatial anomaly mark is generated. The overheated area is identified based on the temperature field distribution information output by the third type of intelligent sensor. When the identified overheated area is not empty, an overheated area mark is generated. A joint determination is performed based on electrical anomaly markers, vibration space anomaly markers, and overheating zone markers. When the electrical anomaly marker is in an abnormal state, or when the electrical anomaly marker is in a normal state and both the vibration space anomaly marker and the overheating zone marker are in an abnormal state, a sensing structure is determined to be generated.

[0012] Preferably, the method for determining the priority of the sensing structure and its corresponding timing alignment range includes: When generating a sensing structure, if the electrical anomaly is marked as an abnormal state, the sensing structure is determined as the first priority; if the electrical anomaly is marked as a normal state and the vibration space anomaly and the overheating area anomaly are both marked as abnormal states, the sensing structure is determined as the second priority. When the perception structure is the first priority, the multi-source perception information of the first type of intelligent sensor, the second type of intelligent sensor, and the third type of intelligent sensor within the preset time window corresponding to the target timestamp is called, and the complete time sequence alignment is performed according to the unified time reference. When the sensing structure is of the second priority, the multi-source sensing information of the second and third type of intelligent sensors within the preset time window corresponding to the target timestamp is called, and local timing alignment is performed according to a unified time reference, thereby determining the priority of the sensing structure and the corresponding timing alignment range.

[0013] Preferably, the method for calculating the violation vector corresponding to each stage includes: The range of multi-source sensing information to be analyzed is determined based on the priority identifier carried in the sensing structure. When the sensing structure is of the first priority, the electrical characteristic information, vibration state information, and temperature field distribution information after complete time-series alignment are called; when the sensing structure is of the second priority, the vibration state information and temperature field distribution information after local time-series alignment are called. Construct a causal chain based on the working process of the flow meter, including a sensing unit, a signal conditioning unit, a signal conversion unit, and a transmitter output unit; and allocate the multi-source sensing information to the corresponding links in the causal chain according to the correspondence between various multi-source sensing information and each functional unit. For each link in the causal chain of work, the multi-source sensing information assigned to that link is substituted into the pre-established operational constraint relationship of that link to calculate the difference between the actual observation results and the predicted results of the constraint relationship, and the various difference quantities are combined in a preset order to form the violation vector corresponding to the link.

[0014] Preferably, the method for generating the corresponding fault criteria and matching score includes: A fault logic space is pre-constructed, which includes several criteria entries corresponding to different fault modes of the flow meter. Each criterion entry includes three dimensions of judgment conditions: fault severity, fault evolution stage, and fault impact range. Receive the violation vectors corresponding to each link in the working causal chain, determine the severity of the fault based on the magnitude of each violation vector, determine the fault evolution stage based on the changing trend of the violation vectors in several consecutive acquisition times, and determine the scope of the fault impact based on the number of functional links whose violation vectors exceed the preset violation vector threshold and their distribution relationship in the working causal chain. The determined fault severity, fault evolution stage, and fault impact range are mapped to the fault logic space and matched with the judgment conditions of each criterion item. The matching degree score of each criterion item is calculated to generate the corresponding fault criterion and matching degree score.

[0015] Preferably, the method for constructing a fault cause-effect graph includes: The sensing unit, signal conditioning unit, signal conversion unit, and transmitter output unit of the flow meter are respectively identified as nodes in the fault cause-effect graph; according to the signal transmission relationship between each functional unit, a directed edge is established between the corresponding node of the previous functional unit and the corresponding node of the next functional unit. Receive the fault criteria and the corresponding matching score, and use the matching score as the global credibility coefficient; obtain the violation vector corresponding to each functional unit, calculate the magnitude of each violation vector, and combine the magnitude of each violation vector with the global credibility coefficient to obtain the activation value corresponding to each node; the nodes, directed edges and the activation values ​​corresponding to each node together constitute the fault cause-effect graph.

[0016] Preferably, the method for outputting the root cause localization result includes: Based on the activation values ​​of each node in the fault cause-effect graph, nodes with activation values ​​greater than a preset activation threshold are identified as symptom nodes. Starting from the symptom node, trace back level by level along the opposite direction of the directed edges in the fault cause-effect graph, and add upstream nodes with activation values ​​greater than the preset activation threshold to the root cause candidate set. For each candidate node in the root cause candidate set, the candidate node is defined as the root cause node. Forward deduction is performed along the direction of the directed edge in the fault cause graph to obtain the predicted abnormal distribution of each downstream node. The predicted abnormal distribution is then compared with the currently observed node activation status. When the comparison results meet the preset consistency conditions, the corresponding candidate node is determined to have passed the consistency verification. Based on the node activation value, fault criterion matching score, and consistency verification results of the candidate node that has passed the consistency verification, the root cause confidence score of each candidate node is calculated. The candidate node with the highest root cause confidence score is determined as the fault root cause node, and the fault root cause location result is output.

[0017] A method for fault detection of flow meters based on smart sensors, comprising: S1. First, second and third type intelligent sensors are respectively installed at the signal output end, structural connection part and outer shell surface of the flow meter to output corresponding multi-source sensing information; S2. Generate electrical anomaly markers, vibration spatial anomaly markers, and overheating area markers based on the three types of intelligent sensors, and perform joint judgment to determine whether to generate a sensing structure, and determine the priority of the sensing structure and the corresponding timing alignment range. S3. Based on the priority of the sensing structure, using the working causal chain of each functional unit of the flow meter as a framework, the multi-source sensing information is allocated to the corresponding link of the working causal chain, and the violation vector corresponding to each link is calculated. S4. Map the violation vector to the pre-constructed fault logic space to generate the corresponding fault criteria and matching score; S5. Based on the matching degree score, construct a fault cause-effect graph with each functional unit of the flow meter as a node and the signal transmission relationship of each functional unit as a directed edge. Perform reverse tracing and consistency verification in the fault cause-effect graph and output the fault root cause location result.

[0018] Compared with the prior art, the present invention has the following beneficial effects: By converting multi-source continuous sensing data into binary anomaly markers and logically combining them for judgment, a shift from data-driven to event-driven processing is achieved. Subsequent analysis is initiated only for data meeting preset anomaly patterns, reducing the number of invalid sensing structures generated. A dual-path judgment mechanism—"direct triggering of electrical anomalies" and "joint triggering of vibration and thermal anomalies"—is employed to cover electrical and mechanical-thermal coupling faults, improving the ability to identify complex fault modes and reducing the probability of false positives. The anomaly judgment process is moved forward to before timing alignment, filtering invalid data in advance, reducing unnecessary operations, and improving overall processing efficiency and real-time response capabilities. The logical relationships of multi-source anomalies are described in a structured Boolean expression, with clear and interpretable judgment rules, facilitating the adjustment of thresholds and logical relationships, and exhibiting good engineering adaptability and scalability. A unified triggering basis is provided for subsequent sensing structure generation and priority determination, enhancing the connectivity and synergy of various modules in the fault detection system.

[0019] Establish a mapping relationship between anomaly triggering paths, priorities, and timing alignment ranges to automatically adjust subsequent processing scopes and achieve differentiated resource scheduling. Assign electrical anomalies first priority to ensure complete analysis of critical anomalies; assign second priority to mechanical and thermal coupling anomalies to reduce computational and storage overhead. Use priority determination results as the basis for subsequent module data calls to improve the synergy between functional modules. Use explicit logical expressions to define priority rules, ensuring that the processing strategy is interpretable and allows for flexible adjustment of rules and scope. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of a flow meter fault detection system based on an intelligent sensor according to the present invention. Figure 2 This is a schematic diagram of a flow meter fault detection method based on intelligent sensors according to the present invention. Detailed Implementation

[0021] 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. Example

[0022] Please see Figure 1 As shown, this embodiment provides a flow meter fault detection system based on intelligent sensors, specifically including the following steps: The sensor deployment module is used to install first-class, second-class, and third-class smart sensors at the signal output end, structural connection parts, and outer surface of the flow meter, respectively, and output corresponding multi-source sensing information. The anomaly labeling and filtering module is used to generate electrical anomaly labels, vibration space anomaly labels and overheating area labels based on three types of smart sensors, and to perform joint judgment to determine whether to generate a sensing structure, and to determine the priority of the sensing structure and the corresponding timing alignment range. The instrument status analysis module is used to allocate multi-source sensing information to the corresponding links of the working causal chain based on the priority of the sensing structure and the working causal chain of each functional unit of the flow meter, and to calculate the violation vector corresponding to each link. The fault criterion generation module is used to map the violation vector to a pre-built fault logic space and generate corresponding fault criteria and matching scores. The fault tracing and location module is used to construct a fault cause-effect graph based on the matching degree score, using each functional unit of the flow meter as a node and the signal transmission relationship between each functional unit as directed edges. It then performs reverse tracing and consistency verification within the fault cause-effect graph and outputs the fault root cause location result. The modules are connected via wired and / or wireless means to achieve data transmission between them.

[0023] The methods for setting up Class I, Class II, and Class III smart sensors respectively include: After the flow meter is installed, the intelligent sensors are deployed according to the structure of the instrument. A first type of intelligent sensor is set at the signal output end of the flow meter. The first type of intelligent sensor is electrically connected to the output signal loop of the flow meter to collect output voltage, current and signal phase information, and output the corresponding electrical characteristic information. Specifically, the signal output terminal can be a 4~20mA current output terminal, a pulse output terminal, a frequency output terminal, or a digital communication interface. The first type of intelligent sensor is electrically connected to the signal output circuit and is used to collect the electrical characteristic information of the instrument's output signal in real time. The electrical characteristic information includes the effective value of the output voltage, the effective value of the output current, the current ripple amplitude, the signal frequency, the signal phase angle, and the phase change between adjacent sampling periods. Through this electrical characteristic information, the operating status of the signal conditioning circuit, the analog-to-digital conversion circuit, and the output drive circuit can be reflected.

[0024] For example, for an electromagnetic flowmeter with an output of 4~20mA, the first type of intelligent sensor detects an output current of 12.04mA, a current ripple amplitude of 0.18mA, and a phase angle offset of 2.6°, and outputs the above parameters as electrical characteristic information.

[0025] A second type of intelligent sensor is installed at the structural connection points of the flow meter, including the meter head, pipe connection flange, and transmitter housing. The second type of intelligent sensor is used to collect vibration information of each structural connection point and output the corresponding vibration status information. The second type of intelligent sensor is used to collect vibration state information at various locations. This vibration state information includes the root mean square value of vibration acceleration, the effective value of vibration velocity, the dominant frequency, the peak amplitude of the spectral spectrum, and the vibration correlation coefficient between different installation locations. This vibration state information reflects the vibration conditions of the instrument's internal sensitive elements, mechanical connection structures, and electronic components. For example, a second-type intelligent sensor installed at a pipe connection flange measured a root mean square value of vibration acceleration of 1.8g, a dominant frequency of 96Hz, and a vibration correlation coefficient of 0.42 with the meter head location.

[0026] A third type of intelligent sensor is installed on the surface of the flow meter housing. The third type of intelligent sensor consists of an array structure composed of several temperature detection units, which collects temperature information at different locations on the surface of the flow meter housing and outputs the corresponding temperature field distribution information.

[0027] The third type of intelligent sensor consists of an array of multiple temperature detection units distributed at preset intervals on the surface of the housing to simultaneously collect temperature values ​​at different locations. The measurement results from each temperature detection unit are combined and processed to form temperature field distribution information. This information includes the temperature values ​​at each sampling location, the temperature gradient between adjacent locations, the highest temperature value, the lowest temperature value, and the location and range of overheated areas. For example, the third type of intelligent sensor detects a highest temperature of 78.5℃, a lowest temperature of 52.3℃, and a maximum temperature gradient of 6.8℃ / cm on the housing surface, and identifies a localized overheated area located in the upper right region of the transmitter housing.

[0028] Methods for outputting corresponding multi-source sensing information include: The electrical characteristic information output by the first type of intelligent sensor, the vibration state information output by the second type of intelligent sensor, and the temperature field distribution information output by the third type of intelligent sensor are timestamped according to a unified time base. The various types of information after the timestamping are combined and encapsulated according to a preset data format to output the corresponding multi-source sensing information.

[0029] In this embodiment, it should be noted that, in order to enable the electrical characteristic information output by the first type of intelligent sensor, the vibration state information output by the second type of intelligent sensor, and the temperature field distribution information output by the third type of intelligent sensor to be analyzed simultaneously, a unified time reference is preset in the system. The unified time reference is provided by the internal controller of the flow meter. After completing their respective feature extraction, each type of intelligent sensor reads the time value corresponding to the current unified time reference and appends this time value as a timestamp to its respective output feature information.

[0030] Specifically, the first type of intelligent sensor appends a timestamp to the electrical characteristic information after obtaining it; the second type of intelligent sensor appends a timestamp from the same time reference to the vibration state information after obtaining it; and the third type of intelligent sensor appends a timestamp from the same time reference to the temperature field distribution information after obtaining it. For information with the same timestamp or a time difference not exceeding a preset synchronization window, the system determines it as a valid sensing result corresponding to the same moment.

[0031] Subsequently, the electrical characteristic information, vibration state information, and temperature field distribution information corresponding to the same moment are combined and encapsulated according to a preset data format. The preset data format includes at least a timestamp field, an information category field, and a feature parameter field. The timestamp field records the acquisition time, the information category field identifies the information source, and the feature parameter field records the corresponding feature values. After encapsulation, multi-source sensing information is formed and output to the anomaly labeling and filtering module as input for subsequent anomaly identification processing.

[0032] For example, at timestamp 2026-04-14 10:15:30: The first type of intelligent sensor has an output current ripple amplitude of 0.18mA and a phase angle offset of 2.6°. The second type of intelligent sensor outputs a main frequency of 96Hz and a root mean square value of vibration acceleration of 1.8g. The third type of intelligent sensor outputs a maximum temperature of 78.5℃ and a maximum temperature gradient of 6.8℃ / cm.

[0033] The system combines and encapsulates the above information as follows: { Timestamp: 2026-05-14 10:15:30, / / Collection time ElectricalFeatures:{...}, / / Electrical feature information VibrationFeatures:{...}, / / Vibration state information ThermalFeatures:{...} / / Temperature field distribution information } This structured data is multi-source sensing information. Among them, Timestamp is used to record the acquisition time under a unified time reference, ElectricalFeatures is used to record the electrical feature information extracted by the first type of smart sensor, VibrationFeatures is used to record the vibration state information extracted by the second type of smart sensor, and ThermalFeatures is used to record the temperature field distribution information extracted by the third type of smart sensor.

[0034] Methods for determining whether a perceptual structure has been generated include: The deviation between the current electrical characteristic information and the corresponding reference value is calculated based on the electrical characteristic information output by the first type of intelligent sensor. When the deviation is greater than the preset deviation threshold, an electrical anomaly mark is generated. Specifically, for the electrical characteristic information output by the first type of intelligent sensor, the system pre-establishes a reference feature set under normal operating conditions. The reference feature set includes reference values ​​for the effective value of the output voltage, the effective value of the output current, the current ripple amplitude, the signal frequency, and the phase angle. The system compares each electrical characteristic at the current acquisition moment with its corresponding reference value, calculates the difference between each characteristic, and performs weighted summation or root mean square calculation on each difference to obtain the deviation between the current electrical characteristic information and the reference value. This deviation is used to characterize the overall degree of deviation of the current output signal from the normal state.

[0035] Electrical anomaly marker: ;in, This indicates an electrical anomaly flag. A value of 1 indicates that the first type of smart sensor has detected an electrical anomaly, and a value of 0 indicates that no electrical anomaly has been detected. Indicates the deviation between the current electrical characteristic information and the corresponding normal reference value; indicates the preset deviation threshold. This indicates an indicator function that takes the value 1 when the condition within the parentheses is true, and 0 otherwise.

[0036] For example, if the difference between the current current ripple amplitude and the corresponding reference value is 0.22mA, and the preset electrical anomaly threshold is 0.15mA, then since 0.22 > 0.15, therefore... .

[0037] The vibration space consistency index between each structural connection part is calculated based on the vibration state information output by the second type of intelligent sensor. When the vibration space consistency index corresponding to any structural connection part is less than the preset consistency index threshold, a vibration space anomaly mark is generated. Based on the vibration state information output by the second type of intelligent sensor, the system acquires vibration characteristic parameters located at the instrument head, pipe connection flanges, and transmitter housing, and calculates the vibration similarity between different locations. Vibration similarity can be characterized using correlation coefficients, spectral overlap, or dominant frequency differences. Subsequently, the vibration similarity between each location is synthesized to obtain the vibration spatial consistency index corresponding to each structural connection. The vibration spatial consistency index reflects the consistency of the vibration response of each structural component.

[0038] Vibration spatial anomaly marker: ;in, The value of 1 indicates that there is an abnormal vibration space, and the value of 0 indicates that no abnormal vibration space was detected. A quantifier indicating existence, meaning that at least one exists; This indicates the installation location of the second type of intelligent sensor, including the instrument head, pipe connection flange, and transmitter housing; Indicates position The corresponding vibration spatial consistency index; This indicates the preset consistency index threshold; For example, if the vibration spatial consistency index corresponding to the flange location is 0.31, and the preset threshold is 0.45, then since 0.31 < 0.45, therefore... .

[0039] The overheated region is identified based on the temperature field distribution information output by the third type of intelligent sensor. When the identified overheated region is not empty, an overheated region marker is generated. Based on the temperature field distribution information output by the third type of intelligent sensor, the system reconstructs the temperature values ​​according to the spatial location of each temperature detection unit, forming a two-dimensional temperature distribution map of the outer casing surface. The system compares the temperature values ​​at each location with a preset temperature threshold and, combined with the temperature gradient between adjacent locations, identifies continuous overheated areas, thus obtaining a set of overheated areas. This set of overheated areas includes the location range, area, and highest temperature value of each overheated area. For example, if the measured values ​​from three consecutive temperature detection units above the transmitter housing are 79.2℃, 80.1℃, and 78.8℃, respectively, and all exceed the preset temperature threshold of 75℃, the system identifies this continuous area as an overheated area.

[0040] Overheated area marking: ;in, This indicates an overheated area marker, with a value of 1 indicating the presence of an overheated area and a value of 0 indicating the absence of an overheated area. This represents the set of overheated regions identified based on temperature field distribution information. Indicates the empty set; This indicates that the set of overheated regions contains at least one overheated region; for example, when a localized overheated region located above the transmitter housing is identified, the set... Not empty, therefore .

[0041] A joint determination is performed based on electrical anomaly markers, vibration space anomaly markers, and overheating zone markers. When the electrical anomaly marker is in an abnormal state, or when the electrical anomaly marker is in a normal state and both the vibration space anomaly marker and the overheating zone marker are in an abnormal state, a sensing structure is determined to be generated.

[0042] Execute joint judgment: ;in, Indicates the joint judgment result; Represents a logical OR operation; Represents the logical AND operation; when When this occurs, it indicates that the preset exception triggering conditions are met, and a perception structure needs to be generated; when When this occurs, it indicates that the abnormal triggering condition has not been met, and no perception structure is generated. For example, when... , , hour, ;when , , hour, ;when , , hour, ; The meaning of the above joint determination logic is: when an electrical abnormality is marked When an electrical anomaly is detected, it is considered that there is a significant abnormality in the instrument output link, directly triggering subsequent processing; when an electrical anomaly is flagged... However, when both the vibration space anomaly marker and the overheating area marker are set simultaneously, it indicates a coupling phenomenon between mechanical and thermal anomalies in the instrument, triggering subsequent processing. This logic ensures that subsequent analysis is initiated only for acquisition results with clearly defined anomaly characteristics, thereby reducing the generation of invalid sensing structures and improving system processing efficiency.

[0043] Methods for determining the priority of a sensing structure and its corresponding timing alignment range include: When generating a sensing structure, if the electrical anomaly is marked as an abnormal state, the sensing structure is determined as the first priority; if the electrical anomaly is marked as a normal state and the vibration space anomaly and the overheating area anomaly are both marked as abnormal states, the sensing structure is determined as the second priority. In this embodiment, it should be noted that the priority of the sensing structure is determined according to the following formula: ;in, Indicates the priority of the perception structure; This indicates that the electrical abnormality marker is in a normal state, while the vibration space abnormality marker and the overheating area marker are both in an abnormal state; when When this occurs, it indicates a significant anomaly in the instrument output link, and the sensing structure is prioritized as the first priority; when and , When this occurs, it indicates that the system has detected the coupling phenomenon between mechanical and thermal anomalies, and the sensing structure is marked as the second priority.

[0044] When the perception structure is the first priority, the multi-source perception information of the first type of intelligent sensor, the second type of intelligent sensor, and the third type of intelligent sensor within the preset time window corresponding to the target timestamp is called, and the complete time sequence alignment is performed according to the unified time reference. In this embodiment, it should be noted that when the anomaly marker filtering module determines that a sensing structure needs to be generated based on the joint judgment result of electrical anomaly markers, vibration space anomaly markers, and overheated area markers, the timestamp corresponding to the current multi-source sensing information that triggered the joint judgment is determined as the target timestamp. The target timestamp represents the acquisition time when the current anomaly state first meets the preset triggering condition, and serves as the time reference point for subsequent timing alignment processing.

[0045] After determining the target timestamp, the system constructs a preset time window centered on the target timestamp. The preset time window consists of a first time length before the target timestamp and a second time length after the target timestamp. The first time length represents the time range tracing back before the target timestamp, used to extract state evolution information before the anomaly occurs; the second time length represents the time range extending forward after the target timestamp, used to extract response information after the anomaly is triggered. The first and second time lengths can be preset according to the flow meter's sampling period, data storage capacity, and fault evolution speed.

[0046] The system extracts sensing information whose timestamps fall within the preset time window from data output by various smart sensors and sorts them using a unified time base. Data with the same timestamp or a time difference not exceeding a preset synchronization tolerance are identified as data corresponding to the same moment and time-series aligned, thereby forming a continuous time-series data set centered around the target timestamp.

[0047] For example, if the multi-source sensing information at a certain acquisition moment meets the joint judgment condition, and its timestamp is 10:15:30, then 10:15:30 is determined as the target timestamp. If the first time length is preset to 5 seconds and the second time length to 2 seconds, then the preset time window is 10:15:25 to 10:15:32. The system extracts the multi-source sensing information within this time interval for time sequence alignment to obtain the complete state change process before and after the anomaly occurs.

[0048] When the sensing structure is of the second priority, the multi-source sensing information of the second and third type of intelligent sensors within the preset time window corresponding to the target timestamp is called, and local timing alignment is performed according to a unified time reference, thereby determining the priority of the sensing structure and the corresponding timing alignment range.

[0049] For example, when the anomaly marker at a certain acquisition time satisfies , , When the system determines the sensing structure as the first priority, it retrieves data from the three types of intelligent sensors near the acquisition time for complete time-series alignment; when the anomaly flag meets the requirements... , , At that time, the system will determine the perception structure as the second priority and only retrieve data from the second and third types of intelligent sensors for local timing alignment.

[0050] Methods for calculating the violation vector corresponding to each stage include: The range of multi-source sensing information to be analyzed is determined based on the priority identifier carried in the sensing structure. When the sensing structure is of the first priority, the electrical characteristic information, vibration state information, and temperature field distribution information after complete time-series alignment are called; when the sensing structure is of the second priority, the vibration state information and temperature field distribution information after local time-series alignment are called. Construct a causal chain based on the working process of the flow meter, including a sensing unit, a signal conditioning unit, a signal conversion unit, and a transmitter output unit; and allocate the multi-source sensing information to the corresponding links in the causal chain according to the correspondence between various multi-source sensing information and each functional unit. In this embodiment, it should be noted that the instrument status analysis module first constructs a causal chain based on the actual signal formation process of the flowmeter. Specifically, according to the processing path of the flowmeter from fluid action to the final output measurement result, four functional units are sequentially determined: a sensing unit, a signal conditioning unit, a signal conversion unit, and a transmitter output unit. Directed connections are established based on the relationship that the output of the previous functional unit serves as the input of the next functional unit. Specifically, the sensing unit converts the physical changes caused by fluid flow into the raw measurement signal; the signal conditioning unit amplifies, filters, and performs anti-interference processing on the raw measurement signal; the signal conversion unit performs analog-to-digital conversion and flow calculation on the conditioned signal; and the transmitter output unit converts the calculation result into a standard current signal, pulse signal, or digital communication data. This forms a causal chain that is sequentially transmitted along the instrument's operating process.

[0051] After establishing the causal chain, the multi-source sensing information is allocated to corresponding links based on the physical relationship between the multi-source sensing information and each functional unit. Specifically, vibration state information and temperature field distribution information, used to characterize the stress state and thermal environment state of the instrument structure, are preferentially allocated to the sensing and signal conditioning units; electrical characteristic information, used to characterize the state of the signal processing circuit and output loop, is allocated to the signal conversion unit and the transmitter output unit. For characteristic information that is simultaneously related to multiple functional units, it is allocated to multiple links according to a preset mapping relationship to reflect the coupling influence between different links.

[0052] For each link in the causal chain of work, the multi-source sensing information assigned to that link is substituted into the pre-established operational constraint relationship of that link to calculate the difference between the actual observation results and the predicted results of the constraint relationship, and the various difference quantities are combined in a preset order to form the violation vector corresponding to the link.

[0053] After information allocation is completed, for each link in the causal chain, the multi-source sensing information allocated to that link is substituted into pre-established operational constraints. Operational constraints describe the correspondence that each observation should satisfy under normal operating conditions for that functional link, and are determined by equipment design parameters, factory calibration results, or historical operating data. The system calculates the prediction result under the current input conditions based on the operational constraints and compares the prediction result with the actual observation result to obtain the difference. The difference is used to characterize the degree of deviation of the current link from the normal operating pattern.

[0054] For example, in the transmitter output unit, a pre-established correspondence between the output current and the calculated flow rate is established. When the theoretical output current corresponding to the current calculated flow rate is 12.00 mA, while the measured output current of the first type of intelligent sensor is 12.28 mA, the difference of 0.28 mA is taken as a difference quantity for this stage. In the sensing unit, when the root mean square value of vibration acceleration and the local temperature gradient exceed the normal operating range, structural vibration difference and thermal difference are obtained, respectively. The system combines these difference quantities in a preset order to form a deviation vector for the corresponding functional stage, which is used to represent the deviation of the functional stage from the normal operating state.

[0055] Methods for generating corresponding fault criteria and matching scores include: A fault logic space is pre-constructed, which includes several criteria entries corresponding to different fault modes of the flow meter. Each criterion entry includes three dimensions of judgment conditions: fault severity, fault evolution stage, and fault impact range. Specifically, the system establishes several criterion entries corresponding to fault modes based on typical fault types that may occur in flowmeters during long-term operation. Fault modes include contamination of sensing elements, loose structural connections, signal conditioning circuit drift, abnormal signal conversion, abnormal transmitter output, and localized overheating. Each criterion entry is described using a unified three-dimensional structure, which includes fault severity, fault evolution stage, and fault impact range. Fault severity characterizes the degree to which the current state deviates from normal operating conditions; fault evolution stage characterizes the development of the anomaly over time; and fault impact range characterizes the spread of the anomaly within the operational causal chain. The system pre-defines the judgment conditions and their logical combinations corresponding to the above three dimensions for each fault mode, thereby forming a fault logic space that can be matched.

[0056] For example, for a "loose structural connection" fault, its criterion can be predefined as "second-degree anomaly + development stage + chain effect"; for a "signal conditioning drift" fault, its criterion can be predefined as "first-degree anomaly + stable stage + local effect"; and for a "transmitter output anomaly" fault, its criterion can be predefined as "second-degree anomaly + stable stage + local effect". The set of each criterion constitutes the fault logic space.

[0057] Receive the violation vectors corresponding to each link in the working causal chain, determine the severity of the fault based on the magnitude of each violation vector, determine the fault evolution stage based on the changing trend of the violation vectors in several consecutive acquisition times, and determine the scope of the fault impact based on the number of functional links whose violation vectors exceed the preset violation vector threshold and their distribution relationship in the working causal chain. After receiving the violation vectors corresponding to each link in the causal chain, the system first determines the severity of the fault based on the magnitude of each violation vector. Specifically, it calculates the magnitude of each violation vector and takes the maximum value as the current anomaly intensity index. Then, it compares the anomaly intensity index with preset classification thresholds to determine the severity of the fault. For example, when the anomaly intensity index is lower than the first anomaly intensity index threshold, it is determined to be a normal state; when the anomaly intensity index is between the first and second anomaly intensity index thresholds, it is determined to be a first-degree anomaly; and when the anomaly intensity index is greater than the second anomaly intensity index threshold, it is determined to be a second-degree anomaly.

[0058] For example, if the threshold for the first anomaly intensity index is 0.30 and the threshold for the second anomaly intensity index is 0.70, when the magnitude of the violation vector of the sensing unit, signal conditioning unit, signal conversion unit, and transmitter output unit is 0.92, 0.78, 0.21, and 0.18 respectively, the current maximum value is 0.92, which is greater than the second anomaly intensity index threshold of 0.70. Therefore, the severity of the fault is determined to be the second level of anomaly.

[0059] The system tracks the changes in the violation degree vector over several consecutive acquisition times and calculates the trend of the anomaly intensity index over time. If the anomaly intensity index continues to rise, the fault evolution stage is determined to be the development stage; if the anomaly intensity index is basically stable, it is determined to be the stable stage; if the anomaly intensity index just exceeds the preset threshold and the duration is short, it is determined to be the initial stage. For example, in five consecutive acquisition times, the anomaly intensity index is 0.41, 0.56, 0.68, 0.81, and 0.92 respectively, showing a continuous increasing trend. Therefore, the fault evolution stage is determined to be the development stage.

[0060] Furthermore, the system counts the number of functional components whose violation vector exceeds a preset violation threshold and determines the scope of the fault's impact based on the positional relationship of these functional components in the causal chain. When only one functional component exceeds the preset violation threshold, it is determined to be a local impact; when multiple adjacent functional components exceed the preset violation threshold, it is determined to be a chain impact; when multiple non-adjacent functional components simultaneously exceed the threshold, it is determined to be a multi-point impact. For example, in the above data, the magnitudes of the sensing unit and the signal conditioning unit are 0.92 and 0.78, respectively, both exceeding 0.50, and they are adjacent in the causal chain; therefore, the scope of the fault's impact is determined to be a chain impact.

[0061] The determined fault severity, fault evolution stage, and fault impact range are mapped to the fault logic space and matched with the judgment conditions of each criterion item. The matching degree score of each criterion item is calculated to generate the corresponding fault criterion and matching degree score.

[0062] After determining the severity, evolution stage, and scope of the fault, the system combines these three results to form a current state description, maps it to the fault logic space, and matches it against the judgment conditions of each criterion item. Specifically, the matching results of the three dimensions can be calculated separately and then combined according to preset weights to obtain a matching degree score for each criterion item.

[0063] For example, the current state is described as "Second-degree anomaly + development stage + chain effect"; it is completely consistent with the "loose structural connection" criterion, so its three-dimensional matching values ​​are 1.00, 1.00, and 1.00 respectively, with a comprehensive matching score of 1.00; compared with the "transmitter output anomaly" criterion, only the severity is consistent, so the matching value for this dimension is 1.00; in the fault evolution stage, "development stage" and "stable stage" are not completely consistent, but they are adjacent in evolution order, so the matching value for this dimension can be 0.40; in the fault impact range, "chain effect" and "local effect" are inconsistent, so the matching value for this dimension can be 0.20. If the three dimensions are calculated using an equal-weighted method, the comprehensive matching score is (1.00 + 0.40 + 0.20) / 3 ≈ 0.53. Since the matching score corresponding to "loose structural connection" is the highest, "loose structural connection" is generated as the current fault criterion, and a matching score of 1.00 is output.

[0064] Methods for constructing cause-effect graphs include: The sensing unit, signal conditioning unit, signal conversion unit, and transmitter output unit of the flow meter are respectively identified as nodes in the fault cause-effect graph; according to the signal transmission relationship between each functional unit, a directed edge is established between the corresponding node of the previous functional unit and the corresponding node of the next functional unit. Receive the fault criteria and the corresponding matching score, and use the matching score as the global credibility coefficient; obtain the violation vector corresponding to each functional unit, calculate the magnitude of each violation vector, and combine the magnitude of each violation vector with the global credibility coefficient to obtain the activation value corresponding to each node; the nodes, directed edges and the activation values ​​corresponding to each node together constitute the fault cause-effect graph.

[0065] For example, if the current fault criterion is "loose structural connection" and its matching score is 0.96, the system will use 0.96 as the global reliability coefficient. If the violation vector magnitudes of the sensing unit, signal conditioning unit, signal conversion unit, and transmitter output unit are 0.92, 0.78, 0.21, and 0.18, respectively, the system can determine the activation values ​​of each node as 0.8832, 0.7488, 0.2016, and 0.1728, respectively, where each node activation value is obtained by multiplying the corresponding violation vector magnitude by the matching score.

[0066] Methods for outputting root cause analysis results include: Based on the activation values ​​of each node in the fault cause-effect graph, nodes with activation values ​​greater than a preset activation threshold are identified as symptom nodes. Starting from the symptom node, trace back level by level along the opposite direction of the directed edges in the fault cause-effect graph, and add upstream nodes with activation values ​​greater than the preset activation threshold to the root cause candidate set. For example, when the preset activation threshold is 0.60, if the activation values ​​of the nodes corresponding to the sensing unit and the signal conditioning unit are 0.8832 and 0.7488 respectively, then these two nodes are identified as symptom nodes. In the example above, the sensing unit is the starting node of the working causal chain and has no upstream nodes; the upstream node of the signal conditioning unit is the sensing unit, and its activation value is 0.8832, which is higher than the candidate threshold of 0.50. Therefore, the sensing unit is added to the root cause candidate set. The final root cause candidate set includes the sensing unit.

[0067] For each candidate node in the root cause candidate set, the candidate node is defined as the root cause node. Forward deduction is performed along the direction of the directed edge in the fault cause graph to obtain the predicted abnormal distribution of each downstream node. The predicted abnormal distribution is then compared with the currently observed node activation status. In this embodiment, it should be noted that after obtaining the root cause candidate set through reverse tracing, the system performs consistency verification for each candidate node in the root cause candidate set. Specifically, it first assumes that the current candidate node is the actual root cause node of the fault, and uses the node activation value corresponding to the candidate node as the initial anomaly intensity. Then, the anomaly intensity is propagated downstream along the direction of the directed edges in the fault cause-effect graph to obtain the predicted anomaly distribution of each downstream node.

[0068] Specifically, for any path in the fault cause-effect graph that originates from the upstream node... Pointing to downstream nodes For directed edges, the system pre-sets the propagation coefficient of the directed edge. ,in The value ranges from 0 to 1. The propagation coefficient is used to characterize the proportion of the impact of an anomaly in an upstream functional unit being transmitted to a downstream functional unit. Let the upstream node... The predicted anomaly intensity is Then the downstream node Predicted anomaly intensity Calculate according to the following formula: When a node has multiple upstream nodes, the predicted anomaly intensities propagated from each upstream node can be summed or the maximum value can be taken as the final predicted anomaly intensity of that node.

[0069] For example, assuming the sensing unit is a candidate root cause node with an activation value of 0.8832; the propagation coefficient from the sensing unit to the signal conditioning unit is 0.85, from the signal conditioning unit to the signal conversion unit is 0.40, and from the signal conversion unit to the transmitter output unit is 0.30, then the predicted anomaly intensities for each node are as follows: sensing unit 0.8832, signal conditioning unit 0.8832 × 0.85 ≈ 0.7507, signal conversion unit 0.7507 × 0.4 ≈ 0.3003, and transmitter output unit 0.3003 × 0.3 ≈ 0.0901. This yields the predicted anomaly distribution along the working causal chain.

[0070] After obtaining the predicted anomaly distribution, the system compares the predicted anomaly intensity of each node with the actual observed node activation value and calculates a consistency index. Specifically, the sum of the absolute differences between the predicted and actual values ​​of each node can be used. When the consistency index is less than or equal to a preset consistency index threshold, the current candidate node is determined to meet the preset consistency condition; otherwise, the current candidate node is determined to have failed the consistency verification.

[0071] When the comparison results meet the preset consistency conditions, the corresponding candidate node is determined to have passed the consistency verification. Based on the node activation value, fault criterion matching score, and consistency verification results of the candidate node that has passed the consistency verification, the root cause confidence score of each candidate node is calculated. The candidate node with the highest root cause confidence score is determined as the fault root cause node, and the fault root cause location result is output.

[0072] The root cause localization results include the functional unit name corresponding to the root cause node, the root cause confidence score, and the fault propagation path from the root cause node to the symptom node.

[0073] It should be noted that for candidate nodes that pass the consistency verification, the system calculates the root cause confidence score based on the node activation value, the matching score of the fault criterion, and the consistency verification result. Specifically, it can be calculated using the following formula: ;in, Indicates the root cause credibility score; This represents the node activation value of the candidate node; This represents the matching score of the corresponding fault criterion; Indicates consistency index; This indicates the preset consistency index threshold; , and Indicates the preset weight, and satisfies , and The sum is 1.

[0074] It should be noted that the root cause credibility score is calculated jointly by the node activation value of the candidate node, the matching score of the fault criterion, and the consistency score. Among them, the node activation value C and the matching score are both normalized, and the consistency index and the preset consistency index threshold have the same dimension. The division between the two yields a dimensionless ratio, so the consistency score is also a dimensionless value between 0 and 1.

[0075] When the consistency index exceeds a preset consistency index threshold, the corresponding candidate node can be directly determined to have failed the consistency verification. This is due to the weighting... , and The values ​​of all three are between 0 and 1, and the sum of the three is equal to 1. Therefore, the root cause credibility score is actually a weighted average of multiple dimensionless indicators, and its value range is always limited to between 0 and 1, so there is no problem of inconsistent dimensions.

[0076] For example, let , , With a candidate node activation value of 0.8832, a fault criterion matching score of 0.96, a consistency index of 0.1833, and a consistency threshold of 0.25, the root cause confidence score is: The system compares the root cause confidence scores of each candidate node and identifies the candidate node with the highest score as the root cause node of the failure.

[0077] The preset deviation threshold is set by staff based on historical data analysis results. This historical analysis process includes the system collecting multiple deviation values ​​and calculating their average value as a reference to obtain the preset deviation threshold. Similarly, the preset consistency index threshold, preset violation vector threshold, and preset activation threshold are also set by staff based on the system's historical operating data and specific application scenario requirements. These preset thresholds are adjusted by staff during system operation according to the actual situation.

[0078] This embodiment transforms data-driven processing into event-driven processing by converting multi-source continuous sensing data into binary anomaly markers and logically combining them for judgment. Subsequent analysis is initiated only for data meeting preset anomaly patterns, reducing the number of invalid sensing structures generated. A dual-path judgment mechanism—"direct triggering of electrical anomalies" and "joint triggering of vibration and thermal anomalies"—covers electrical and mechanical-thermal coupling faults, improving the ability to identify complex fault modes and reducing the probability of misjudgment. The anomaly judgment process is moved forward to before timing alignment, filtering invalid data in advance, reducing unnecessary operations, and improving overall processing efficiency and real-time response capabilities. The logical relationships of multi-source anomalies are described in a structured Boolean expression, with clear and interpretable judgment rules, facilitating the adjustment of thresholds and logical relationships, and exhibiting good engineering adaptability and scalability. This provides a unified triggering basis for subsequent sensing structure generation and priority determination, enhancing the connectivity and synergy of various modules in the fault detection system.

[0079] Establish a mapping relationship between anomaly triggering paths, priorities, and timing alignment ranges to automatically adjust subsequent processing scopes and achieve differentiated resource scheduling. Assign electrical anomalies first priority to ensure complete analysis of critical anomalies; assign second priority to mechanical and thermal coupling anomalies to reduce computational and storage overhead. Use priority determination results as the basis for subsequent module data calls to improve the synergy between functional modules. Use explicit logical expressions to define priority rules, ensuring that the processing strategy is interpretable and allows for flexible adjustment of rules and scope. Example

[0080] Please see Figure 2 As shown, parts not described in detail in this embodiment are described in Embodiment 1. A method for fault detection of flow meters based on intelligent sensors is provided, including: S1. First, second and third type intelligent sensors are respectively installed at the signal output end, structural connection part and outer shell surface of the flow meter to output corresponding multi-source sensing information; S2. Generate electrical anomaly markers, vibration spatial anomaly markers, and overheating area markers based on the three types of intelligent sensors, and perform joint judgment to determine whether to generate a sensing structure, and determine the priority of the sensing structure and the corresponding timing alignment range. S3. Based on the priority of the sensing structure, using the working causal chain of each functional unit of the flow meter as a framework, the multi-source sensing information is allocated to the corresponding link of the working causal chain, and the violation vector corresponding to each link is calculated. S4. Map the violation vector to the pre-constructed fault logic space to generate the corresponding fault criteria and matching score; S5. Based on the matching degree score, construct a fault cause-effect graph with each functional unit of the flow meter as a node and the signal transmission relationship of each functional unit as a directed edge. Perform reverse tracing and consistency verification in the fault cause-effect graph and output the fault root cause location result. Example

[0081] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the flow meter fault detection system based on intelligent sensors described above.

[0082] Since the electronic device described in this embodiment is the electronic device used to implement the flow meter fault detection system and method based on intelligent sensors described in this application, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the flow meter fault detection system and method based on intelligent sensors described in this application. Therefore, how the electronic device implements the method in this application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the flow meter fault detection system and method based on intelligent sensors described in this application, it falls within the scope of protection of this application.

[0083] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0084] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A flow meter fault detection system based on intelligent sensors, characterized in that, include: The sensor deployment module is used to install first-class, second-class, and third-class smart sensors at the signal output end, structural connection parts, and outer surface of the flow meter, respectively, and output corresponding multi-source sensing information. The anomaly labeling and filtering module is used to generate electrical anomaly labels, vibration space anomaly labels and overheating area labels based on three types of smart sensors, and to perform joint judgment to determine whether to generate a sensing structure, and to determine the priority of the sensing structure and the corresponding timing alignment range. The instrument status analysis module is used to allocate multi-source sensing information to the corresponding links of the working causal chain based on the priority of the sensing structure and the working causal chain of each functional unit of the flow meter, and to calculate the violation vector corresponding to each link. The fault criterion generation module is used to map the violation vector to a pre-built fault logic space and generate corresponding fault criteria and matching scores. The fault tracing and location module is used to construct a fault cause-effect graph based on the matching degree score, with each functional unit of the flow meter as a node and the signal transmission relationship of each functional unit as a directed edge. It performs reverse tracing and consistency verification in the fault cause-effect graph and outputs the fault root cause location result.

2. The flow meter fault detection system based on intelligent sensors according to claim 1, characterized in that, The method for setting up the first, second, and third types of smart sensors respectively includes: After the flow meter is installed, the intelligent sensors are deployed according to the structure of the instrument. A first type of intelligent sensor is set at the signal output end of the flow meter. The first type of intelligent sensor is electrically connected to the output signal loop of the flow meter to collect output voltage, current and signal phase information, and output the corresponding electrical characteristic information. A second type of intelligent sensor is installed at the structural connection points of the flow meter, including the meter head, pipe connection flange, and transmitter housing. The second type of intelligent sensor is used to collect vibration information of each structural connection point and output the corresponding vibration status information. A third type of intelligent sensor is installed on the surface of the flow meter housing. The third type of intelligent sensor consists of an array structure composed of several temperature detection units, which collects temperature information at different locations on the surface of the flow meter housing and outputs the corresponding temperature field distribution information.

3. The flow meter fault detection system based on intelligent sensors according to claim 2, characterized in that, The method for outputting multi-source sensing information includes: The electrical characteristic information output by the first type of intelligent sensor, the vibration state information output by the second type of intelligent sensor, and the temperature field distribution information output by the third type of intelligent sensor are timestamped according to a unified time base. The various types of information after the timestamping are combined and encapsulated according to a preset data format to output the corresponding multi-source sensing information.

4. The flow meter fault detection system based on intelligent sensors according to claim 3, characterized in that, The method for determining whether a perceptual structure has been generated includes: The deviation between the current electrical characteristic information and the corresponding reference value is calculated based on the electrical characteristic information output by the first type of intelligent sensor. When the deviation is greater than the preset deviation threshold, an electrical anomaly mark is generated. The vibration spatial consistency index between the structural connection parts is calculated based on the vibration state information output by the second type of intelligent sensor. When the vibration spatial consistency index corresponding to any structural connection part is less than the preset consistency index threshold, a vibration spatial anomaly mark is generated. The overheated area is identified based on the temperature field distribution information output by the third type of intelligent sensor. When the identified overheated area is not empty, an overheated area mark is generated. A joint determination is performed based on electrical anomaly markers, vibration space anomaly markers, and overheating zone markers. When the electrical anomaly marker is in an abnormal state, or when the electrical anomaly marker is in a normal state and both the vibration space anomaly marker and the overheating zone marker are in an abnormal state, a sensing structure is determined to be generated.

5. A flow meter fault detection system based on intelligent sensors according to claim 4, characterized in that, The method for determining the priority of the sensing structure and its corresponding timing alignment range includes: When generating a sensing structure, if the electrical anomaly is marked as an abnormal state, the sensing structure is determined as the first priority; if the electrical anomaly is marked as a normal state and the vibration space anomaly and the overheating area anomaly are both marked as abnormal states, the sensing structure is determined as the second priority. When the perception structure is the first priority, the multi-source perception information of the first type of intelligent sensor, the second type of intelligent sensor, and the third type of intelligent sensor within the preset time window corresponding to the target timestamp is called, and the complete time sequence alignment is performed according to the unified time reference. When the sensing structure is of the second priority, the multi-source sensing information of the second and third type of intelligent sensors within the preset time window corresponding to the target timestamp is called, and local timing alignment is performed according to a unified time reference, thereby determining the priority of the sensing structure and the corresponding timing alignment range.

6. A flow meter fault detection system based on intelligent sensors according to claim 5, characterized in that, The method for calculating the violation vector corresponding to each stage includes: The range of multi-source sensing information to be analyzed is determined based on the priority identifier carried in the sensing structure. When the sensing structure is of the first priority, the electrical characteristic information, vibration state information, and temperature field distribution information after complete time-series alignment are called; when the sensing structure is of the second priority, the vibration state information and temperature field distribution information after local time-series alignment are called. Construct a causal chain based on the working process of the flow meter, including a sensing unit, a signal conditioning unit, a signal conversion unit, and a transmitter output unit; and allocate the multi-source sensing information to the corresponding links in the causal chain according to the correspondence between various multi-source sensing information and each functional unit. For each link in the causal chain of work, the multi-source sensing information assigned to that link is substituted into the pre-established operational constraint relationship of that link to calculate the difference between the actual observation results and the predicted results of the constraint relationship, and the various difference quantities are combined in a preset order to form the violation vector corresponding to the link.

7. A flow meter fault detection system based on intelligent sensors according to claim 6, characterized in that, The method for generating the corresponding fault criteria and matching score includes: A fault logic space is pre-constructed, which includes several criteria entries corresponding to different fault modes of the flow meter. Each criterion entry includes three dimensions of judgment conditions: fault severity, fault evolution stage, and fault impact range. Receive the violation vectors corresponding to each link in the working causal chain, determine the severity of the fault based on the magnitude of each violation vector, determine the fault evolution stage based on the changing trend of the violation vectors in several consecutive acquisition times, and determine the scope of the fault impact based on the number of functional links whose violation vectors exceed the preset violation vector threshold and their distribution relationship in the working causal chain. The determined fault severity, fault evolution stage, and fault impact range are mapped to the fault logic space and matched with the judgment conditions of each criterion item. The matching degree score of each criterion item is calculated to generate the corresponding fault criterion and matching degree score.

8. A flow meter fault detection system based on intelligent sensors according to claim 7, characterized in that, The method for constructing a fault cause-effect graph includes: The sensing unit, signal conditioning unit, signal conversion unit, and transmitter output unit of the flow meter are respectively identified as nodes in the fault cause-effect graph; according to the signal transmission relationship between each functional unit, a directed edge is established between the corresponding node of the previous functional unit and the corresponding node of the next functional unit. Receive the fault criteria and the corresponding matching score, and use the matching score as the global credibility coefficient; obtain the violation vector corresponding to each functional unit, calculate the magnitude of each violation vector, and combine the magnitude of each violation vector with the global credibility coefficient to obtain the activation value corresponding to each node; the nodes, directed edges and the activation values ​​corresponding to each node together constitute the fault cause-effect graph.

9. A flow meter fault detection system based on intelligent sensors according to claim 8, characterized in that, The method for outputting fault root cause localization results includes: Based on the activation values ​​of each node in the fault cause-effect graph, nodes with activation values ​​greater than a preset activation threshold are identified as symptom nodes. Starting from the symptom node, trace back level by level along the opposite direction of the directed edges in the fault cause-effect graph, and add upstream nodes with activation values ​​greater than the preset activation threshold to the root cause candidate set. For each candidate node in the root cause candidate set, the candidate node is defined as the root cause node. Forward deduction is performed along the direction of the directed edge in the fault cause graph to obtain the predicted abnormal distribution of each downstream node. The predicted abnormal distribution is then compared with the currently observed node activation status. When the comparison results meet the preset consistency conditions, the corresponding candidate node is determined to have passed the consistency verification. Based on the node activation value, fault criterion matching score, and consistency verification results of the candidate node that has passed the consistency verification, the root cause confidence score of each candidate node is calculated. The candidate node with the highest root cause confidence score is determined as the fault root cause node, and the fault root cause location result is output.

10. A method for fault detection of flow meters based on intelligent sensors, implemented by a fault detection system for flow meters based on intelligent sensors as described in any one of claims 1 to 9, characterized in that, include: S1. First, second and third type intelligent sensors are respectively installed at the signal output end, structural connection part and outer shell surface of the flow meter to output corresponding multi-source sensing information; S2. Generate electrical anomaly markers, vibration spatial anomaly markers, and overheating area markers based on the three types of intelligent sensors, and perform joint judgment to determine whether to generate a sensing structure, and determine the priority of the sensing structure and the corresponding timing alignment range. S3. Based on the priority of the sensing structure, using the working causal chain of each functional unit of the flow meter as a framework, the multi-source sensing information is allocated to the corresponding link of the working causal chain, and the violation vector corresponding to each link is calculated. S4. Map the violation vector to the pre-constructed fault logic space to generate the corresponding fault criteria and matching score; S5. Based on the matching degree score, construct a fault cause-effect graph with each functional unit of the flow meter as a node and the signal transmission relationship of each functional unit as a directed edge. Perform reverse tracing and consistency verification in the fault cause-effect graph and output the fault root cause location result.

Images