Gas ultrasonic flowmeter system with fault diagnosis function

Abstract

The invention discloses a gas ultrasonic flowmeter system with a fault diagnosis function. Relates to the technical field of ultrasonic flowmeters. Comprising a measurement module, a data acquisition module, a fault diagnosis module and a decision output module, an ultrasonic sensor is arranged in the measuring module and is used for measuring the flowing state of gas in a sound channel; the data acquisition module acquires and pre-processes actually measured sound velocity, temperature, pressure and sensor working state parameters according to the gas flow state; the fault diagnosis module performs fault diagnosis on the gas ultrasonic flowmeter according to the preprocessed data to obtain a fault diagnosis result; and the decision output module outputs an early warning prompt according to the fault diagnosis result. The problems that a traditional gas ultrasonic flowmeter fault diagnosis system is one-sided in measurement, distorted in data, inaccurate in diagnosis and inconvenient to operate and maintain are solved, and the operation reliability and metering accuracy of the flowmeter are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic flow meter technology, and more specifically to a gas ultrasonic flow meter system with fault diagnosis. Background Technology

[0002] Traditional gas ultrasonic flow meter fault diagnosis systems often use single-channel or unreasonable sensor arrangements in their measurement modules, which cannot fully capture the gas flow state of different channels, resulting in insufficient raw data support for fault diagnosis. Existing systems lack targeted acquisition modes for different gas flow states, and the acquisition and calculation methods for key parameters are not standardized, resulting in poor data quality input to the fault diagnosis module and affecting diagnostic accuracy.

[0003] Traditional diagnostic solutions often employ a single model, failing to simultaneously consider static parameters and dynamic temporal characteristics. Furthermore, they lack effective mechanisms for fusing and verifying diagnostic results, leading to inaccurate fault type identification and ambiguous fault location, making it difficult to meet the precise diagnostic needs under complex operating conditions. Some existing systems lack clearly defined data interaction logic between modules and are deficient in supporting modules such as stable power supply, remote communication, and data storage. This results in poor system stability, inefficient transmission of fault warning information, and an inability to adapt to the remote monitoring and maintenance requirements of industrial scenarios.

[0004] In summary, existing fault diagnosis systems for gas ultrasonic flowmeters suffer from problems such as incomplete measurements, poor data quality, low diagnostic accuracy, and insufficient system stability, making it difficult to meet the reliability assurance requirements of industrial-grade gas metering.

[0005] Therefore, it is necessary to propose a gas ultrasonic flow meter system with a clear architecture, standardized data processing, and accurate fault diagnosis. Summary of the Invention

[0006] In view of this, the present invention provides a gas ultrasonic flow meter system with fault diagnosis, which effectively solves the problems of limited measurement, data distortion, inaccurate diagnosis, and inconvenient operation and maintenance of traditional gas ultrasonic flow meter fault diagnosis systems, and greatly improves the reliability and measurement accuracy of the flow meter.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A gas ultrasonic flow meter system with fault diagnosis includes: a measurement module, a data acquisition module, a fault diagnosis module, and a decision output module; The measurement module has a built-in ultrasonic sensor for measuring the flow state of gas in the acoustic channel; The data acquisition module collects measured sound velocity, temperature, pressure, and sensor operating status parameters based on the gas flow state, and performs preprocessing. The fault diagnosis module performs fault diagnosis on the gas ultrasonic flow meter based on the preprocessed data and obtains the fault diagnosis results. The decision output module outputs early warning prompts based on the fault diagnosis results.

[0008] Preferably, the ultrasonic sensor adopts a multi-channel distributed arrangement structure, including at least two sets of paired ultrasonic transducers. Each set of ultrasonic transducers corresponds to a different preset measurement channel on the flow meter body and outputs the corresponding original echo signal for measuring the flow state of gas in different measurement channels.

[0009] Preferably, the data acquisition module includes a status recognition unit, a multi-parameter synchronous acquisition unit, and a data preprocessing unit; The state recognition unit is used to receive the raw echo signal from the ultrasonic sensor, extract the time-domain and frequency-domain features of the raw echo signal, and determine the gas flow state type through a preset threshold; the multi-parameter synchronous acquisition unit starts the corresponding acquisition mode for different flow states to obtain raw data; the data preprocessing unit performs outlier removal, data calibration, data normalization, and data compression on the acquired raw data.

[0010] Preferably, the raw data includes measured sound velocity, temperature parameters, pressure parameters, and sensor operating status parameters; The acquisition process for the measured sound velocity includes: a multi-parameter synchronous acquisition unit receiving the transmitted and received signals from each group of ultrasonic transducers, calculating the signal transmission time difference based on a cross-correlation algorithm, and calculating the measured sound velocity in combination with the channel length, as shown in the following formula: ; Where L is the duct length, t 顺 t 逆 These represent the downstream and upstream transmission times of the ultrasonic signal, respectively. The temperature parameter acquisition process includes: the gas medium temperature output by the temperature sensor built into the flow meter pipeline; The pressure parameter acquisition process includes: acquiring the real-time pressure value of the gas medium through a pipeline pressure sensor, and performing pressure stabilization and filtering on the pressure signal during the acquisition process to eliminate pressure fluctuation interference caused by pipeline vibration; The process of acquiring sensor operating status parameters includes: acquiring the ultrasonic transducer's transmission power, received signal amplitude, impedance matching value, and operating voltage and current, while recording the sensor's cumulative operating time to form a basic dataset of sensor health status.

[0011] Preferably, the fault diagnosis module includes a feature extraction unit, a parallel diagnosis unit, a result fusion unit, and a fault level determination unit; The feature extraction unit is used to receive preprocessed data and extract static features and dynamic temporal features from the data. The static features include sound velocity deviation rate, temperature fluctuation coefficient and pressure fluctuation coefficient. The dynamic temporal features include the temporal change slope of sensor operating state parameters. The parallel diagnostic unit includes a multilayer perceptron subunit and a gated attention network subunit. Static features are input into the multilayer perceptron subunit, which outputs a probability distribution of fault types. Dynamic temporal features are input into the gated attention network subunit, which outputs the correlation between the fault occurrence time and the feature parameters. The result fusion unit sets weight coefficients based on the historical diagnostic accuracy of each subunit according to the various diagnostic results output by the parallel diagnostic unit, and completes the result fusion through a weighted voting method to obtain candidate diagnostic results. It then determines whether to trigger a secondary diagnosis by using a preset consistency check threshold. If the consistency of each preliminary diagnostic result is lower than the threshold, the feature extraction unit is called again to extract additional features and then diagnose again. The fault level determination unit classifies the candidate diagnostic results into levels to obtain fault levels, and encapsulates the fault type, fault level, the corresponding measurement channel position and the fault duration into a fault diagnosis result, which is then transmitted to the decision output module.

[0012] Preferably, the multilayer perceptron subunit adopts a fully connected network architecture consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the dimension of the static features. The hidden layer has two layers, and the number of neurons in each layer is 2 to 3 times that of the input layer and uses the ReLU activation function. The number of neurons in the output layer matches the number of preset fault types and uses the Softmax activation function. The multilayer perceptron subunit calculates the probability distribution of the output fault types through forward propagation. The gated attention network subunit adopts a combined architecture of GRU gated unit and attention mechanism. The gated unit filters effective temporal information, and the attention mechanism focuses on key feature points related to the fault, outputting the fault occurrence time, duration, and correlation values ​​of each temporal feature parameter. The multilayer perceptron subunit and the gated attention network subunit operate independently and in parallel, and their output results are synchronously transmitted to the result fusion unit.

[0013] Preferably, it also includes a power module, a communication module, and a storage module. The power supply module provides a stable operating voltage for the measurement module, data acquisition module, fault diagnosis module, and decision output module; the communication module enables local transmission of early warning signals from the decision output module and the sending of remote alarm signals; and the storage module stores fault diagnosis results and early warning information.

[0014] As can be seen from the above technical solution, compared with the prior art, this invention discloses a gas ultrasonic flow meter system with fault diagnosis. The multi-channel distributed ultrasonic sensor arrangement can comprehensively capture the gas flow state of each channel, making up for the one-sidedness of traditional single-channel data, providing accurate raw data support, and reducing missed diagnoses. The data acquisition module adapts the acquisition mode according to the state, improves data quality through preprocessing, standardizes sound velocity calculation and eliminates pressure fluctuation interference, ensuring the accuracy of the data input to the diagnostic module. The fault diagnosis module collaboratively extracts static and dynamic time-series features, accurately identifies the fault type and locates the time through a parallel diagnostic architecture; the fusion and consistency verification mechanism improves the reliability of diagnosis, avoids the limitations of a single model, and reduces the false diagnosis rate. Furthermore, the system has high adaptability and ease of operation and maintenance. The auxiliary module ensures stable operation, the communication module realizes two-way early warning transmission, and the storage module facilitates data traceability; the diagnostic results include key information such as fault type, level, and location, which can guide precise operation and maintenance, shorten downtime, reduce costs, and improve the safety and reliability of the gas metering system. Attached Figure Description

[0015] 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 The system structure connection diagram provided for this invention. Detailed Implementation

[0017] 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.

[0018] like Figure 1 As shown in the figure, an embodiment of the present invention discloses a gas ultrasonic flow meter system with fault diagnosis, including: a measurement module, a data acquisition module, a fault diagnosis module and a decision output module; The measurement module has a built-in ultrasonic sensor for measuring the flow state of gas in the acoustic channel; The data acquisition module collects measured sound velocity, temperature, pressure, and sensor operating status parameters based on the gas flow state, and performs preprocessing. The fault diagnosis module performs fault diagnosis on the gas ultrasonic flow meter based on the preprocessed data and obtains the fault diagnosis results. The decision output module outputs early warning prompts based on the fault diagnosis results.

[0019] Specifically, the ultrasonic sensor adopts a multi-channel distributed arrangement structure, including at least two sets of paired ultrasonic transducers. Each set of ultrasonic transducers corresponds to a different preset measurement channel on the flow meter body and outputs the corresponding raw echo signal for measuring the flow state of gas in different measurement channels.

[0020] In a specific embodiment of the present invention, for industrial gas pipelines with diameters of DN80-DN1000mm, a 4-channel distributed arrangement is adopted, with the channels symmetrically distributed in an X-shape. The angles between adjacent channels and the pipeline axis are 30° and 60°, and the channel spacing is 1 / 3 of the pipeline diameter. The ultrasonic transducer is a piezoelectric transducer with a working frequency of 1MHz and a bandwidth of 200kHz, and the transmission power is adjustable from 5-20W, which is suitable for common industrial gas media such as natural gas and coalbed methane.

[0021] Each quarter, the length L of each channel is calibrated using a laser rangefinder, and the deviation between the calibrated value and the initial value is recorded. If the deviation exceeds 0.5mm, the L parameter in the fault diagnosis module is automatically updated. The calibration data is synchronously stored in the storage module as a basis for equipment maintenance.

[0022] Specifically, the data acquisition module includes a status recognition unit, a multi-parameter synchronous acquisition unit, and a data preprocessing unit; The state recognition unit is used to receive the raw echo signal from the ultrasonic sensor, extract the time-domain and frequency-domain features of the raw echo signal, and determine the gas flow state type through a preset threshold; the multi-parameter synchronous acquisition unit starts the corresponding acquisition mode for different flow states to obtain raw data; the data preprocessing unit performs outlier removal, data calibration, data normalization, and data compression on the acquired raw data.

[0023] Specifically, the raw data includes measured sound velocity, temperature parameters, pressure parameters, and sensor operating status parameters; The acquisition process for the measured sound velocity includes: a multi-parameter synchronous acquisition unit receiving the transmitted and received signals from each group of ultrasonic transducers, calculating the signal transmission time difference based on a cross-correlation algorithm, and calculating the measured sound velocity in combination with the channel length, as shown in the following formula: ; Where L is the duct length, t 顺 t 逆 These represent the downstream and upstream transmission times of the ultrasonic signal, respectively. The temperature parameter acquisition process includes: the gas medium temperature output by the temperature sensor built into the flow meter pipeline; The pressure parameter acquisition process includes: acquiring the real-time pressure value of the gas medium through a pipeline pressure sensor, and performing pressure stabilization and filtering on the pressure signal during the acquisition process to eliminate pressure fluctuation interference caused by pipeline vibration; The process of acquiring sensor operating status parameters includes: acquiring the ultrasonic transducer's transmission power, received signal amplitude, impedance matching value, and operating voltage and current, while recording the sensor's cumulative operating time to form a basic dataset of sensor health status.

[0024] In a specific embodiment of the present invention, an adaptive threshold algorithm combining a sliding window and statistical learning is adopted, with the window size set to 5s and the threshold updated every 1s. A threshold benchmark model is trained based on normal operating data of the past 30 days, and the deviation between the current flow state parameters (such as flow velocity variance) and the benchmark model is calculated in real time. When the deviation exceeds the dynamic threshold (dynamic threshold = mean of benchmark deviation + 2 × standard deviation of benchmark deviation), it is determined to be a sudden change in flow state, triggering the fast acquisition mode of the multi-parameter synchronous acquisition unit (the acquisition frequency is increased from 10Hz to 50Hz).

[0025] The duct length calibration is performed using a laser rangefinder (measurement accuracy ±0.1mm). Calibration points are marked at both ends of the duct corresponding to the installation positions of the ultrasonic transducers. Measurements are taken quarterly. If the measured value deviates from the initial value by more than 0.5mm, the L value in the formula is automatically corrected.

[0026] Furthermore, using the 3σ criterion, the mean μ and standard deviation σ of the measured sound velocity, temperature, and pressure data are calculated, and data exceeding the range of [μ-3σ, μ+3σ] are discarded; the temperature data are calibrated using the linear calibration formula T. new =a×T orig +b (a and b are calibration coefficients, obtained through standard temperature source calibration), pressure data uses the secondary calibration formula P. new =c×P orig 2 +d×P orig +e (c, d, and e are calibration coefficients, calibrated using a standard pressure source); min-max normalization is used to map the data to the [0, 1] interval, as shown in the formula x. norm =(xx min ) / (x max -x min Wavelet transform compression is used to retain the first 20% of the wavelet coefficients.

[0027] Specifically, the fault diagnosis module includes a feature extraction unit, a parallel diagnosis unit, a result fusion unit, and a fault level determination unit; The feature extraction unit is used to receive preprocessed data and extract static features and dynamic temporal features from the data. The static features include sound velocity deviation rate, temperature fluctuation coefficient and pressure fluctuation coefficient. The dynamic temporal features include the temporal change slope of sensor operating state parameters. The parallel diagnostic unit includes a multilayer perceptron subunit and a gated attention network subunit. Static features are input into the multilayer perceptron subunit, which outputs a probability distribution of fault types. Dynamic temporal features are input into the gated attention network subunit, which outputs the correlation between the fault occurrence time and the feature parameters. The result fusion unit sets weight coefficients based on the historical diagnostic accuracy of each subunit according to the various diagnostic results output by the parallel diagnostic unit, and completes the result fusion through a weighted voting method to obtain candidate diagnostic results. It then determines whether to trigger a secondary diagnosis by using a preset consistency check threshold. If the consistency of each preliminary diagnostic result is lower than the threshold, the feature extraction unit is called again to extract additional features and then diagnose again. The fault level determination unit classifies the candidate diagnostic results into levels to obtain fault levels, and encapsulates the fault type, fault level, the corresponding measurement channel position and the fault duration into a fault diagnosis result, which is then transmitted to the decision output module.

[0028] Specifically, the multilayer sensing subunit adopts a fully connected network architecture consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the dimension of the static features (corresponding to three types of static features: sound velocity deviation rate, temperature fluctuation coefficient, and pressure fluctuation coefficient, with three input neurons). The hidden layer has two layers, with each layer having 2 to 3 times the number of neurons as the input layer and using the ReLU activation function. The number of neurons in the output layer matches the number of preset fault types, and the ReLU activation function is used to achieve nonlinear mapping of static features, enhancing the model's ability to express fault features. The number of neurons in the output layer matches the number of preset fault types (corresponding to four typical faults: sensor performance degradation, sound channel blockage, parameter drift, and multi-channel anomaly, with four output neurons), and the Softmax activation function is used to normalize the output results into a probability distribution.

[0029] The gated attention network subunit adopts a combined architecture of GRU gated unit and attention mechanism. The gated unit filters effective temporal information, and the attention mechanism focuses on key feature points related to the fault, outputting the fault occurrence time, duration, and correlation values ​​of each temporal feature parameter. The multilayer perceptron subunit and the gated attention network subunit operate independently and in parallel, and their output results are synchronously transmitted to the result fusion unit.

[0030] During the model training phase, static features output by the feature extraction unit are used as input samples, along with manually labeled fault types to construct a training set. The network weights and biases are optimized through the backpropagation algorithm to enable the model to fit the mapping relationship between static features and fault types. During actual diagnosis, the static features extracted in real time are input into the trained multilayer sensing unit. After forward propagation calculation, the probability values ​​of various faults are output (sensor performance degradation probability 0.85, sound channel blockage probability 0.12), completing the preliminary determination of the fault type.

[0031] The implementation process of the gated attention network subunit includes: The architecture employs a combination of gating units and an attention mechanism. First, the dynamic temporal features (the temporal change slope of sensor operating state parameters) are processed for temporal alignment to form a fixed-length temporal feature sequence (e.g., a sequence window of 100 sampling points). The gating unit adopts a GRU (Gated Recurrent Unit) structure, which filters valid temporal information and suppresses invalid temporal redundancy (e.g., repeated data during parameter stabilization phases) through reset and update gates. The attention mechanism module calculates the attention weight of each sampling point in the temporal sequence, focusing on key feature points related to faults, such as parameter mutations and sudden slope changes (the higher the weight, the stronger the correlation between the point and the fault).

[0032] During model training, the network is trained using time-series feature sequences as input, coupled with labeled data from the actual fault occurrence times, to optimize the parameters of the gating unit and attention mechanism. During actual diagnosis, the real-time dynamic time-series feature sequences are input into the network. After gating and attention weighting, the start time and duration of the fault are determined using a time-series localization algorithm. Simultaneously, the correlation values ​​of various time-series feature parameters (such as received signal amplitude and transmitted power) are output (higher values ​​indicate a greater impact of the parameter on the fault), completing the time-series localization and feature correlation analysis of the fault. The outputs of both sub-units are synchronously transmitted to the result fusion unit.

[0033] Furthermore, the sound speed deviation rate = |v 实测 -v 理论 | / v 理论 ×100% (v 理论 (The theoretical speed of sound is calculated based on the ideal gas law). Temperature fluctuation coefficient = standard deviation of temperature data / mean temperature; pressure fluctuation coefficient = standard deviation of pressure data / mean pressure.

[0034] Dynamic time series features: The extraction window size is set to 10s, and the slope of the time series change = (the current 10s mean - the previous 10s mean) / 10s.

[0035] Dynamic weight update mechanism: The initial weights are set based on the historical diagnostic accuracy. The weight of the multilayer sensor subunit is w1=0.6, and the weight of the gated attention network subunit is w2=0.4. After every 100 diagnoses, the accuracy of the two subunits under different fault types is calculated (such as the accuracy of subunit w1 and subunit w2 under sensor aging faults). The weighted update formula is used: w1'= w1×(a1 / A), w2' = w2×(a2 / A) (where a1 and a2 are the current accuracy of the two subunits, and A=a1+a2) to ensure that the weights are adapted to the diagnostic needs of different fault types.

[0036] Secondary diagnostic refinement rules: When the consistency of the two output results of the parallel diagnostic unit is less than 80% (consistency = probability intersection / probability union of the same fault type), secondary diagnosis is triggered; two types of features are extracted: the difference of adjacent channel parameters and the frequency features of parameter changes, and then re-input into the parallel diagnostic unit; if the consistency is still less than 80% after secondary diagnosis, a suspected fault result is output and marked as requiring manual review; secondary diagnosis is the upper limit of iteration, and no further iterations are performed.

[0037] The fault severity classification criteria include: Minor fault: Abnormal parameters of a single channel, which does not affect the overall measurement accuracy (measurement error change < ±1%), and the fault duration is < 5 minutes, such as slight fluctuations in the amplitude of the sensor received signal; Moderate fault: Abnormal parameters in 2 or more channels, measurement error variation of ±1% to ±3%, fault duration of 5-30 minutes, such as slight deviation in sensor impedance matching or excessive pressure fluctuation; Severe fault: All channel parameters are abnormal or core parameters (such as sound velocity) change abruptly, measurement error changes > ±3%, fault duration > 30 minutes, such as ultrasonic transducer damage or channel blockage.

[0038] Specifically, it also includes power modules, communication modules, and storage modules. The power supply module provides a stable operating voltage for the measurement module, data acquisition module, fault diagnosis module, and decision output module; the communication module enables local transmission of early warning signals from the decision output module and the sending of remote alarm signals; and the storage module stores fault diagnosis results and early warning information.

[0039] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0040] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A gas ultrasonic flow meter system with fault diagnosis, characterized in that, include: Measurement module, data acquisition module, fault diagnosis module, and decision output module; The measurement module has a built-in ultrasonic sensor for measuring the flow state of gas in the acoustic channel; The data acquisition module collects measured sound velocity, temperature, pressure, and sensor operating status parameters based on the gas flow state, and performs preprocessing. The fault diagnosis module performs fault diagnosis on the gas ultrasonic flow meter based on the preprocessed data and obtains the fault diagnosis results. The decision output module outputs early warning prompts based on the fault diagnosis results.

2. The gas ultrasonic flow meter system with fault diagnosis according to claim 1, characterized in that, The ultrasonic sensor adopts a multi-channel distributed arrangement structure, including at least two sets of paired ultrasonic transducers. Each set of ultrasonic transducers corresponds to a different preset measurement channel on the flow meter body and outputs the corresponding original echo signal for measuring the flow state of gas in different measurement channels.

3. The gas ultrasonic flow meter system with fault diagnosis according to claim 2, characterized in that, The data acquisition module includes a status recognition unit, a multi-parameter synchronous acquisition unit, and a data preprocessing unit; The state recognition unit is used to receive the raw echo signal from the ultrasonic sensor, extract the time-domain and frequency-domain features of the raw echo signal, and determine the gas flow state type through a preset threshold; the multi-parameter synchronous acquisition unit starts the corresponding acquisition mode for different flow states; the data preprocessing unit performs outlier removal, data calibration, data normalization, and data compression on the acquired data.

4. A gas ultrasonic flow meter system with fault diagnosis according to claim 3, characterized in that, The acquisition process of the measured sound velocity includes: a multi-parameter synchronous acquisition unit receives the transmitted and received signals from each group of ultrasonic transducers, calculates the signal transmission time difference based on a cross-correlation algorithm, and calculates the measured sound velocity by combining the channel length, as shown in the following formula: ； Where L is the duct length, t 顺 t 逆 These represent the downstream and upstream transmission times of the ultrasonic signal, respectively. The temperature parameter acquisition process includes: the gas medium temperature output by the temperature sensor built into the flow meter pipeline; The pressure parameter acquisition process includes: acquiring the real-time pressure value of the gas medium through a pipeline pressure sensor, and performing pressure stabilization and filtering on the pressure signal during the acquisition process to eliminate pressure fluctuation interference caused by pipeline vibration; The process of acquiring sensor operating status parameters includes: acquiring the ultrasonic transducer's transmission power, received signal amplitude, impedance matching value, and operating voltage and current, while recording the sensor's cumulative operating time to form a basic dataset of sensor health status.

5. A gas ultrasonic flow meter system with fault diagnosis according to claim 3, characterized in that, The fault diagnosis module includes a feature extraction unit, a parallel diagnosis unit, a result fusion unit, and a fault level determination unit. The feature extraction unit is used to receive preprocessed data and extract static features and dynamic temporal features from the data. The static features include sound velocity deviation rate, temperature fluctuation coefficient and pressure fluctuation coefficient. The dynamic temporal features include the temporal change slope of sensor operating state parameters. The parallel diagnostic unit includes a multilayer perceptron subunit and a gated attention network subunit. Static features are input into the multilayer perceptron subunit, which outputs a probability distribution of fault types. Dynamic temporal features are input into the gated attention network subunit, which outputs the correlation between the fault occurrence time and the feature parameters. The result fusion unit sets weight coefficients based on the historical diagnostic accuracy of each subunit according to the various diagnostic results output by the parallel diagnostic unit, and completes the result fusion through a weighted voting method to obtain candidate diagnostic results. It then determines whether to trigger a secondary diagnosis by using a preset consistency check threshold. If the consistency of each preliminary diagnostic result is lower than the threshold, the feature extraction unit is called again to extract additional features and then diagnose again. The fault level determination unit classifies the candidate diagnostic results into levels to obtain fault levels, and encapsulates the fault type, fault level, the corresponding measurement channel position and the fault duration into a fault diagnosis result, which is then transmitted to the decision output module.

6. A gas ultrasonic flow meter system with fault diagnosis according to claim 5, characterized in that, The multilayer perceptron subunit adopts a fully connected network architecture consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the dimension of the static features. The hidden layer has two layers, with each layer having 2 to 3 times the number of neurons as the input layer and using the ReLU activation function. The number of neurons in the output layer matches the number of preset fault types and uses the Softmax activation function. The multilayer perceptron subunit calculates the probability distribution of the output fault types through forward propagation. The gated attention network subunit adopts a combined architecture of GRU gated unit and attention mechanism. The gated unit filters effective temporal information, and the attention mechanism focuses on key feature points related to the fault, outputting the fault occurrence time, duration, and correlation values ​​of each temporal feature parameter. The multilayer perceptron subunit and the gated attention network subunit operate independently and in parallel, and their output results are synchronously transmitted to the result fusion unit.

7. A gas ultrasonic flow meter system with fault diagnosis according to claim 1, characterized in that, It also includes a power module, a communication module, and a storage module. The power supply module provides a stable operating voltage for the measurement module, data acquisition module, fault diagnosis module, and decision output module; the communication module enables local transmission of early warning signals from the decision output module and the sending of remote alarm signals; and the storage module stores fault diagnosis results and early warning information.

Images