Valve fault data detection method and system

Abstract

The invention discloses a data detection method and system for a valve fault, and relates to the technical field of fault detection, and the method comprises the steps: obtaining multi-path sensing data in a valve operation state, carrying out the time synchronization and the delay alignment processing based on the valve working condition of the multi-path sensing data, carrying out the spatial coherence analysis of a multi-path signal set subjected to the space-time alignment, and carrying out the detection of the valve fault. Constructing a coherence relation matrix; by taking the coherence relation matrix as a transformation constraint, carrying out coherence-preserving spatial transformation processing on the multi-path signal set subjected to space-time alignment; adaptive screening based on valve fault features is carried out on the converted signal representation to obtain an enhanced signal representation; according to the method, the coherent relation matrix capable of describing the related cooperative relation between the signals and the valve fault is extracted and constructed from multiple paths of signals subjected to space-time alignment, and targeted transformation of the signal space is carried out by taking the coherent relation matrix as a constraint; the method can effectively enhance the aggregation and separability of the weak cooperative features representing the multi-physics coupling fault in the transform domain.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology, and in particular to a data detection method and system for valve faults. Background Technology

[0002] In the field of industrial valve condition monitoring based on multi-sensor information fusion, existing fault detection methods typically rely on independent analysis of signals from a single type of sensor, or simple time-domain and frequency-domain feature extraction and threshold comparison of multiple signals. The core of these methods lies in capturing the abnormal state of the valve from a single physical dimension.

[0003] However, as a mechanical system consisting of a drive component, a sealing component, and a fluid passage, the potential failures of a valve, such as valve stem jamming, sealing surface wear, and valve body leakage, are essentially the result of the coupling effect of multiple physical fields such as mechanical stress, fluid dynamics, and thermodynamics. Its early and weak fault characteristics are often not fully characterized by a single sensor signal.

[0004] Existing methods cannot effectively uncover and utilize the collaborative change patterns hidden in temporal correlation and spatial distribution caused by the same fault root cause among different sensor signals. This leads to problems such as insensitivity to early fault response, difficulty in distinguishing compound faults, and high false alarm rate under high background noise, which limits the accuracy and reliability of predictive maintenance of valves. Summary of the Invention

[0005] This invention provides a data detection method and system for valve faults to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a data detection method for valve faults, comprising:

[0007] S1. Acquire multi-channel sensor data under valve operating conditions, perform time synchronization and delay alignment processing on the multi-channel sensor data based on valve operating conditions, and obtain a spatiotemporally aligned multi-channel signal set;

[0008] S2. Perform spatial coherence analysis on the spatiotemporally aligned multi-channel signal set to obtain the coherence relationship between each signal and the valve fault, and construct the coherence relationship matrix;

[0009] S3. Using the coherence relation matrix as a transformation constraint, perform a coherence-preserving spatial transformation on the spatiotemporally aligned multi-channel signal set to obtain the transformed signal representation;

[0010] S4. Perform adaptive filtering based on valve fault characteristics on the transformed signal representation to obtain the enhanced signal representation;

[0011] S5. Perform time-frequency domain feature extraction on the enhanced signal representation to obtain a fault feature set related to the valve fault type;

[0012] S6. Perform feature evolution trend analysis on the fault feature set to match the valve structural damage. When the feature evolution trend shows a deterministic law related to the valve structural damage, generate and output the valve fault identification information.

[0013] Preferably, the time synchronization of multi-channel sensor data includes:

[0014] It receives a unified timing signal and generates a corresponding timestamp for each data point in the multi-channel sensing data from different sensors based on the signal.

[0015] Based on the timestamps, the data streams from all sensors are aligned to the same time base to obtain time-synchronized multi-channel data streams.

[0016] Preferably, the time-delay alignment process based on valve operating conditions includes:

[0017] Determine the start time of valve operating condition changes based on valve opening and closing signals;

[0018] Using the start time as the alignment point, time shift compensation is performed on the time-synchronized multi-channel data streams to align the signal segments corresponding to the same valve action in each data stream in time.

[0019] The time-shift compensated data streams are segmented and spliced ​​to obtain a spatiotemporally aligned multi-channel signal set.

[0020] Preferably, the step of performing spatial coherence analysis on the spatiotemporally aligned multi-channel signal set to obtain the coherence relationship between each signal and the valve fault includes:

[0021] Based on the start time of the valve operating condition change, a signal window for coherence analysis is determined on a spatiotemporally aligned multi-channel signal set;

[0022] Within the signal window, the coherence degree calculation based on the cross power spectral density is performed on any two signals to obtain the corresponding coherence coefficients;

[0023] Based on the magnitude of the coherence coefficients, the pairing relationships that characterize the strong linkage between signals are extracted and used as the coherence relationships related to valve failure.

[0024] Preferably, constructing the coherence relation matrix includes:

[0025] Using each signal as a row and column of a matrix, the coherence relationship related to valve failure is mapped to the element at the corresponding position in the matrix;

[0026] Based on the correlation priority between the coherence relationship and different valve failure modes, the order of elements in the matrix is ​​adjusted by weighting.

[0027] After mapping and sorting all coherent relationships, a coherent relationship matrix is ​​formed, which is based on the contribution of failure modes.

[0028] Preferably, the spatial transformation processing of the spatiotemporally aligned multi-channel signal set, using the coherence relation matrix as a transformation constraint, includes:

[0029] The coherence relation matrix is ​​decomposed into eigenvalues ​​to obtain the eigenvalues ​​and eigenvectors of the coherence relation matrix, and the eigenvectors are used as basis vectors to guide the transformation.

[0030] Based on the basis vectors, a linear projection is performed on the spatiotemporally aligned set of multiple signals, mapping the multiple signals at each time moment to a coordinate point in the space spanned by the basis vectors.

[0031] The coordinates of all points at all times are collected to form the transformed signal representation.

[0032] Preferably, the method further includes optimizing and recombining the transformed signal representation, including:

[0033] Based on the basis vectors and the coherence relation matrix, the matrix elements are weighted and synthesized by using the components of the basis vectors as weights to obtain the comprehensive coherence strength value of each basis vector.

[0034] All basis vectors and their corresponding coordinate components are reordered according to the order of comprehensive coherence strength from strongest to weakest.

[0035] The reordered coordinate component sequence is used to construct an optimized transformed signal representation.

[0036] Preferably, the step of adaptively filtering the transformed signal representation based on valve fault characteristics to obtain the enhanced signal representation includes:

[0037] The energy contribution distribution is obtained by analyzing the energy contribution of each sequential segment, sorted by comprehensive coherence strength, to the overall signal in the optimized transformed signal representation.

[0038] Based on the trend of energy contribution distribution with ranking position, the transition segment from concentrated distribution to dispersed distribution of energy contribution is identified;

[0039] Extract the signal components corresponding to all sequential segments before the transition segment to form the enhanced signal representation.

[0040] Preferably, the feature evolution trend analysis of the fault feature set matching the valve structural damage includes:

[0041] Arrange each feature in the fault feature set in chronological order to form a feature value time series;

[0042] For each feature value time series, divide it into consecutive time windows, extract the representative value of the feature value distribution within each time window, and calculate the difference between the representative values ​​of the distribution distribution of adjacent time windows to obtain the statistical difference value;

[0043] Analyze the trend of statistical difference values ​​over time. If the trend shows a monotonically increasing trend, it is determined that the feature has an evolutionary trend related to structural damage.

[0044] To address the above problems, the present invention also provides a data detection system for valve faults, the system comprising:

[0045] The data processing module is used to acquire multi-channel sensor data under the valve's operating status, perform time synchronization and delay alignment processing based on the valve's operating conditions on the multi-channel sensor data, and obtain a spatiotemporally aligned multi-channel signal set.

[0046] The spatial coherence analysis module performs spatial coherence analysis on a spatiotemporally aligned set of multiple signals to obtain the coherence relationships between each signal and the valve fault and constructs a coherence relationship matrix.

[0047] The signal transformation module is used to perform coherence-preserving spatial transformation processing on a spatiotemporally aligned set of signals, using the coherence relation matrix as the transformation constraint, to obtain the transformed signal representation.

[0048] An adaptive filtering module is used to adaptively filter the transformed signal representation based on valve fault characteristics to obtain an enhanced signal representation.

[0049] The feature extraction module is used to extract time-frequency domain features from the enhanced signal representation to obtain a fault feature set related to the valve fault type;

[0050] The evolution analysis module is used to perform feature evolution trend analysis on the fault feature set and match it with valve structural damage. When the feature evolution trend shows a deterministic law related to valve structural damage, it generates and outputs valve fault identification information.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. By extracting and constructing a coherent relation matrix from spatiotemporally aligned multi-channel signals to characterize the collaborative relationship between signals and valve faults, and using this matrix as a constraint for targeted transformation of the signal space, the aggregation and separability of weak collaborative features characterizing multi-physics coupled faults in the transform domain can be effectively enhanced. This significantly improves the signal-to-noise ratio of the input signal for subsequent fault feature extraction, laying a reliable data foundation for accurate identification of early composite valve faults.

[0053] 2. By adopting an adaptive component screening method based on dynamic identification of transition zones according to energy distribution trends, and combining it with parametric statistical tests to determine the feature evolution trend, the method can accurately separate early fault components of valves and provide objective trend warnings. At the same time, it can also make the entire detection system more adaptable to operating conditions, significantly reduce the risk of misjudgment caused by individual differences or changes in operating conditions, and improve the universality and reliability of the method. Attached Figure Description

[0054] Figure 1 A flowchart of a data detection method for valve faults provided by the present invention;

[0055] Figure 2 A module structure diagram of a data detection system for valve faults provided by the present invention;

[0056] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0057] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0058] Example 1, referring to Figure 1 The diagram shown is a flowchart illustrating a data detection method for valve faults according to an embodiment of the present invention. In this embodiment, the data detection method for valve faults includes:

[0059] S1. Acquire multi-channel sensor data under valve operating conditions, perform time synchronization and delay alignment processing on the multi-channel sensor data based on valve operating conditions, and obtain a spatiotemporally aligned multi-channel signal set;

[0060] S2. Perform spatial coherence analysis on the spatiotemporally aligned multi-channel signal set to obtain the coherence relationship between each signal and the valve fault, and construct the coherence relationship matrix;

[0061] S3. Using the coherence relation matrix as a transformation constraint, perform a coherence-preserving spatial transformation on the spatiotemporally aligned multi-channel signal set to obtain the transformed signal representation;

[0062] S4. Perform adaptive filtering based on valve fault characteristics on the transformed signal representation to obtain the enhanced signal representation;

[0063] S5. Perform time-frequency domain feature extraction on the enhanced signal representation to obtain a fault feature set related to the valve fault type;

[0064] S6. Perform feature evolution trend analysis on the fault feature set to match the valve structural damage. When the feature evolution trend shows a deterministic law related to the valve structural damage, generate and output the valve fault identification information.

[0065] In embodiments of the present invention, time synchronization of multi-channel sensor data includes:

[0066] It receives a unified timing signal and generates a corresponding timestamp for each data point in the multi-channel sensing data from different sensors based on the signal.

[0067] Based on the timestamps, the data streams from all sensors are aligned to the same time base to obtain time-synchronized multi-channel data streams.

[0068] In embodiments of the present invention, the delay alignment processing based on valve operating conditions includes:

[0069] Determine the start time of valve operating condition changes based on valve opening and closing signals;

[0070] Using the start time as the alignment point, time shift compensation is performed on the time-synchronized multi-channel data streams to align the signal segments corresponding to the same valve action in each data stream in time.

[0071] The time-shift compensated data streams are segmented and spliced ​​to obtain a spatiotemporally aligned multi-channel signal set.

[0072] In one embodiment, taking the operation monitoring scenario of an industrial chemical pipeline gate valve as an example, multi-channel sensor data output from the valve body vibration sensor, valve stem torque sensor, valve cavity pressure sensor, and sealing surface temperature sensor are first collected.

[0073] Simultaneously, the raw data transmission streams of each sensor are collected to ensure that each sensor data stream contains continuous physical quantity detection values ​​and raw data acquisition time information, thereby completing the acquisition of multiple sensor data under valve operating conditions.

[0074] Then, the unified timing signal from the BeiDou satellite is received and used as a time reference to generate a unique timestamp for each data point in the multi-channel sensor data output by the valve body vibration sensor, valve stem torque sensor, valve cavity pressure sensor, and sealing surface temperature sensor. The timestamp can contain time information such as year, month, day, hour, minute, second, and millisecond.

[0075] Furthermore, based on the timestamps corresponding to each data point, the data streams of the valve body vibration sensor, valve stem torque sensor, valve cavity pressure sensor, and sealing surface temperature sensor are all aligned to the same time reference of BeiDou satellite timing, so that different sensors have corresponding detection data at the same time node, resulting in time-synchronized multi-channel data streams.

[0076] Furthermore, the opening and closing action electrical signal emitted by the electric actuator of the gate valve in the industrial chemical pipeline is extracted as the valve's opening and closing action signal, and the starting time of the valve's operating condition change is determined by identifying the level transition node of this opening and closing action electrical signal.

[0077] Furthermore, taking this starting moment as the alignment point, based on the actual transmission delay and response delay of the signals detected by the valve body vibration sensor, valve stem torque sensor, valve cavity pressure sensor, and sealing surface temperature sensor, the time-shift compensation operation is performed on the multiple data streams after time synchronization, so that the signal segments in each data stream corresponding to the same opening and closing action of the gate valve completely overlap in the time dimension.

[0078] Finally, for each data stream after time-shift compensation, signal segments are extracted according to the time period of a single opening and closing action of the gate valve. Then, the extracted signal segments corresponding to multiple consecutive opening and closing actions are spliced ​​together in chronological order to obtain a spatiotemporally aligned multi-channel signal set containing the continuous operating state of the gate valve.

[0079] In summary, this solution ensures that multi-channel sensor data from different physical locations and with different response characteristics are aligned under a strictly unified spatiotemporal reference through time synchronization and delay alignment based on valve operating conditions.

[0080] This solution effectively eliminates non-fault-related time delays and phase deviations introduced by differences in sensor sampling clocks, different signal transmission paths, and valve mechanical action response delays. As a result, it obtains a set of multi-physics field signals that can truly and synchronously reflect the valve's same operating state at any instant. This provides a crucial high-quality data foundation for subsequent accurate analysis of the intrinsic coherence and co-evolution patterns between signals, fundamentally avoiding distortion and misjudgment in correlation analysis caused by asynchronous data.

[0081] In embodiments of the present invention, spatial coherence analysis is performed on a spatiotemporally aligned set of multiple signals to obtain the coherence relationships between the signals related to valve faults, including:

[0082] Based on the start time of the valve operating condition change, a signal window for coherence analysis is determined on a spatiotemporally aligned multi-channel signal set;

[0083] Within the signal window, the coherence degree calculation based on the cross power spectral density is performed on any two signals to obtain the corresponding coherence coefficients;

[0084] Based on the magnitude of the coherence coefficients, the pairing relationships that characterize the strong linkage between signals are extracted and used as the coherence relationships related to valve failure.

[0085] In embodiments of the present invention, constructing a coherence relation matrix includes:

[0086] Using each signal as a row and column of a matrix, the coherence relationship related to valve failure is mapped to the element at the corresponding position in the matrix;

[0087] Based on the correlation priority between the coherence relationship and different valve failure modes, the order of elements in the matrix is ​​adjusted by weighting.

[0088] After mapping and sorting all coherent relationships, a coherent relationship matrix is ​​formed, which is based on the contribution of failure modes.

[0089] In one embodiment, taking the industrial chemical pipeline gate valve scenario as an example, based on the starting time of the change in the operating conditions of the industrial chemical pipeline gate valve, the stable signal segment when the gate valve is not in operation is intercepted forward as the front window, and the stable signal segment after the gate valve completes one opening and closing action is intercepted backward as the back window. The front window and the back window are combined to form a signal window for coherence analysis. This signal window only contains the spatiotemporally aligned gate valve body vibration signal, valve stem torque signal, valve cavity pressure signal, and sealing surface temperature signal.

[0090] Furthermore, within this signal window, any two signals are selected. First, all detection data of the two signals within the window are obtained separately. Then, the covariance of the two signals is obtained by calculating the product of the two signals point by point and accumulating the results.

[0091] Simultaneously, the sum of squares of the detection data of each signal is calculated and accumulated to obtain the variance of each of the two signals. The covariance of the two signals is divided by the square root of the product of the variances of the two signals to obtain the coherence coefficient of the corresponding two signals.

[0092] The coherence calculations are performed sequentially between the valve body vibration signal and the valve stem torque signal, the valve body vibration signal and the valve cavity pressure signal, the valve body vibration signal and the sealing surface temperature signal, the valve stem torque signal and the valve cavity pressure signal, the valve stem torque signal and the sealing surface temperature signal, and the valve cavity pressure signal and the sealing surface temperature signal, resulting in six sets of corresponding coherence coefficients.

[0093] Furthermore, through multiple rounds of calibration using measured data from normal and faulty operation of industrial chemical pipeline gate valves, a coherence coefficient standard for distinguishing strong and weak linkages between signals was determined. The calculated coherence coefficients were compared with this standard. If the coherence coefficients of two signals reached the standard, the pairing relationship between the two signals was extracted as the coherence relationship related to the gate valve fault. Finally, all signal pairing relationships that meet the requirements were obtained.

[0094] Furthermore, the gate valve body vibration signal, valve stem torque signal, valve cavity pressure signal, and sealing surface temperature signal, which are spatiotemporally aligned and aggregated from multiple signals, are used as rows and columns of a matrix, respectively. The intersection of each row and column corresponds to the pairing relationship between two signals. The extracted coherent relationships related to the gate valve fault are mapped to the corresponding intersection positions in the matrix, and intersection positions without coherent relationships are marked as unrelated.

[0095] Furthermore, common failure modes of industrial chemical pipeline gate valves were preset, including valve body leakage, valve stem jamming, and sealing surface wear. Through multiple rounds of actual measurements, the correlation between each coherence relationship and the three failure modes was determined. Among them, the coherence relationship between valve cavity pressure signal and sealing surface temperature signal was most correlated with valve body leakage, the coherence relationship between valve stem torque signal and valve body vibration signal was most correlated with valve stem jamming, and the coherence relationship between valve body vibration signal and sealing surface temperature signal was most correlated with sealing surface wear.

[0096] The order of rows and columns in the matrix is ​​adjusted by weighting based on the correlation priority, so that the rows and columns corresponding to high-priority correlations are placed at the beginning of the matrix.

[0097] After completing the mapping and sorting adjustment of all coherent relationships, a coherent relationship matrix is ​​formed, which is based on the contribution of three failure modes: valve body leakage, valve stem jamming, and sealing surface wear. This matrix clearly shows the degree of correlation between the coherent relationships of each signal and different failure modes of the gate valve.

[0098] It should be noted that this solution systematically extracts the coherent relationships characterizing the coordinated changes between signals by performing spatial coherence analysis on spatiotemporally aligned multi-channel signal sets. This effectively achieves a key leap from monitoring raw physical quantities to mining fault correlation patterns, thereby accurately identifying and quantifying the coordinated response patterns caused by damage to the same valve structure, such as valve stem jamming or sealing surface wear, on different sensors such as vibration, torque, and temperature. This transforms the implicit fault correlations in multidimensional signals into an explicit and quantifiable coherent relationship network.

[0099] Furthermore, the coherence relation matrix constructed based on the extracted coherence relations provides a core structured feature representation for the entire fault diagnosis process. It not only objectively records the fault-related coupling strength between multi-physics field signals, but also provides a data foundation that can centrally reflect the contribution of fault modes through matrix organization. This enables subsequent transformations, screening, and feature extraction to directly revolve around fault-related collaborative modes, greatly improving the pertinence and efficiency of the diagnosis process.

[0100] In embodiments of the present invention, a spatial transformation process that preserves coherence is performed on a spatiotemporally aligned set of multiple signals, using the coherence relation matrix as a transformation constraint. This process includes:

[0101] The coherence relation matrix is ​​decomposed into eigenvalues ​​to obtain the eigenvalues ​​and eigenvectors of the coherence relation matrix, and the eigenvectors are used as basis vectors to guide the transformation.

[0102] Based on the basis vectors, a linear projection is performed on the spatiotemporally aligned set of multiple signals, mapping the multiple signals at each time moment to a coordinate point in the space spanned by the basis vectors.

[0103] The coordinates of all points at all times are collected to form the transformed signal representation.

[0104] In embodiments of the present invention, the method further includes optimizing and recombining the transformed signal representation, including:

[0105] Based on the basis vectors and the coherence relation matrix, the matrix elements are weighted and synthesized by using the components of the basis vectors as weights to obtain the comprehensive coherence strength value of each basis vector.

[0106] All basis vectors and their corresponding coordinate components are reordered according to the order of comprehensive coherence strength from strongest to weakest.

[0107] The reordered coordinate component sequence is used to construct an optimized transformed signal representation.

[0108] Specifically, the overall coherence strength value is calculated using the following formula:

[0109]

[0110] In the formula, Indicates the first basis vectors The corresponding composite coherence strength value. This scalar value quantifies the degree to which the signal mode represented by the basis vector dominates the entire signal set; a larger value indicates that the mode is more significant and more correlated with the fault.

[0111] Represents the eigenvalue decomposition obtained by the first eigenvalue decomposition. basis vectors Indicates the first The contribution weight of each sensor signal in this mode. Represents the coherence relation matrix, matrix elements Indicates the first Road and the First The coherence coefficient between the signals, Indicates the total number of sensors. This represents the matrix transpose operation.

[0112] In one embodiment, the coherence relation matrix corresponding to the gate valve of the industrial chemical pipeline is decomposed into eigenvalues. First, the set of eigenvalues ​​and the set of eigenvectors of the matrix are decomposed. Then, each eigenvector in the set of eigenvectors is directly used as the basis vector to guide the spatial transformation. The dimension of the basis vector is consistent with the number of channels of the gate valve sensing data. Each component in the basis vector corresponds to the transformation weight of different sensor signals of the gate valve.

[0113] Furthermore, using the basis vector as the projection reference, linear projection processing is performed on the spatiotemporally aligned multi-channel signal set of the industrial chemical pipeline gate valve to extract the detection values ​​of four signals at each moment in the signal set: valve body vibration, valve stem torque, valve cavity pressure, and sealing surface temperature.

[0114] Then, the detected value is linearly weighted and summed with the basis vectors to obtain a coordinate point in the space spanned by the basis vectors at that moment. The dimension of the coordinate point is consistent with the number of basis vectors.

[0115] Furthermore, the detection values ​​of all moments in the spatiotemporal aligned multi-channel signal set of the industrial chemical pipeline gate valve are extracted to complete the linear projection operation. All the obtained coordinate points are arranged in chronological order and directly aggregated to form the transformed signal representation. This representation contains the coordinate information and time correlation information of the gate valve at each moment in the transformed space.

[0116] Furthermore, the comprehensive coherence strength value is calculated using a formula, where the components of the basis vector are the contribution weights of different sensor signals of the gate valve in the industrial chemical pipeline, the elements of the coherence relation matrix are the coherence coefficients between the two signals corresponding to the gate valve, and the total number of sensors is the number of sensor data channels configured for the gate valve.

[0117] It should be noted that this comprehensive coherence strength value can quantify the dominance of the signal mode represented by a single basis vector of the gate valve in the overall signal set of the industrial chemical pipeline. The larger the value, the more significant the corresponding signal mode is in the gate valve signal set and the stronger its correlation with the gate valve fault. In other words, the magnitude of this value can directly determine the correlation between each signal mode of the gate valve and the fault.

[0118] Furthermore, the comprehensive coherence intensity values ​​corresponding to all basis vectors of the industrial chemical pipeline gate valve are compared, and the basis vectors are rearranged in order of strength to weakness. At the same time, the coordinate components corresponding to each basis vector in the transformed signal representation are synchronously reordered to maintain the correspondence between basis vectors and coordinate components.

[0119] Finally, the reordered coordinate components are arranged into a continuous sequence according to time order, and directly constructed into an optimized transformed signal representation. This representation prioritizes the retention of signal pattern information with strong correlation to faults in industrial chemical pipeline gate valves, removes signal pattern information with weak correlation, and only includes the sorted coordinate components and time correlation information of the gate valve at each time.

[0120] In summary, this scheme uses the coherence relation matrix as a transformation constraint for spatial transformation processing, which realizes the targeted extraction and enhancement of fault-related cooperative modes in multiple signals. It maps the original signal to a new space dominated by fault coupling features, so that the weak correlation signal components that were originally scattered in the time domain and different sensor channels can be gathered and amplified in the transform domain, thereby significantly improving the signal-to-noise ratio and identifiability of fault features.

[0121] Furthermore, this coherence-preserving transformation not only enhances fault characteristics but also generates a structured transformed signal representation with clear physical meaning. In this representation, signal components are organized in an orderly manner according to their correlation strength with the fault cooperative mode, providing a data foundation for subsequent adaptive screening steps that can be efficiently operated based on fault contribution, thus optimizing the efficiency and accuracy of fault feature separation and selection.

[0122] In embodiments of the present invention, the transformed signal representation is adaptively filtered based on valve fault characteristics to obtain an enhanced signal representation, including:

[0123] The energy contribution distribution is obtained by analyzing the energy contribution of each sequential segment, sorted by comprehensive coherence strength, to the overall signal in the optimized transformed signal representation.

[0124] Based on the trend of energy contribution distribution with ranking position, the transition segment from concentrated distribution to dispersed distribution of energy contribution is identified;

[0125] Extract the signal components corresponding to all sequential segments before the transition segment to form the enhanced signal representation.

[0126] Among them, energy contribution rate Calculated using the following formula:

[0127]

[0128] In the formula, Indicates the first Each sequential segment contains a series of continuous coordinate components in the optimized transformed signal representation. The optimized transformed signal representation represents the first... Each coordinate component Representing its energy, Indicates the first The energy contribution rate of each sequential segment, that is, the proportion of energy in that segment to the total energy. This represents the summation over all coordinate components.

[0129] In one embodiment, for the optimized and transformed signal representation of gate valves in industrial chemical pipelines, the energy contribution rate of each sequential segment sorted by comprehensive coherence intensity is first calculated according to the above formula, and then the energy contribution rates of all segments are sorted according to their sorting positions to obtain the energy contribution distribution.

[0130] Specifically, the energy contribution rate can quantify the proportion of each signal segment of the gate valve in an industrial chemical pipeline to the overall signal energy. Moreover, this value gradually decreases as the ranking position moves forward. That is, the earlier the ranking segment is, the more significant its energy support for the gate valve fault-related signals. The change characteristics of this value can be used as the basis for identifying the gate valve fault-related signal segments.

[0131] Therefore, based on the energy contribution distribution of gate valves in industrial chemical pipelines, the energy contribution rate changes of adjacent sections are compared sequentially along the sorting position. When it is found that the decrease in energy contribution rate suddenly increases from a gradual decrease, the position is determined to be the transition section where energy contribution transitions from a concentrated distribution to a dispersed distribution.

[0132] Finally, the coordinate components and time correlation information of all sequential segments before the turning section of the gate valve in the industrial chemical pipeline are extracted. These signal components are then integrated according to their original order and time sequence to form an enhanced signal representation that retains only the highly correlated signal components of the gate valve fault.

[0133] In summary, this solution achieves accurate and automatic separation and enhancement of fault-related signal components by adaptively filtering the transformed signal representation based on valve fault characteristics. Furthermore, the filtering process dynamically determines the retention interval based on the inherent statistical characteristics of the signal in the transform domain, without relying on preset thresholds or fixed templates. This effectively eliminates noise and interference components unrelated to valve structural damage, significantly improving the signal-to-noise ratio and purity of fault characteristics.

[0134] Furthermore, this adaptive filtering mechanism not only enhances the quality of signal representation but also enables it to generalize to different valve types, operating conditions, and fault modes. In the enhanced signal representation obtained after filtering, fault features are more concentrated and prominent, providing highly purified and targeted input for subsequent time-frequency domain feature extraction, thereby improving the accuracy of fault feature extraction and the overall reliability of the subsequent diagnostic process.

[0135] In embodiments of the present invention, feature evolution trend analysis of the fault feature set matching valve structural damage is performed, including:

[0136] Arrange each feature in the fault feature set in chronological order to form a feature value time series;

[0137] For each feature value time series, divide it into consecutive time windows, extract the representative value of the feature value distribution within each time window, and calculate the difference between the representative values ​​of the distribution distribution of adjacent time windows to obtain the statistical difference value;

[0138] Analyze the trend of statistical difference values ​​over time. If the trend shows a monotonically increasing trend, it is determined that the feature has an evolutionary trend related to structural damage.

[0139] In one embodiment, each feature in the fault feature set of the gate valve of the industrial chemical pipeline is arranged sequentially according to the time information of its collection, forming a feature value time series. Each feature value time series corresponds to a fault feature of the gate valve, such as the torque fluctuation feature corresponding to valve stem jamming and the pressure abnormality feature corresponding to valve body leakage. Each series contains continuous feature values ​​and corresponding time node information.

[0140] Furthermore, the time series of each characteristic value of the gate valve in the industrial chemical pipeline can be divided into continuous time windows, and the duration of the time window can be set according to the normal opening and closing cycle of the gate valve to ensure that each window contains complete gate valve action characteristic data.

[0141] Then, the average value of all feature values ​​within each time window is extracted as the distribution representative value. The distribution representative value of the next time window is subtracted from the distribution representative value of the previous adjacent time window, and the statistical difference value corresponding to all adjacent windows is calculated one by one.

[0142] Furthermore, all statistical difference values ​​of industrial chemical pipeline gate valves are arranged in chronological order according to the corresponding time windows, and the size of adjacent statistical difference values ​​is compared one by one. If the later statistical difference value is always greater than the previous statistical difference value, that is, the statistical difference value continues to increase as the time window progresses and shows a monotonically increasing trend, then it is determined that this feature has an evolutionary trend related to gate valve structural damage.

[0143] Finally, when the evolution trend of the gate valve in the industrial chemical pipeline exhibits a deterministic pattern related to structural damage, the valve fault identification information, which includes the fault location, fault type, and evolution trend, is generated by combining the fault type corresponding to this feature. This identification information is then transmitted to the gate valve monitoring terminal to complete the output of the valve fault identification information.

[0144] In summary, this solution achieves a key leap from static feature identification to dynamic damage process determination by analyzing the feature evolution trend of the fault feature set and matching it with valve structural damage, and generating fault identification information when the feature evolution trend shows a deterministic pattern.

[0145] By tracking the evolution pattern of characteristic values ​​over time, it is possible to effectively distinguish between random changes caused by fluctuations in normal operating conditions and directional, continuous deterioration trends caused by structural damage. This enables early warning and deterministic diagnosis of potential valve failures based on continuous monitoring of valve health status.

[0146] Furthermore, this solution is based on an evolutionary analysis mechanism using parameter-free statistical trend criteria. It can objectively and adaptively determine whether features exhibit deterministic evolutionary patterns related to structural damage without the need for preset fixed thresholds and damage templates. This significantly reduces the reliance on expert prior knowledge and historical fault data, and effectively avoids missed or false alarms caused by improper threshold settings. It also significantly improves the objectivity and reliability of fault warning conclusions and the system's generalization ability in different application scenarios.

[0147] Example 2, as Figure 2 The diagram shown is a module structure diagram of a data detection system for valve faults provided by the present invention, which includes:

[0148] The data processing module is used to acquire multi-channel sensor data under the valve's operating status, perform time synchronization and delay alignment processing based on the valve's operating conditions on the multi-channel sensor data, and obtain a spatiotemporally aligned multi-channel signal set.

[0149] The spatial coherence analysis module performs spatial coherence analysis on a spatiotemporally aligned set of multiple signals to obtain the coherence relationships between each signal and the valve fault and constructs a coherence relationship matrix.

[0150] The signal transformation module is used to perform coherence-preserving spatial transformation processing on a spatiotemporally aligned set of signals, using the coherence relation matrix as the transformation constraint, to obtain the transformed signal representation.

[0151] An adaptive filtering module is used to adaptively filter the transformed signal representation based on valve fault characteristics to obtain an enhanced signal representation.

[0152] The feature extraction module is used to extract time-frequency domain features from the enhanced signal representation to obtain a fault feature set related to the valve fault type;

[0153] The evolution analysis module is used to perform feature evolution trend analysis on the fault feature set and match it with valve structural damage. When the feature evolution trend shows a deterministic law related to valve structural damage, it generates and outputs valve fault identification information.

[0154] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A data detection method for valve faults, characterized in that, The method includes: S1. Acquire multi-channel sensor data under valve operating conditions, perform time synchronization and delay alignment processing on the multi-channel sensor data based on valve operating conditions, and obtain a spatiotemporally aligned multi-channel signal set; S2. Perform spatial coherence analysis on the spatiotemporally aligned multi-channel signal set to obtain the coherence relationship between each signal and the valve fault, and construct the coherence relationship matrix; S3. Using the coherence relation matrix as a transformation constraint, perform a coherence-preserving spatial transformation on the spatiotemporally aligned multi-channel signal set to obtain the transformed signal representation; S4. Perform adaptive filtering based on valve fault characteristics on the transformed signal representation to obtain the enhanced signal representation; S5. Perform time-frequency domain feature extraction on the enhanced signal representation to obtain a fault feature set related to the valve fault type; S6. Perform feature evolution trend analysis on the fault feature set to match the valve structural damage. When the feature evolution trend shows a deterministic law related to the valve structural damage, generate and output the valve fault identification information.

2. The data detection method for valve faults as described in claim 1, characterized in that, The time synchronization of multi-channel sensor data includes: It receives a unified timing signal and generates a corresponding timestamp for each data point in the multi-channel sensing data from different sensors based on the signal. Based on the timestamps, the data streams from all sensors are aligned to the same time base to obtain time-synchronized multi-channel data streams.

3. The data detection method for valve faults as described in claim 2, characterized in that, The time-delay alignment process based on valve operating conditions includes: Determine the start time of valve operating condition changes based on valve opening and closing action signals; Using the start time as the alignment point, time shift compensation is performed on the time-synchronized multi-channel data streams to align the signal segments corresponding to the same valve action in each data stream in time. The time-shift compensated data streams are segmented and spliced ​​to obtain a spatiotemporally aligned multi-channel signal set.

4. The data detection method for valve faults as described in claim 1, characterized in that, The spatial coherence analysis of the spatiotemporally aligned multi-channel signal set yields the coherence relationships between each signal related to valve faults, including: Based on the start time of the valve operating condition change, a signal window for coherence analysis is determined on a spatiotemporally aligned multi-channel signal set; Within the signal window, the coherence degree calculation based on the cross power spectral density is performed on any two signals to obtain the corresponding coherence coefficients; Based on the magnitude of the coherence coefficients, the pairing relationships that characterize the strong linkage between signals are extracted and used as the coherence relationships related to valve failure.

5. The data detection method for valve faults as described in claim 1, characterized in that, The construction of the coherence relation matrix includes: Using each signal as a row and column of a matrix, the coherence relationship related to valve failure is mapped to the element at the corresponding position in the matrix; Based on the correlation priority between the coherence relationship and different valve failure modes, the order of elements in the matrix is ​​adjusted by weighting. After mapping and sorting all coherent relationships, a coherent relationship matrix is ​​formed, which is based on the contribution of failure modes.

6. The data detection method for valve faults as described in claim 1, characterized in that, The spatial transformation process, which uses the coherence relation matrix as a transformation constraint to perform coherence-preserving transformation on a spatiotemporally aligned multi-channel signal set, includes: The coherence relation matrix is ​​decomposed into eigenvalues ​​to obtain the eigenvalues ​​and eigenvectors of the coherence relation matrix, and the eigenvectors are used as basis vectors to guide the transformation. Based on the basis vectors, a linear projection is performed on the spatiotemporally aligned set of multiple signals, mapping the multiple signals at each moment to a coordinate point in the space spanned by the basis vectors. The coordinates of all points at all times are collected to form the transformed signal representation.

7. The data detection method for valve faults as described in claim 6, characterized in that, The method also includes optimizing and recombining the transformed signal representation, including: Based on the basis vectors and the coherence relation matrix, the matrix elements are weighted and synthesized by using the components of the basis vectors as weights to obtain the comprehensive coherence strength value of each basis vector. All basis vectors and their corresponding coordinate components are reordered according to the order of comprehensive coherence intensity from strongest to weakest. The reordered sequence of coordinate components is used to construct an optimized transformed signal representation.

8. The data detection method for valve faults as described in claim 7, characterized in that, The process of adaptively filtering the transformed signal representation based on valve fault characteristics to obtain the enhanced signal representation includes: The energy contribution distribution is obtained by analyzing the energy contribution of each sequential segment, sorted by comprehensive coherence strength, to the overall signal in the optimized transformed signal representation. Based on the trend of energy contribution distribution changing with ranking position, the transition segment from concentrated distribution to dispersed distribution of energy contribution is identified; Extract the signal components corresponding to all sequential segments before the transition segment to form the enhanced signal representation.

9. The data detection method for valve faults as described in claim 1, characterized in that, The feature evolution trend analysis of the fault feature set matching valve structural damage includes: Arrange each feature in the fault feature set in chronological order to form a feature value time series; For each feature value time series, divide it into consecutive time windows, extract the representative value of the feature value distribution within each time window, and calculate the difference between the representative values ​​of the distribution distribution of adjacent time windows to obtain the statistical difference value; Analyze the trend of statistical difference values ​​over time. If the trend shows a monotonically increasing trend, it is determined that the feature has an evolutionary trend related to structural damage.

10. A data detection system for valve faults, used to implement the data detection method for valve faults as described in any one of claims 1-9, characterized in that, The system includes: The data processing module is used to acquire multi-channel sensor data under the valve's operating status, perform time synchronization and delay alignment processing based on the valve's operating conditions on the multi-channel sensor data, and obtain a spatiotemporally aligned multi-channel signal set. The spatial coherence analysis module performs spatial coherence analysis on a spatiotemporally aligned multi-channel signal set to obtain the coherence relationship between each signal and the valve fault and construct a coherence relationship matrix. The signal transformation module is used to perform coherence-preserving spatial transformation processing on a spatiotemporally aligned set of signals, using the coherence relation matrix as the transformation constraint, to obtain the transformed signal representation. An adaptive filtering module is used to adaptively filter the transformed signal representation based on valve fault characteristics to obtain an enhanced signal representation. The feature extraction module is used to extract time-frequency domain features from the enhanced signal representation to obtain a fault feature set related to the valve fault type. The evolution analysis module is used to perform feature evolution trend analysis on the fault feature set and match it with valve structural damage. When the feature evolution trend shows a deterministic law related to valve structural damage, it generates and outputs valve fault identification information.

Images