Oil and gas pipeline valve inner leakage online diagnosis method based on voiceprint analysis

Abstract

The invention relates to the technical field of oil and gas pipeline valve inner leakage online diagnosis, and discloses an oil and gas pipeline valve inner leakage online diagnosis method based on voiceprint analysis. According to the method, valve operation voiceprint data are obtained through a distributed interface to construct a feature database, a diagnosis framework is designed according to the database, and a multi-source voiceprint collection scheme is determined. And the voiceprint acquisition terminal acquires a multi-dimensional voiceprint signal according to the scheme, inputs the multi-dimensional voiceprint signal into a model after feature extraction to complete inner leakage level judgment and abnormal mode recognition, and outputs a diagnosis result. A diagnosis result is transmitted to the management optimization module, and diagnosis strategy adjustment is achieved through timestamp calibration, mapping table construction and the like. And a three-level evaluation sub-frame is also constructed to assist diagnosis. According to the method, accurate diagnosis and strategy optimization of oil and gas pipeline valve inner leakage are realized, and the diagnosis management level is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of online diagnosis of internal leakage of oil and gas pipeline valves, and in particular to an online diagnosis method of internal leakage of oil and gas pipeline valves based on voiceprint analysis. Background Art

[0002] In the oil and gas transportation industry, the proper operation of pipeline valves plays a crucial role in the safety and efficiency of the entire transportation system. Valve internal leakage not only wastes oil and gas resources but can also cause serious safety incidents such as fires and explosions, threatening personnel safety and the ecological environment. Therefore, accurate and timely diagnosis of valve internal leakage is crucial.

[0003] Traditional methods for diagnosing internal leaks in oil and gas pipeline valves have many limitations. Some rely on regular on-site inspection personnel to listen, see, and touch to determine the valve status. This method is greatly influenced by the experience and subjective judgment of the inspectors, making it difficult to detect small internal leaks in the early stages. Manual inspections are also inefficient and costly, and cannot achieve real-time monitoring. Other methods use physical parameter detection methods such as pressure and flow monitoring to determine whether there is an internal leak by analyzing changes in pressure and flow within the pipeline. However, during the oil and gas transportation process, the physical parameters within the pipeline are easily affected by various factors such as changes in the transportation volume and fluctuations in ambient temperature, resulting in inaccurate detection results and misjudgments and missed judgments.

[0004] With technological advancements, some sensor-based monitoring technologies have been gradually adopted. However, most of these technologies monitor a single parameter and cannot fully reflect the status of valve internal leakage. Furthermore, these technologies have limited capabilities in data processing and analysis, failing to fully tap into the value of data, making it difficult to accurately diagnose and effectively manage valve internal leakage. Furthermore, existing diagnostic methods lack a dynamic adjustment mechanism for diagnostic strategies, preventing timely optimization of diagnostic processes and resource allocation based on actual diagnostic conditions. This makes them unable to meet the growing demand for diagnosing internal leakage in oil and gas pipeline valves. Therefore, a more efficient, accurate, and intelligently managed online method for diagnosing internal leakage in oil and gas pipeline valves is urgently needed. Summary of the Invention

[0005] The purpose of the present invention is to provide an online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides an online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis, the method comprising:

[0007] Acquire voiceprint data of oil and gas pipeline valves in operation, and construct a voiceprint feature database based on the voiceprint data;

[0008] Designing an endoleak diagnosis framework based on the voiceprint feature database, and determining a multi-source voiceprint collection scheme by combining the diagnosis framework with the voiceprint feature database;

[0009] The voiceprint acquisition terminal acquires the multi-dimensional voiceprint signal of the valve according to the acquisition scheme, performs internal leakage level determination and abnormal pattern recognition based on the multi-dimensional voiceprint signal, and outputs a diagnosis result;

[0010] The diagnosis results are transmitted to the management optimization module for diagnosis strategy adjustment.

[0011] Preferably, the voiceprint data of the oil and gas pipeline valves under the operating state is obtained, and a voiceprint feature database is constructed based on the voiceprint data, specifically:

[0012] Connect the valve body sensor and the pipeline attached microphone array through a distributed interface to obtain the time domain waveform information of the valve operation soundprint and the frequency domain distribution information of the steady-state working condition soundprint;

[0013] The time-domain waveform information of the valve operation soundprint is framed and windowed using time-frequency conversion technology, and the frequency-domain distribution information of the steady-state working condition soundprint is subjected to background noise filtering using an energy threshold method to obtain a denoised soundprint feature unit.

[0014] Constructing a voiceprint feature mapping network based on feature association rules, inputting the denoised voiceprint feature units into the mapping network for temporal association analysis, and extracting core feature dimensions that are strongly correlated with the endoleak state;

[0015] The core feature dimensions are standardized and labeled with corresponding internal leakage status labels. Feature aggregation is performed according to valve type and operating time period to construct a voiceprint feature database including time domain features, frequency domain features and device features.

[0016] Preferably, the endoleak diagnosis framework is designed based on the voiceprint feature database, and a multi-source voiceprint collection scheme is determined by combining the diagnosis framework and the voiceprint feature database, specifically:

[0017] Decomposing the voiceprint feature database into three dimensions, extracting three core evaluation dimensions: voiceprint energy distribution, frequency characteristic offset, and time-varying regularity fluctuation, and assigning initial weight parameters to each dimension based on expert experience;

[0018] The core evaluation dimensions are topologically mapped to the spatial structure of oil and gas pipelines, and a three-level evaluation subframework consisting of valve body layer, pipe section layer, and system layer is constructed. The synergistic influence weights between the subframeworks are calculated through feature cross-validation.

[0019] Check the matching degree between the core feature dimensions in the voiceprint feature database and the three-level evaluation sub-framework, and filter out the feature coverage missing items corresponding to each sub-framework;

[0020] Determine the types of voiceprint signals that need to be collected based on the missing features, and calculate the collection frequency and layout location of each type of signal based on the sensor layout density and data transmission bandwidth limitation in the pipeline environment;

[0021] Based on the decision tree algorithm, a comprehensive judgment is made on the collection frequency, layout location and feature coverage missing items to generate a multi-source voiceprint collection plan that includes signal type, collection frequency and layout location.

[0022] Preferably, the decision tree algorithm is used to comprehensively judge the acquisition frequency, layout location and feature coverage missing items to generate a multi-source voiceprint acquisition solution including signal type, acquisition frequency and layout location, specifically:

[0023] Obtaining hardware performance indicators of each voiceprint collection terminal in the pipeline environment, including signal sensitivity, sampling rate, and battery life;

[0024] Constructing a collection constraint condition set, wherein the constraint condition set includes a single-terminal maximum sampling rate constraint, a single-terminal maximum storage capacity constraint, and a system total transmission bandwidth constraint;

[0025] Taking the acquisition frequency, layout location, and feature coverage missing items as decision factors, the correlation function between each decision factor and the acquisition constraint condition set is established respectively;

[0026] The importance scores of each decision factor are integrated through the decision tree splitting rule to obtain the feasibility scores of each signal type at different acquisition frequencies;

[0027] Filter out the collection frequency combinations with feasibility scores higher than the set threshold, dynamically configure the sensor resources of each signal type based on the layout location, and generate a multi-source voiceprint collection plan.

[0028] Preferably, the voiceprint acquisition terminal acquires the valve multi-dimensional voiceprint signal according to the acquisition scheme, performs internal leakage level determination and abnormal pattern recognition based on the multi-dimensional voiceprint signal, and outputs a diagnosis result, specifically:

[0029] The voiceprint acquisition terminal synchronously acquires the vibration voiceprint data of the valve body, the flow voiceprint data of the pipeline fluid, and the interference voiceprint data of the environmental background according to the acquisition frequency and signal type requirements of the acquisition scheme;

[0030] The short-time zero-crossing rate algorithm is used to extract time domain features of the vibration voiceprint data, the Mel-frequency cepstral coefficient method is used to extract frequency domain features of the flow voiceprint data, and the envelope analysis method is used to extract noise features of the interference voiceprint data to obtain a fused feature vector;

[0031] Inputting the fused feature vector into a pre-trained voiceprint classification model, wherein the classification model constructs an endoleak level determination layer based on a feedforward neural network and an abnormal pattern recognition layer based on a recurrent neural network;

[0032] The inner leakage level determination layer outputs the inner leakage level classification result, the abnormality pattern recognition layer outputs the abnormality feature matching result, and a logical AND operation is performed on the classification result and the matching result to output the diagnosis result.

[0033] Preferably, the transmitting of the diagnosis result to the management optimization module for adjusting the diagnosis strategy is specifically as follows:

[0034] Encapsulating the diagnostic results based on a data transmission protocol to generate structured feedback information including endoleak level, abnormality pattern, and occurrence location;

[0035] The structured feedback information is transmitted to the storage database of the management optimization module through the data transmission middleware, and the historical strategy data and the current feedback information in the storage database are timestamp-calibrated;

[0036] Constructing a policy adjustment mapping table, which includes the correspondence between internal leakage level and inspection priority, the correspondence between abnormal mode and repair process, and the correspondence between occurrence location and resource allocation;

[0037] Parameters of the historical strategy data are adjusted according to the mapping table to generate an updated diagnosis strategy.

[0038] Preferably, the structured feedback information is transmitted to the storage database of the management optimization module through the data transmission middleware, and the timestamp calibration operation is performed on the historical policy data and the current feedback information in the storage database, specifically:

[0039] Obtain the timestamp distribution characteristics of historical policy data in the management optimization module storage database, and calculate the average interval of timestamps as the benchmark calibration period;

[0040] Rounding down the timestamp of the structured feedback information according to the reference calibration period to obtain a calibration timestamp;

[0041] Performing exponential weighted averaging on the data in the historical policy data whose timestamps are within a period before and after the calibration timestamp to calculate the time calibration value of the historical policy data;

[0042] Nonlinear interpolation is performed on the numerical data of the structured feedback information and the time calibration value to obtain a time-calibrated feedback data sequence.

[0043] Preferably, the construction strategy adjustment mapping table includes the correspondence between the internal leakage level and the inspection priority, the correspondence between the abnormal mode and the repair process, and the correspondence between the occurrence location and the resource allocation, specifically:

[0044] Perform correlation analysis on the time-calibrated feedback data sequence to calculate the correlation coefficient between internal leakage level and inspection priority, the correlation coefficient between abnormal pattern and repair process, and the correlation coefficient between occurrence location and resource allocation;

[0045] Filter out the variable pairs whose absolute value of correlation coefficient is greater than the set threshold as the core correlation items of the mapping table;

[0046] Perform interval division statistics on the core related items, determine the value range of another variable corresponding to each variable interval, and construct a policy adjustment mapping table containing discrete interval mapping relationships.

[0047] Preferably, the parameter adjustment of the historical strategy data according to the mapping table to generate an updated diagnostic strategy is specifically as follows:

[0048] For the inspection priority parameters in the historical policy data, adjust the priority value according to the mapping relationship between the internal leakage level and the inspection priority;

[0049] For the repair process parameters in the historical strategy data, replace the matching standard process based on the mapping relationship between the abnormal pattern and the repair process;

[0050] For resource allocation parameters in historical strategy data, the resource allocation ratio is modified based on the mapping relationship between the occurrence location and resource allocation;

[0051] Perform consistency verification on the adjusted inspection priorities, repair processes, and resource allocation parameters to generate an updated diagnostic strategy.

[0052] Preferably, the construction includes a three-level evaluation subframework comprising valve body layer, pipe section layer, and system layer, specifically:

[0053] The valve body layer evaluation subframework includes three evaluation indicators: the soundprint energy value, frequency offset, and duration of a single valve;

[0054] The pipe segment layer evaluation subframe includes three evaluation indicators: the acoustic print attenuation of the associated pipe segment, the impact length value, and the number of affected nodes;

[0055] The system-level evaluation subframework includes three evaluation indicators: global voiceprint coverage, redundant detection rate, and recovery time value;

[0056] The comprehensive scores of the evaluation sub-frameworks at each level are calculated based on the factor analysis method, and a three-level evaluation sub-framework including the valve body layer, pipe section layer, and system layer is constructed through linear weighted summation.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] This method, based on voiceprint analysis, diagnoses internal leaks in oil and gas pipeline valves online. By acquiring voiceprint data from valve operation to build a voiceprint feature database, the method utilizes a distributed interface to connect to multiple devices to acquire time and frequency domain information. Through operations such as denoising and feature extraction, the method comprehensively and accurately integrates the valve operation voiceprint data, providing a rich data foundation for subsequent diagnosis. When determining a multi-source voiceprint acquisition scheme, hardware performance, acquisition constraints, and feature requirements are comprehensively considered. The acquisition parameters are dynamically configured using a decision tree algorithm, ensuring that the acquisition scheme is tailored to the actual pipeline environment and that the collected voiceprint data is representative and effective.

[0059] The voiceprint acquisition terminal simultaneously acquires multidimensional voiceprint signals, applies multiple algorithms to extract features, and inputs them into pre-trained models for internal leakage level determination and abnormal pattern recognition. This analyzes voiceprint signals from multiple perspectives, improving diagnostic accuracy and comprehensiveness. After the diagnostic results are transmitted to the management optimization module, dynamic adjustments to the diagnostic strategy are achieved through operations such as timestamp calibration and the construction of a policy adjustment mapping table. Inspection priorities are adjusted based on internal leakage level, repair processes are matched based on abnormal pattern matching, and resources are allocated based on the location of the leakage. This ensures that the diagnostic strategy is more aligned with actual needs, improves the efficiency and management of internal leakage diagnosis of oil and gas pipeline valves, and reduces resource waste and safety hazards.

[0060] In addition, the constructed three-level assessment sub-framework conducts multi-level assessment of valve internal leakage from the valve body layer, pipe section layer, and system layer. Different assessment indicators are set at each layer to comprehensively reflect the impact of valve internal leakage at different levels, providing a more comprehensive reference basis for diagnosis and decision-making, and helping to achieve systematic and scientific management of oil and gas pipeline valve internal leakage diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a working principle diagram of the online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to the present invention;

[0062] Figure 2 A working diagram of the process of building a voiceprint feature database;

[0063] Figure 3 Generate a flowchart for a multi-source voiceprint collection solution based on a decision tree algorithm;

[0064] Figure 4 Workflow diagram for diagnosis result transmission and diagnosis strategy adjustment. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] See also Figure 1-Figure 4 The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis of the present invention is specifically implemented as follows:

[0067] Execute the operation of obtaining the voiceprint data of the oil and gas pipeline valve under the operating state, and construct a voiceprint feature database based on the voiceprint data. Through the distributed interface, the connection with the valve body sensor and the pipeline attached microphone array is realized to obtain the time domain waveform information of the valve operation soundprint and the frequency domain distribution information of the steady-state working condition soundprint. Use the time-frequency conversion technology to perform frame and window processing on the time domain waveform information, and combine the energy threshold method to filter the background noise of the frequency domain distribution information to obtain the denoised voiceprint feature unit. Construct a voiceprint feature mapping network based on the feature association rules, input the denoised voiceprint feature unit into the network for time series association analysis, extract the core feature dimensions that are strongly correlated with the internal leakage state, and then standardize the core feature dimensions and mark the corresponding internal leakage state labels. Perform feature aggregation according to the valve type and operating time period to complete the construction of the voiceprint feature database.

[0068] Based on the constructed voiceprint feature database, an internal leak diagnosis framework was designed. This framework was then combined with the voiceprint feature database to determine a multi-source voiceprint acquisition plan. The voiceprint feature database was dimensionally decomposed to extract three core evaluation dimensions: voiceprint energy distribution, frequency characteristic offset, and time-varying regular fluctuation. Initial weight parameters were assigned to each dimension based on expert experience. The core evaluation dimensions were topologically aligned with the spatial structural hierarchy of the oil and gas pipeline, and a three-level evaluation subframework consisting of the valve body layer, the pipe section layer, and the system layer was constructed. The synergistic influence weights between the subframeworks were calculated through feature cross-validation. The core feature dimensions in the voiceprint feature database were checked for match with the three-level evaluation subframeworks. Missing feature coverage items corresponding to each subframework were identified. Based on these missing items, the types of voiceprint signals that needed to be collected were determined. The acquisition frequency and location of each type of signal were calculated, taking into account the sensor density and data transmission bandwidth limitations in the pipeline environment. Finally, a decision tree algorithm was used to comprehensively determine the acquisition frequency, location, and missing feature coverage items, generating a multi-source voiceprint acquisition plan.

[0069] The voiceprint acquisition terminal then acquires the valve's multidimensional voiceprint signal according to the acquisition plan. Based on this multidimensional voiceprint signal, it determines the internal leakage level and identifies abnormal patterns, outputting the diagnostic results. The voiceprint acquisition terminal simultaneously acquires the valve's vibration voiceprint data, the pipeline fluid's flow voiceprint data, and the environmental background's interference voiceprint data, as required by the acquisition plan. It then extracts time-domain features from the vibration voiceprint data using a short-time zero-crossing rate algorithm, extracts frequency-domain features from the flow voiceprint data using the Mel-frequency cepstral coefficient method, and extracts noise features from the interference voiceprint data using envelope analysis, generating a fused feature vector. This fused feature vector is then fed into a pre-trained voiceprint classification model, which constructs an internal leakage level determination layer based on a feedforward neural network and an abnormal pattern recognition layer based on a recurrent neural network. The internal leakage level determination layer outputs the internal leakage level classification result, while the abnormal pattern recognition layer outputs the abnormal feature matching result. The classification and matching results are logically ANDed together to output the diagnostic result.

[0070] Finally, the diagnostic results are transmitted to the management optimization module for diagnostic strategy adjustment. The diagnostic results are encapsulated based on the data transmission protocol to generate structured feedback information containing the internal leakage level, abnormality pattern, and location of occurrence. This structured feedback information is transmitted to the storage database of the management optimization module through the data transmission middleware. The historical policy data in the storage database is timestamped with the current feedback information. A policy adjustment mapping table is constructed, which contains the correspondence between the internal leakage level and the inspection priority, the abnormality pattern and the repair process, and the location of occurrence and resource allocation. The parameters of the historical policy data are adjusted according to the mapping table to generate an updated diagnostic strategy.

[0071] Example 1:

[0072] When determining a multi-source voiceprint collection solution, the first step is to obtain the hardware performance indicators of each voiceprint collection terminal in the pipeline environment. These indicators cover signal sensitivity, sampling rate, and battery life. Signal sensitivity determines the collection terminal's ability to capture weak voiceprint signals, and different models of collection terminals have different performance indicators. The sampling rate affects the amount of data collected per unit time. Although a higher sampling rate can obtain richer data, it also requires higher computing and storage capabilities of the terminal. Battery life is related to the continuous working time of the collection terminal, affecting the continuity of data collection.

[0073] Then, a set of acquisition constraint conditions is constructed, where the maximum sampling rate constraint for a single terminal is set based on the computing power and data processing capabilities of the acquisition terminal hardware. If the sampling rate is too high, the acquisition terminal may not be able to process and store the collected data in a timely manner, resulting in data loss or lag in the operation of the acquisition terminal. The maximum storage capacity constraint for a single terminal takes into account the capacity limitations of the storage device of the acquisition terminal itself, preventing the collected data from exceeding its storage capacity and avoiding interruption of data acquisition due to insufficient storage space. The total transmission bandwidth constraint of the system is to ensure the stable operation of the entire data transmission system. In a pipeline environment, multiple acquisition terminals transmit data simultaneously. If the total transmission bandwidth is insufficient, congestion, delays, and even data loss will occur during the data transmission process.

[0074] Taking collection frequency, deployment location, and feature coverage omission as decision factors, a correlation function is established between each decision factor and the set of collection constraints. The collection frequency is closely related to the maximum sampling rate constraint for a single terminal. A higher collection frequency means an increase in the amount of data collected per unit time. Exceeding the maximum sampling rate for a single terminal violates this constraint. The collection frequency also affects the amount of data transmitted and is related to the total system bandwidth constraint. The deployment location determines the quality and integrity of the voiceprint signal captured by the collection terminal. Different deployment locations will subject the collection terminal to different interference, affecting signal sensitivity. The deployment location also affects the power supply and data transmission of the collection terminal, and is related to the battery life and the total system bandwidth. The feature coverage omission reflects the key voiceprint features not captured under the current collection scheme and determines the types of voiceprint signals that require additional collection. This decision factor interacts with the collection frequency and deployment location, as different types of voiceprint signals may require different collection frequencies and appropriate deployment locations for effective acquisition.

[0075] The importance scores of each decision factor are integrated using a decision tree splitting rule. This decision tree splitting rule evaluates and classifies the importance of the three decision factors: acquisition frequency, deployment location, and feature coverage gaps, according to specific logic and criteria. During this evaluation process, the constraints of the acquisition constraint set, as well as the requirements of different oil and gas pipeline environments and valve characteristics for each decision factor, are comprehensively considered. For example, in complex pipeline environments, feature coverage gaps may be more important, as obtaining comprehensive and accurate voiceprint features is crucial for endoleak diagnosis. In situations with limited hardware performance, the acquisition frequency and the maximum sampling rate constraint for a single terminal may be more closely correlated, prioritizing the impact of the acquisition frequency on the hardware. This method generates a feasibility score for each signal type at different acquisition frequencies. This feasibility score is derived based on the comprehensive relationship between each decision factor and the acquisition constraint set, reflecting the feasibility and rationality of acquiring a particular signal type at a specific acquisition frequency.

[0076] Finally, the acquisition frequency combinations with feasibility scores higher than the set threshold are screened out, and the sensor resources for each signal type are dynamically configured in combination with the layout location. The set threshold is a standard value determined based on actual conditions and experience. The acquisition frequency combination higher than the threshold is considered to be a more feasible solution under the current conditions. Dynamic configuration combined with the layout location is to reasonably allocate sensor resources according to the signal acquisition requirements and hardware conditions of different locations. For example, in key locations such as near valves, it may be necessary to increase the density of sensor layout to obtain more accurate voiceprint signals; for some locations that are far away and have greater difficulty in signal transmission, it is necessary to select appropriate sensor types and layout methods to ensure that the required voiceprint signals can be effectively collected. Finally, a multi-source voiceprint collection solution that includes signal type, acquisition frequency, and layout location is generated, so that the collection solution can collect the voiceprint signals related to oil and gas pipeline valves as comprehensively and accurately as possible under the premise of meeting the limitations of hardware performance and transmission conditions.

[0077] Example 2:

[0078] The voiceprint acquisition terminal adheres to the acquisition frequency and signal type requirements specified in the acquisition plan when acquiring and diagnosing multidimensional valve voiceprint signals. This acquisition plan, developed based on an analysis of the operational characteristics of oil and gas pipeline valves and diagnostic requirements, aims to comprehensively capture voiceprint information related to valve internal leakage. The acquisition terminal simultaneously collects vibration voiceprint data from the valve itself, flow data from the pipeline fluid, and background interference data.

[0079] After acquiring the vibration soundprint data, it is processed using a short-time zero-crossing rate algorithm. This algorithm extracts time-domain features by counting the number of times the vibration soundprint signal crosses the zero axis per unit time. The vibration characteristics of a valve change during normal operation and during abnormal conditions such as internal leakage. These changes are reflected in the time-domain characteristics of the vibration soundprint signal. By calculating the short-time zero-crossing rate of the vibration soundprint data, these subtle changes can be captured, thereby obtaining effective characteristic information related to valve vibration.

[0080] For pipeline fluid flow soundprint data, the Mel-frequency cepstral coefficient method is used to extract frequency domain features. Based on the human ear's perception of sound frequency, the Mel-frequency cepstral coefficient method converts the actual sound frequency into a Mel-frequency scale. In oil and gas pipelines, normal fluid flow and abnormal flow caused by factors such as valve internal leakage produce soundprint signals with different frequency distributions. Processing flow soundprint data using the Mel-frequency cepstral coefficient method can analyze these signals from a frequency domain perspective, extracting features related to the pipeline fluid flow state, and revealing the characteristics of the fluid soundprint in a way that better aligns with human auditory perception.

[0081] Envelope analysis is used to extract noise features from environmental background interference soundprint data. This method performs a series of processing operations, including demodulation, on the interference soundprint signal. In real-world oil and gas pipeline environments, noise generated by various interfering factors can interfere with the valid soundprint signal required for valve internal leakage diagnosis. Envelope analysis can extract noise features from complex interference soundprint signals and distinguish them from valid soundprint signals.

[0082] The time-domain features of the vibration soundprint data, the frequency-domain features of the flow soundprint data, and the noise features of the interference soundprint data extracted by the above different methods are fused to form a fused feature vector. The fused feature vector integrates the soundprint feature information from different aspects and comprehensively reflects the soundprint generated during valve operation.

[0083] The pre-trained voiceprint classification model plays a crucial role in the diagnostic process. This model comprises an endoleak classification layer based on a feedforward neural network and an anomaly pattern recognition layer based on a recurrent neural network. After the fused feature vector is input into the voiceprint classification model, the feedforward neural network in the endoleak classification layer performs multi-layer processing and calculations on the input feature vector. Each layer of the feedforward neural network performs specific transformations and processing on the input information. Through layer-by-layer transmission and calculation, it gradually extracts features related to the endoleak level and outputs the endoleak classification results, categorizing the valve endoleak into different levels.

[0084] The recurrent neural network in the abnormal pattern recognition layer leverages its ability to process sequence data to analyze the fused feature vector. During valve operation, different abnormal patterns generate voiceprint signals with specific sequence characteristics. The recurrent neural network captures these sequence characteristics and matches them with pre-trained abnormal pattern features. It then outputs the abnormal feature matching results and identifies the specific abnormal pattern of the valve.

[0085] Finally, a logical AND operation is performed on the internal leakage classification results output by the internal leakage level determination layer and the abnormal feature matching results output by the abnormal pattern recognition layer. This logical AND operation comprehensively evaluates the two results. Only when both the internal leakage level classification result and the abnormal feature matching result meet certain conditions is the final diagnosis result output. This diagnosis result comprehensively reflects the valve internal leakage level and the abnormal pattern that has occurred, providing detailed and accurate information for subsequent treatment of oil and gas pipeline valve internal leakage.

[0086] Example 3:

[0087] When transmitting diagnostic results to the management optimization module for diagnostic strategy adjustment, the results must first be encapsulated using the data transmission protocol. The results contain information such as the endoleak level, anomaly pattern, and location. The data transmission protocol specifies the encapsulation format for this information, creating structured feedback for easy transmission and processing within the system.

[0088] The structured feedback information is transmitted to the storage database of the management optimization module through the data transmission middleware. In this process, it is necessary to perform timestamp calibration on the historical policy data and the current feedback information in the storage database. First, obtain the timestamp distribution characteristics of the historical policy data in the storage database of the management optimization module. The timestamp is the information that marks the time when the data is generated or the event occurs. By analyzing the timestamps of the historical policy data, the average interval of the timestamps is calculated, and this average interval is set as the benchmark calibration period, which is represented by the symbol express.

[0089] Process the timestamp of the structured feedback information. The timestamp of the structured feedback information is calibrated according to the reference period. Perform the rounding operation down to get the calibration timestamp. Let the original timestamp of the structured feedback information be , the calibration timestamp is , the calculation method of rounding down can be expressed as:

[0090] in, Represents the floor function, that is, taking the largest integer not greater than the value in the brackets.

[0091] Then, the exponentially weighted average of the historical strategy data with timestamps within a period before and after the calibration timestamp is performed to calculate the time calibration value of the historical strategy data. The value at , the weight coefficient of the exponential weighted average is ( ), calculate the time calibration value The process is as follows: Determine the time range , the historical strategy data within this range participates in the calculation. For each timestamp Data at , assign weight , weights are updated with the calibration timestamp The time calibration value decays exponentially with the increase of the time interval. The calculation formula is:

[0092] This formula indicates that the historical strategy data within one period before and after the calibration timestamp is assigned different weights according to the time interval with the calibration timestamp, and then the sum is divided by the sum of all weights to obtain a time calibration value that comprehensively considers time correlation.

[0093] Finally, the numerical data of the structured feedback information is nonlinearly interpolated with the time-calibrated values. The numerical data in the structured feedback information, such as the quantitative values ​​corresponding to the endoleak level and the coded values ​​of the abnormal patterns, are processed with the calculated time-calibrated values. Based on the distribution characteristics of the data, the nonlinear interpolation method establishes a nonlinear correspondence between the two, and through this relationship, a time-calibrated feedback data sequence is obtained. The specific nonlinear interpolation method can select an appropriate algorithm based on the actual data characteristics, such as polynomial interpolation and spline interpolation. This series of operations ensures that the structured feedback information is consistent with the historical strategy data in the time dimension, allowing data generated at different times to be compared and analyzed on the same time scale, facilitating subsequent adjustments to the diagnostic strategy based on this data.

[0094] Example 4:

[0095] When constructing the strategy adjustment mapping table, the time-calibrated feedback data sequence is first processed. Consider a set of time-calibrated feedback data containing multiple data records, each of which contains information such as the internal leakage level, anomaly pattern, and location. For example, one record indicates that the internal leakage level of an oil and gas pipeline valve is "medium," the anomaly pattern is "seal wear," and the location is "V-101 valve on pipeline A." Another record indicates that the internal leakage level is "minor," the anomaly pattern is "loose connection due to pipeline vibration," and the location is "V-202 valve on pipeline B."

[0096] Calculate the correlation coefficients between the internal leakage level and inspection priority, the correlation coefficient between the abnormal pattern and the repair process, and the correlation coefficient between the occurrence location and resource allocation. Taking the calculation of the correlation coefficient between the internal leakage level and the inspection priority as an example, traverse all time-calibrated feedback data records. The internal leakage level is divided into different categories, such as "minor," "medium," and "severe." For the inspection priority, corresponding levels are also set, such as "low," "medium," and "high." In the data records, examine the records with a "minor" internal leakage level and count the distribution of the number of "low," "medium," and "high" inspection priorities. Similarly, perform similar statistics for the records with "medium" and "severe" internal leakage levels. By analyzing the correspondence between these data, the correlation coefficient is calculated. The correlation coefficients for the abnormal pattern and repair process, and the occurrence location and resource allocation, are calculated similarly, based on the analysis and calculation of the correspondence between different variables in the feedback data records.

[0097] Next, a threshold is set based on actual conditions and experience. Assuming the threshold is set at 0.5, variable pairs with absolute correlation coefficients greater than the threshold are selected and designated as core association items in the mapping table. For example, if the correlation coefficient between an internal leakage level of "Severe" and an inspection priority of "High" is 0.6, which is greater than the set threshold of 0.5, the variable pair "Internal leakage level - Severe" and "Inspection priority - High" will be identified as a core association item. Another example is that the correlation coefficient between the abnormal pattern "seal wear" and the repair process "seal replacement" is 0.7, which will also be included in the core association items.

[0098] Core correlation items are divided into intervals and statistics are performed. Taking the core correlation item of internal leakage level and inspection priority as an example, the internal leakage level is further subdivided into intervals, such as the "minor" internal leakage level is further subdivided into "minor-early" and "minor-late", and the "medium" internal leakage level is subdivided into "medium-mild" and "medium-severe", etc. For each subdivided internal leakage level interval, the corresponding reasonable inspection priority value range is statistically analyzed. For example, after statistical analysis of a large amount of data, it was found that when the internal leakage level is in the "minor-early" range, the inspection priority is generally in the "low-low" range; while when the internal leakage level is in the "medium-severe" range, the inspection priority is mostly in the "medium-high" range. For the core correlation item of abnormal pattern and repair process, the abnormal pattern is more finely classified according to the specific fault manifestation, such as "seal wear" is divided into "surface wear" and "deep wear", etc., and then the standard repair process corresponding to each abnormal pattern subdivision interval is determined. For the core correlation items between the occurrence location and resource allocation, the occurrence location is divided into intervals based on factors such as the geographical location and importance of the pipeline. For example, the pipeline is divided into "critical transmission sections" and "secondary transmission sections", and then the resource allocation ratio corresponding to each interval is determined. For example, when allocating resources in the "critical transmission section", the allocation ratio of human and material resources is relatively high.

[0099] Through the above series of operations, a strategy adjustment mapping table containing discrete interval mapping relationships is constructed. This mapping table records in detail the relationship between each interval of internal leakage level and the corresponding interval of inspection priority, such as the "mild-initial" internal leakage level corresponding to the "low-low" inspection priority; the corresponding relationship between each interval of abnormal mode and the repair process, such as the "seal-surface wear" abnormal mode corresponding to the "clean the sealing surface and apply sealant" repair process; and the corresponding relationship between each interval of occurrence location and resource allocation, such as the "critical conveying section" occurrence location corresponding to a higher proportion of human and material resource allocation. This strategy adjustment mapping table provides a specific reference basis for subsequent adjustment of the diagnostic strategy based on the diagnostic results, allowing the diagnostic strategy to be adjusted according to the actual internal leakage situation.

[0100] Example 5:

[0101] A three-level assessment sub-framework consisting of valve body layer, pipe section layer, and system layer is constructed to evaluate the internal leakage of oil and gas pipeline valves from different levels and dimensions.

[0102] The valve body-level assessment subframework focuses on a single valve and employs three evaluation metrics: soundprint energy, frequency offset, and duration. The soundprint energy value represents the energy contained in the soundprint signal during valve operation. Under normal valve operation, the soundprint energy value remains relatively stable. However, when an abnormal condition such as internal leakage occurs, the vibration and flow changes caused by the leaking fluid cause the soundprint energy value to change, potentially manifesting as an increase in energy or abnormal fluctuations. The frequency offset measures the degree to which the valve soundprint frequency differs from the normal operating frequency. A properly functioning valve has a soundprint frequency that follows a specific pattern and range. However, internal structural damage, seal failure, or interference from foreign objects can cause the soundprint frequency to shift. Monitoring the frequency offset can capture these subtle changes. Duration measures the duration of the abnormal soundprint. While brief soundprint anomalies may be caused by accidental interference, persistent abnormalities are more likely to indicate a substantial problem, such as internal leakage. This metric helps determine the severity and persistence of the abnormality.

[0103] The pipeline-level assessment subframework focuses on associated pipeline segments and includes three evaluation metrics: acoustic print attenuation, impact length, and number of affected nodes. Acoustic print attenuation measures the energy attenuation of acoustic print signals as they propagate through a pipeline segment. In oil and gas pipeline systems, acoustic print signals propagate along a pipeline segment, attenuating due to factors such as pipe material, diameter variations, and curvature. When a valve leak occurs, the attenuation characteristics of the acoustic print signal generated by the leak differ from those under normal conditions. By analyzing the acoustic print attenuation, the leak's location and the extent of its impact on surrounding pipeline segments can be inferred. The impact length represents the length of the pipeline segment affected by the valve leak. Leaks can cause changes in fluid pressure and flow velocity within the pipeline segment, and these effects propagate along the entire pipeline segment. The impact length provides a direct reflection of the spatial extent of the leak's impact. The number of affected nodes reflects the number of pipeline nodes affected by the leak. Pipeline systems have numerous connection nodes and monitoring nodes, and an internal leak can cause parameter anomalies at multiple nodes. By counting the number of affected nodes, the extent of the leak's impact on the overall structure and operational status of the pipeline segment can be assessed.

[0104] The system-level evaluation subframework takes a global perspective and includes three evaluation metrics: global voiceprint coverage, redundant detection rate, and recovery time. The global voiceprint coverage rate reflects the coverage of the entire oil and gas pipeline system by voiceprint collection equipment, specifically the percentage of the system where voiceprint signals can be effectively collected. A higher global voiceprint coverage rate means more comprehensive voiceprint information can be obtained from all parts of the system, enabling the timely detection of potential internal leaks. Conversely, blind spots in voiceprint collection may prevent internal leaks from being detected in a timely manner. The redundant detection rate indicates the percentage of the system with redundant detection capabilities. Implementing redundant detection by setting up multiple voiceprint collection points or using multiple detection methods can improve the accuracy and reliability of internal leak detection. The redundant detection rate reflects the system's fault tolerance and backup capabilities. The recovery time is used to assess the time it takes for the system to recover from a faulty state to normal operation after detecting an internal leak. It comprehensively considers factors such as maintenance processes, resource allocation, and repair technologies, and is a key indicator of the system's resilience to internal leaks.

[0105] After obtaining the evaluation indicator data of each level of evaluation sub-framework, the comprehensive score of each level of evaluation sub-framework is calculated based on the factor analysis method. The factor analysis method extracts a few comprehensive factors by performing dimensionality reduction processing on multiple indicator data. These comprehensive factors can reflect most of the information of the original indicators. Each comprehensive factor is assigned a corresponding weight according to the degree of its explanation of the original indicator. The comprehensive score of each level of evaluation sub-framework is obtained by weighted summing up the scores of each indicator data on the comprehensive factor. Finally, the comprehensive scores of the three evaluation sub-frameworks of valve body layer, pipe section layer and system layer are integrated by linear weighted summation. Each sub-framework is assigned an appropriate weight according to its importance in the overall evaluation. The final evaluation result of the three-level evaluation sub-framework including valve body layer, pipe section layer and system layer is calculated, realizing a multi-level and comprehensive evaluation of the internal leakage of oil and gas pipeline valves from a single valve, associated pipe section to the entire system.

[0106] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0107] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis, characterized in that: The following steps are involved: Acquire voiceprint data of oil and gas pipeline valves in operation, and construct a voiceprint feature database based on the voiceprint data; Designing an endoleak diagnosis framework based on the voiceprint feature database, and determining a multi-source voiceprint collection scheme by combining the diagnosis framework with the voiceprint feature database; The voiceprint acquisition terminal acquires the multi-dimensional voiceprint signal of the valve according to the acquisition scheme, performs internal leakage level determination and abnormal pattern recognition based on the multi-dimensional voiceprint signal, and outputs a diagnosis result; The diagnosis results are transmitted to the management optimization module for diagnosis strategy adjustment.

2. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 1 is characterized in that: The voiceprint data of the oil and gas pipeline valve under the operating state is obtained, and a voiceprint feature database is constructed based on the voiceprint data, specifically: Connect the valve body sensor and the pipeline attached microphone array through a distributed interface to obtain the time domain waveform information of the valve operation soundprint and the frequency domain distribution information of the steady-state working condition soundprint; The time-domain waveform information of the valve operation soundprint is framed and windowed using time-frequency conversion technology, and the frequency-domain distribution information of the steady-state working condition soundprint is subjected to background noise filtering using an energy threshold method to obtain a denoised soundprint feature unit. Constructing a voiceprint feature mapping network based on feature association rules, inputting the denoised voiceprint feature units into the mapping network for temporal association analysis, and extracting core feature dimensions that are strongly correlated with the endoleak state; The core feature dimensions are standardized and labeled with corresponding internal leakage status labels. Feature aggregation is performed according to valve type and operating time period to construct a voiceprint feature database including time domain features, frequency domain features and device features.

3. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 1 is characterized in that: The design of the internal leakage diagnosis framework based on the voiceprint feature database and the determination of the multi-source voiceprint collection scheme in combination with the diagnosis framework and the voiceprint feature database are specifically as follows: Decomposing the voiceprint feature database into three dimensions, extracting three core evaluation dimensions: voiceprint energy distribution, frequency characteristic offset, and time-varying regularity fluctuation, and assigning initial weight parameters to each dimension based on expert experience; The core evaluation dimensions are topologically mapped to the spatial structure of oil and gas pipelines, and a three-level evaluation subframework consisting of valve body layer, pipe section layer, and system layer is constructed. The synergistic influence weights between the subframeworks are calculated through feature cross-validation. Check the matching degree between the core feature dimensions in the voiceprint feature database and the three-level evaluation sub-framework, and filter out the feature coverage missing items corresponding to each sub-framework; Determine the types of voiceprint signals that need to be collected based on the missing features, and calculate the collection frequency and layout location of each type of signal based on the sensor layout density and data transmission bandwidth limitation in the pipeline environment; Based on the decision tree algorithm, a comprehensive judgment is made on the collection frequency, layout location and feature coverage missing items to generate a multi-source voiceprint collection plan that includes signal type, collection frequency and layout location.

4. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 3 is characterized in that: The decision tree algorithm is used to comprehensively judge the acquisition frequency, layout location and feature coverage missing items, and generate a multi-source voiceprint acquisition solution including signal type, acquisition frequency and layout location, specifically: Obtaining hardware performance indicators of each voiceprint collection terminal in the pipeline environment, including signal sensitivity, sampling rate, and battery life; Constructing a collection constraint condition set, wherein the constraint condition set includes a single-terminal maximum sampling rate constraint, a single-terminal maximum storage capacity constraint, and a system total transmission bandwidth constraint; Taking the acquisition frequency, layout location, and feature coverage missing items as decision factors, the correlation function between each decision factor and the acquisition constraint condition set is established respectively; The importance scores of each decision factor are integrated through the decision tree splitting rule to obtain the feasibility scores of each signal type at different acquisition frequencies; Filter out the collection frequency combinations with feasibility scores higher than the set threshold, dynamically configure the sensor resources of each signal type based on the layout location, and generate a multi-source voiceprint collection plan.

5. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 1 is characterized in that: The voiceprint acquisition terminal acquires the valve multi-dimensional voiceprint signal according to the acquisition scheme, performs internal leakage level determination and abnormal pattern recognition based on the multi-dimensional voiceprint signal, and outputs the diagnosis result, specifically: The voiceprint acquisition terminal synchronously acquires the vibration voiceprint data of the valve body, the flow voiceprint data of the pipeline fluid, and the interference voiceprint data of the environmental background according to the acquisition frequency and signal type requirements of the acquisition scheme; The short-time zero-crossing rate algorithm is used to extract time domain features of the vibration voiceprint data, the Mel-frequency cepstral coefficient method is used to extract frequency domain features of the flow voiceprint data, and the envelope analysis method is used to extract noise features of the interference voiceprint data to obtain a fused feature vector; Inputting the fused feature vector into a pre-trained voiceprint classification model, wherein the classification model constructs an endoleak level determination layer based on a feedforward neural network and an abnormal pattern recognition layer based on a recurrent neural network; The inner leakage level determination layer outputs the inner leakage level classification result, the abnormality pattern recognition layer outputs the abnormality feature matching result, and a logical AND operation is performed on the classification result and the matching result to output the diagnosis result.

6. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 1 is characterized in that: The transmitting of the diagnosis result to the management optimization module for adjusting the diagnosis strategy is specifically as follows: Encapsulating the diagnostic results based on a data transmission protocol to generate structured feedback information including endoleak level, abnormality pattern, and occurrence location; The structured feedback information is transmitted to the storage database of the management optimization module through the data transmission middleware, and the historical strategy data and the current feedback information in the storage database are timestamp-calibrated; Constructing a policy adjustment mapping table, which includes the correspondence between internal leakage level and inspection priority, the correspondence between abnormal mode and repair process, and the correspondence between occurrence location and resource allocation; Parameters of the historical strategy data are adjusted according to the mapping table to generate an updated diagnosis strategy.

7. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 6 is characterized in that: The structured feedback information is transmitted to the storage database of the management optimization module through the data transmission middleware, and the timestamp calibration operation is performed on the historical policy data and the current feedback information in the storage database, specifically: Obtain the timestamp distribution characteristics of historical policy data in the management optimization module storage database, and calculate the average interval of timestamps as the benchmark calibration period; Rounding down the timestamp of the structured feedback information according to the reference calibration period to obtain a calibration timestamp; Performing exponential weighted averaging on the data in the historical policy data whose timestamps are within a period before and after the calibration timestamp to calculate the time calibration value of the historical policy data; Nonlinear interpolation is performed on the numerical data of the structured feedback information and the time calibration value to obtain a time-calibrated feedback data sequence.

8. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 7 is characterized in that: The construction strategy adjustment mapping table includes the correspondence between internal leakage level and inspection priority, the correspondence between abnormal mode and repair process, and the correspondence between occurrence location and resource allocation, specifically: Perform correlation analysis on the time-calibrated feedback data sequence to calculate the correlation coefficient between internal leakage level and inspection priority, the correlation coefficient between abnormal pattern and repair process, and the correlation coefficient between occurrence location and resource allocation; Filter out the variable pairs whose absolute value of correlation coefficient is greater than the set threshold as the core correlation items of the mapping table; Perform interval division statistics on the core related items, determine the value range of another variable corresponding to each variable interval, and construct a policy adjustment mapping table containing discrete interval mapping relationships.

9. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 8 is characterized in that: The parameter adjustment of the historical strategy data according to the mapping table to generate an updated diagnostic strategy is specifically as follows: For the inspection priority parameters in the historical policy data, adjust the priority value according to the mapping relationship between the internal leakage level and the inspection priority; For the repair process parameters in the historical strategy data, replace the matching standard process based on the mapping relationship between the abnormal pattern and the repair process; For resource allocation parameters in historical strategy data, the resource allocation ratio is modified based on the mapping relationship between the occurrence location and resource allocation; Perform consistency verification on the adjusted inspection priorities, repair processes, and resource allocation parameters to generate an updated diagnostic strategy.

10. The online diagnosis method for internal leakage of oil and gas pipeline valves based on voiceprint analysis according to claim 3 is characterized in that: The construction includes a three-level evaluation sub-framework at the valve body level, the pipe section level, and the system level, specifically: The valve body layer evaluation subframework includes three evaluation indicators: the soundprint energy value, frequency offset, and duration of a single valve; The pipe segment layer evaluation subframe includes three evaluation indicators: the acoustic print attenuation of the associated pipe segment, the impact length value, and the number of affected nodes; The system-level evaluation subframework includes three evaluation indicators: global voiceprint coverage, redundant detection rate, and recovery time value; The comprehensive scores of the evaluation sub-frameworks at each level are calculated based on the factor analysis method, and a three-level evaluation sub-framework including the valve body layer, pipe section layer, and system layer is constructed through linear weighted summation.

Images