A method for diagnosing the operating state of a pressure regulator

Abstract

The application discloses a method for diagnosing the working state of a pressure regulator, comprising the following steps: S1, judging whether the distance between the data in the running data set and the health centroid is less than the health maximum distance; if not, entering S2; S2, eliminating the data from the running data set; when the distance between the residual data is less than the first threshold value, calculating the first offset between the current centroid and the health centroid, and entering S3; S3, collecting a new running data set and a new centroid after a preset time interval; judging whether the difference between the new centroid and the current centroid is greater than the second threshold value; if yes, returning to S2; if the difference between the new centroid and the current centroid is greater than the second threshold value and the data in the new running data set meets the clustering condition, judging whether the first and second offset vectors are collinear; if not, returning to S3; if yes, entering S4; S4, the fault type corresponding to the minimum distance between the new centroid and the fault centroid is the diagnosis result. The application can accurately identify the working state of the pressure regulator.

Description

Technical Field

[0001] This invention relates to the field of voltage regulators, and in particular to a method for diagnosing the operating status of voltage regulators. Background Technology

[0002] As a key hub in urban gas transmission and distribution networks, pressure regulating stations, equipped with pressure regulators, play a crucial role in regulating and stabilizing gas pressure. Their operational status directly determines the stability and safety of gas supply. Currently, the operation and maintenance model for pressure regulators in medium and large-sized gas stations in China mainly relies on periodic manual inspections and post-fault repairs, lacking 24 / 7 online monitoring and early warning capabilities.

[0003] Traditional operation and maintenance methods rely on maintenance personnel to observe the equipment's appearance and simply record operating parameters on-site. This approach can only detect obvious, overt faults and cannot accurately capture subtle changes in operating parameters such as pressure, flow, temperature, vibration, and noise of the pressure regulator. Problems such as component wear, operating condition deviations, and latent anomalies that occur during long-term operation of the pressure regulator are difficult to identify in a timely manner and can easily develop into sudden failures. These failures can not only cause pressure regulation failure and abnormal pipeline pressure, affecting normal gas supply downstream, but also pose serious gas safety hazards in severe cases.

[0004] Most existing pressure regulator monitoring technologies only provide threshold alarms for single pressure and flow parameters, resulting in a single monitoring dimension and a lack of comprehensive analysis combining multi-dimensional operating signals such as vibration and acoustics, leading to low fault identification accuracy. Summary of the Invention

[0005] To address the aforementioned problems, the present invention aims to provide a method for diagnosing the operating status of a voltage regulator, which can accurately identify the operating status of the voltage regulator.

[0006] Based on this, the present invention provides a method for diagnosing the operating status of a voltage regulator, the method comprising:

[0007] S1: Obtain the operating data set of the voltage regulator to be tested, and determine whether the distance between each operating data in the operating data set and the preset healthy centroid does not exceed the preset healthy maximum distance. If yes, the voltage regulator to be tested is determined to be in a healthy state. If no, proceed to step S2.

[0008] S2: Remove data from the running dataset whose distance from the preset healthy centroid does not exceed the maximum healthy distance. When the distance between the remaining data is less than the first threshold, calculate the current centroid of the current dataset composed of the remaining data and the first offset between the current centroid and the healthy centroid, and then proceed to step S3.

[0009] S3: After a preset time interval, collect the operating data of the voltage regulator under test again to obtain a new operating dataset, and calculate the new centroid of the new operating dataset. Determine whether the difference between the new centroid and the current centroid is not greater than the second threshold. If yes, return to step S2. If the difference between the new centroid and the current centroid is greater than the second threshold and the data in the new operating dataset meets the clustering conditions, determine whether the first offset vector and the second offset vector are collinear, where the second offset vector is the offset vector between the new centroid and the current centroid. If they are not collinear, return to step S3. If they are collinear, proceed to step S4.

[0010] S4: Calculate the distance between the new centroid and each preset fault centroid, and use the fault type corresponding to the minimum distance as the diagnosis result.

[0011] The step of removing data from the running dataset whose distance from the preset healthy centroid does not exceed the maximum healthy distance includes: traversing each running data in the running dataset, calculating the Euclidean distance between the running data and the healthy centroid, and if the Euclidean distance is less than or equal to the maximum healthy distance, then marking the running data as healthy data and removing it from the running dataset.

[0012] The distance between the remaining data is the Euclidean distance between any two data points in the remaining data.

[0013] The process of obtaining the current centroid includes: calculating the arithmetic mean of each dimension of all data in the current dataset composed of the remaining data, and using the arithmetic mean of each dimension as the coordinate component of the current centroid.

[0014] The running dataset consists of several feature vector groups, which include: pressure features, vibration features, noise features, signal kurtosis features, spectral centroid features, spectral variance features, signal margin factor features, and signal supplementary statistical features.

[0015] The data in the new running dataset that meet the clustering conditions include: the distance between any two sets of feature vectors in the new running dataset is less than a preset third threshold.

[0016] After obtaining the first offset, a multi-level anomaly alarm determination step is set up, specifically including:

[0017] The system continuously collects running data and updates the centroid at preset time intervals to form multiple rounds of continuous cyclic detection; it records the offset distance of the centroid in each round of cyclic detection relative to the centroid in the previous round; it summarizes the offset distances and calculates the average offset.

[0018] When the average offset is less than the first set value, a level 1 abnormal alarm is triggered; when the average offset is between the first set value and the second set value, a level 2 abnormal alarm is triggered; when the average offset is greater than the third set value, a level 3 abnormal alarm is triggered.

[0019] The running dataset consists of multiple sets of feature vectors, and the method further includes a warning determination step based on distance ratio.

[0020] Select a single feature vector from the running data, determine the distance between the feature vector and the healthy centroid, and for each type of preset fault centroid, determine the distance between the fault centroid and the healthy centroid, and calculate the ratio of the two corresponding distances in turn.

[0021] When the ratio is less than a first preset value, a level 1 anomaly warning is triggered; when the ratio is within the range of the first preset value to the second preset value, a level 2 anomaly warning is triggered; when the ratio is within the range of the second preset value to the third preset value, a level 3 anomaly warning is triggered.

[0022] The method further includes:

[0023] Before acquiring the running dataset, a basic operating condition verification is performed on the pressure regulator to be tested: the effective value of the inlet pressure of the pressure regulator to be tested is checked in sequence to see if it is greater than the preset minimum allowable value, the degree of inlet pressure fluctuation is less than the preset stable threshold, and the outlet temperature is higher than the preset temperature value. If any one of the checks fails, the data is collected again.

[0024] If a preset number of checks fail, the voltage regulator under test is determined to have a basic operating condition fault.

[0025] The determination of whether the first offset vector and the second offset vector are collinear includes: if there exists a real number such that the second offset vector coincides with the first offset vector after scaling, then the first offset vector and the second offset vector are collinear.

[0026] This invention uses the health centroid and distance threshold as quantitative judgment criteria to accurately distinguish between normal equipment operating conditions and equipment malfunctions, and to promptly detect early faults that are difficult to identify using traditional methods. By filtering and clustering abnormal data, interference caused by instantaneous operating condition fluctuations can be effectively eliminated, improving monitoring reliability. By dynamically tracking centroid offset and vector changes through multi-round data acquisition, the development trend of faults can be clearly judged. Combined with fault centroid distance comparison, the fault type can be automatically determined, greatly improving diagnostic efficiency. This method can achieve all-weather automated online monitoring, promoting the transformation of pressure regulator operation and maintenance from post-event repair to pre-event early warning. While reducing labor costs, it comprehensively enhances the operational safety and stability of the gas transmission and distribution system. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a flowchart of a method for diagnosing the working status of a voltage regulator, provided by an embodiment of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Figure 1 This is a flowchart of a method for diagnosing the operating status of a voltage regulator according to an embodiment of the present invention, the method comprising:

[0031] S1: Obtain the operating data set of the voltage regulator to be tested, and determine whether the distance between each operating data in the operating data set and the preset healthy centroid does not exceed the preset healthy maximum distance. If yes, the voltage regulator to be tested is determined to be in a healthy state. If no, proceed to step S2.

[0032] This invention focuses on data acquisition for a voltage regulator. The signals to be acquired include temperature, pressure, vibration, and sound, specifically temperatures t1 and t2 before and after the regulator, pressures p1 and p2, and the regulator's own sound signal k1 and vibration signal o1. To avoid interference, sensors are added at a distance from the regulator to collect sound signals k2 and k3 from peripheral devices such as manifolds and filters. The default data acquisition frequency is once per hour, but can be adjusted as needed, recording all parameters and their corresponding durations T during normal operation. The signal processing primarily focuses on noise suppression. Algorithms such as adaptive noise cancellation (LMS / NLMS algorithm), spectral subtraction, and Wiener filtering can be used to filter out interference noise from peripheral devices, resulting in a valid sound signal s1. These noise reduction algorithms are existing technologies and will not be detailed further.

[0033] This invention first classifies and filters the collected voltage regulator operating data, distinguishing between normal operating parameters and parameters that are abnormal due to operating conditions, and then uniformly enters the classified data into a comparison database.

[0034] The method further includes:

[0035] Before acquiring the running dataset, a basic operating condition verification is performed on the pressure regulator to be tested: the effective value of the inlet pressure of the pressure regulator to be tested is checked in sequence to see if it is greater than the preset minimum allowable value, the degree of inlet pressure fluctuation is less than the preset stable threshold, and the outlet temperature is higher than the preset temperature value. If any one of the checks fails, the data is collected again.

[0036] If a preset number of checks fail, the voltage regulator under test is determined to have a basic operating condition fault.

[0037] Specifically, three judgment conditions are set around inlet pressure, pressure stability, and outlet temperature. The corresponding indicators are calculated based on integral calculation and compared with the threshold.

[0038] Inlet pressure value verification: The mean value u1 and the effective value (RMS) of the inlet pressure p1 within the collection period T are calculated by formula. It is stipulated that the effective value RMS must be greater than or equal to the minimum allowable inlet pressure of the pressure regulator to ensure that the inlet pressure is within the compliance range.

[0039] Inlet pressure stability verification: The standard deviation of the inlet pressure, denoted by σ, is calculated based on the formula. This value is used to characterize the degree of pressure fluctuation. σ is required to not exceed the threshold set according to the actual operating conditions to ensure stable inlet pressure operation.

[0040] Outlet temperature verification: The set outlet temperature t2 must be higher than the natural gas dew point t0 by a preset number of units, such as 5 units or more, to meet the operating temperature requirements.

[0041] If the data does not meet any of the above judgment criteria, data collection and testing need to be carried out again. Compare the differences between the two sets of data. If the absolute value of the difference between the two sets of data is greater than the preset threshold, it indicates that the deviation is large. If the two sets of data have a large deviation, it is necessary to return to the previous data collection step and repeat the operation. If the results of multiple retests still cannot be unified, the voltage regulator body should be troubleshooted and repaired. Among them, if the absolute value of the difference between the two sets of data is less than the preset threshold, it indicates that the data is unified. The threshold is an empirical value set manually.

[0042] By collecting a large amount of operating data from voltage regulators, a database of the entire life cycle of voltage regulators, from normal to sub-healthy to faulty, is established.

[0043] The running dataset consists of several feature vector groups, which include: pressure features, vibration features, noise features, signal kurtosis features, spectral centroid features, spectral variance features, signal margin factor features, and signal supplementary statistical features.

[0044] Specifically, establish and run the dataset For continuous signals, the definition is as follows:

[0045] ;

[0046] ;

[0047] ;

[0048] ;

[0049] Where YS is a preset value, which can be 3;

[0050] ;

[0051] ;

[0052] ;

[0053] ;

[0054] ;

[0055] in, These are the average values ​​of pressure, vibration, and noise, respectively. The maximum value of the signal. Indicates standard deviation, This represents the frequency value at the k-th point in the vibration signal. FC represents the power spectrum value corresponding to the k-th point in the vibration signal, and FC represents the corresponding spectral center or frequency mean in the vibration signal.

[0056] First, calculate the centroid of the range in the normal state:

[0057] Health dataset collected during normal operation of the voltage regulator common One data point, , i = 1, 2 ... n, where n can be any positive integer greater than 1, corresponding to the previously extracted... .

[0058] Computing health datasets The center of mass, of which, .

[0059] ;

[0060] Calculate the maximum distance between each data point and the centroid, which is the distance farthest from the centroid among all healthy samples, and is also the maximum permissible deviation from the health status:

[0061] ;

[0062] The distance here refers to the Euclidean distance. When the distance between the operating data and the healthy centroid is less than or equal to d, the voltage regulator is considered to be in a healthy state. When it is greater than d, the voltage regulator is in a faulty state.

[0063] For each type of regulator fault state, a separate fault dataset is established, and the same calculation logic as the health data is used to achieve fault classification modeling.

[0064] The dataset for the i-th type of fault is: Among them, the fault dataset Using the mean algorithm of the healthy dataset, the centroid of the current fault dataset is calculated: that is, the fault centroid. Calculate the centroid specific to this type of fault. This represents the typical characteristic state of this type of fault, and a unique label φᵢ is assigned to this fault to distinguish different fault types.

[0065] Similarly, calculate the maximum distance between the data in each type of fault dataset and its respective centroid.

[0066] ;

[0067] in, For fault dataset The center of mass, It is the maximum distance from all samples of this type of fault to their own centroid, which serves as the boundary for determining this type of fault.

[0068] The existing voltage regulator has the following operating data set for a certain period of time: It meets the following conditions:

[0069] ;

[0070] This indicates that all data are within the healthy threshold range, confirming that the device's initial operating state is healthy and without abnormal deviations. In other words, the voltage regulator was in a healthy operating state during this period.

[0071] After the voltage regulator has been running continuously for a period of time, such as D days (D can be set by the user), collect the running data set Z. .

[0072] S2: Remove data from the running dataset whose distance from the preset healthy centroid does not exceed the maximum healthy distance. When the distance between the remaining data is less than the first threshold, calculate the current centroid of the current dataset composed of the remaining data and the first offset between the current centroid and the healthy centroid, and then proceed to step S3.

[0073] The step of removing data from the running dataset whose distance from the preset healthy centroid does not exceed the maximum healthy distance includes: traversing each running data in the running dataset, calculating the Euclidean distance between the running data and the healthy centroid, and if the Euclidean distance is less than or equal to the maximum healthy distance, then marking the running data as healthy data and removing it from the running dataset.

[0074] The distance between the remaining data is the Euclidean distance between any two data points in the remaining data.

[0075] The process of obtaining the current centroid includes: calculating the arithmetic mean of each dimension of all data in the current dataset composed of the remaining data, and using the arithmetic mean of each dimension as the coordinate component of the current centroid.

[0076] If it exists:

[0077] This indicates that the voltage regulator has generated abnormal data that is outside the healthy range;

[0078] Remove those that satisfy the following formula :

[0079] ;

[0080] In other words, to remove all health data and retain only abnormal data, the following conditions must be met simultaneously:

[0081] This means that none of the data is health data;

[0082] That is, the abnormal data features are concentrated and there is no clutter or interference;

[0083] This means that after a fault occurs, the data gradually moves away from its original normal centroid;

[0084] in, and To remove data from the running dataset whose distance from the preset healthy centroid does not exceed the maximum healthy distance (i.e., the remaining data), a first threshold is set. The user can set the parameters as needed. If the above conditions are met, it indicates that the device is shifting towards a sub-optimal state, and the centroid of Z should be recalculated. The current centroid represents the overall operating characteristics of the voltage regulator.

[0085] Recalculate The distance between C and [the other].

[0086] ;

[0087] The threshold value is related to the type of voltage regulator and setting preferences. If the above formula is true, it means that the voltage regulator is in a stable state and the health status of the equipment does not change much.

[0088] To accurately determine which type of sub-health fault the equipment offset is approaching, vector distance calculation is performed:

[0089] That is, defining the offset direction vector, which represents the direction and magnitude of the device's offset from the healthy state.

[0090] That is, the offset vector of the centroid of each type of fault relative to the healthy centroid.

[0091] For each type of fault centroid, the vertical distance from the current offset trajectory of the voltage regulator is: the smaller the distance, the closer the voltage regulator is to that type of fault.

[0092] After obtaining the first offset, a multi-level anomaly alarm determination step is set up, specifically including:

[0093] The system continuously collects running data and updates the centroid at preset time intervals to form multiple rounds of continuous cyclic detection; it records the offset distance of the centroid in each round of cyclic detection relative to the centroid in the previous round; it summarizes the offset distances and calculates the average offset.

[0094] When the average offset is less than the first set value, a level 1 abnormal alarm is triggered; when the average offset is between the first set value and the second set value, a level 2 abnormal alarm is triggered; when the average offset is greater than the third set value, a level 3 abnormal alarm is triggered.

[0095] The presence of valid abnormal data during the operation of the voltage regulator equipment meets the judgment criteria.

[0096] ;

[0097] In the multi-cycle monitoring of the voltage regulator equipment, the offset between the old and new anomalous centroids can be calculated in each iteration. k takes values ​​from 1 to N, where N is a positive integer. The average value is calculated as follows:

[0098] ;

[0099] parameter It can eliminate random errors caused by single data fluctuations and environmental interference, and truly reflect the overall deviation trend and abnormal stability of the voltage regulator equipment in long-term operation. It is the core judgment indicator for graded early warning.

[0100] Abnormal data consistently and stably appears during the monitoring period, always meeting the requirements. The average centroid offset obtained from multiple iterations can be expressed as: .

[0101] like When the value is less than the first set value, such as 1, it indicates that although the voltage regulator continuously deviates from a healthy state, the centroid offset is extremely small over multiple cycles, and the overall system tends to be stable, without a trend of continuous offset or deterioration. The fault is in the nascent steady-state stage, and the abnormal development has stalled, triggering a level one abnormality alarm. If the instantaneous data satisfies ||Z|| m -C‖>The first warning value, such as 1, can be directly included in the category of Level II abnormal warning.

[0102] like When the value is greater than the first set value but less than the second set value (e.g., 2), the regulator's centroid exhibits a continuous and regular shift, triggering a level-two abnormal alarm. If the instantaneous abnormal data satisfies ||Z|| m -C‖> If the second alert value is 2, it will be directly upgraded to a Level 3 anomaly warning.

[0103] like When the value exceeds the third set value, such as 3, the overall displacement of the regulator's centroid is extremely large, and it remains in a state of continuous displacement and deterioration for a long time. The equipment's operating status deviates significantly from the healthy standard and is highly close to a sub-healthy fault. The risk of fault occurrence is extremely high, the equipment's operating stability drops significantly, and a level three abnormality warning is triggered.

[0104] S3: After a preset time interval, collect the operating data of the voltage regulator under test again to obtain a new operating dataset, and calculate the new centroid of the new operating dataset. Determine whether the difference between the new centroid and the current centroid is not greater than the second threshold. If yes, return to step S2. If the difference between the new centroid and the current centroid is greater than the second threshold and the data in the new operating dataset meets the clustering conditions, determine whether the first offset vector and the second offset vector are collinear, where the second offset vector is the offset vector between the new centroid and the current centroid. If they are not collinear, return to step S3. If they are collinear, proceed to step S4.

[0105] The data in the new running dataset that meet the clustering conditions include: the distance between any two sets of feature vectors in the new running dataset is less than a preset third threshold.

[0106] The determination of whether the first offset vector and the second offset vector are collinear includes: if there exists a real number such that the second offset vector coincides with the first offset vector after scaling, then the first offset vector and the second offset vector are collinear.

[0107] Specifically, after a period of time D (D can be n days), the running data... Calculate the new centroid as At this point, two scenarios need to be considered:

[0108] Scenario 1: The situation tends to stabilize;

[0109] ;

[0110] make ;

[0111] n is the number of iterations: n = 1, 2, 3, ... N.

[0112] This indicates that the sub-healthy offset state of the voltage regulator has stabilized and has not deteriorated significantly. There is no need to immediately determine the fault; continued monitoring is sufficient to proceed to the next data acquisition cycle.

[0113] Scenario 2: The condition continues to deteriorate or the direction of deviation changes;

[0114] ;

[0115] ;

[0116] ;

[0117] in, Adjust as needed and recalculate the centroid. .

[0118] New offset vector ;

[0119] New fault relative vector ;

[0120] New matching distance ;

[0121] Judgment Vector and Are they collinear? If so, and If they are collinear, it indicates that they are collinear; otherwise, they are not collinear. If they are not collinear, it means that the fault trend is not stable, and this step should be repeated for continuous tracking.

[0122] S4: Calculate the distance between the new centroid and each preset fault centroid, and use the fault type corresponding to the minimum distance as the diagnosis result.

[0123] Once the voltage regulator's offset trend stabilizes, the final fault matching determination is carried out:

[0124] Calculate the minimum distance between the current equipment centroid and the centroids of all sub-healthy faults:

[0125] ;

[0126] And it meets the threshold condition. ;

[0127] It can be adjusted according to the actual situation.

[0128] Fault centroid corresponding to minimum distance The corresponding fault type is the fault type currently matched by the current voltage regulator.

[0129] The running dataset consists of multiple sets of feature vectors, and the method further includes a warning determination step based on distance ratio.

[0130] Select a single feature vector from the running data, determine the distance between the feature vector and the healthy centroid, and for each type of preset fault centroid, determine the distance between the fault centroid and the healthy centroid, and calculate the ratio of the two corresponding distances in turn.

[0131] When the ratio is less than a first preset value, a level 1 anomaly warning is triggered; when the ratio is within the range of the first preset value to the second preset value, a level 2 anomaly warning is triggered; when the ratio is within the range of the second preset value to the third preset value, a level 3 anomaly warning is triggered.

[0132] The ratio can be expressed as: ,in, This represents the average value of the data processed within a preset time period. This represents the centroid of the i-th type of sub-health fault. It indicates a healthy mindset.

[0133] The first preset value, the second preset value, and the third preset value can all be set independently. Preferably, the first preset value is... The second preset value is The third preset value is .

[0134] Normal state: All operating parameters of the equipment are stable, with no state deviation or abnormal fluctuations, fully meeting the healthy operation standards, and the equipment is operating safely and stably.

[0135] Level 1 Anomaly Warning: The equipment shows minor abnormal signs and is in the early and controllable stage of the fault. The abnormality has stopped developing and shows no trend of aggravation. No emergency repair is required. Routine monitoring is sufficient.

[0136] Level 2 Anomaly Warning: The abnormal characteristics of the equipment are clear and there is a trend of continuous development and accelerated deterioration. The risk is gradually increasing and it needs to be listed as a key monitoring target, and the frequency of data collection and status observation should be increased.

[0137] Level 3 Anomaly Warning: The equipment is highly abnormal and is approaching a sub-healthy failure state. There are clear potential faults, and a comprehensive equipment inspection, hazard identification, and maintenance must be carried out immediately.

[0138] This invention uses the health centroid and distance threshold as quantitative judgment criteria to accurately distinguish between normal equipment operating conditions and equipment malfunctions, and to promptly detect early faults that are difficult to identify using traditional methods. By filtering and clustering abnormal data, interference caused by instantaneous operating condition fluctuations can be effectively eliminated, improving monitoring reliability. By dynamically tracking centroid offset and vector changes through multi-round data acquisition, the development trend of faults can be clearly judged. Combined with fault centroid distance comparison, the fault type can be automatically determined, greatly improving diagnostic efficiency. This method can achieve all-weather automated online monitoring, promoting the transformation of pressure regulator operation and maintenance from post-event repair to pre-event early warning. While reducing labor costs, it comprehensively enhances the operational safety and stability of the gas transmission and distribution system.

[0139] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for diagnosing the operating status of a voltage regulator, characterized in that, include: S1: Obtain the operating data set of the voltage regulator to be tested, and determine whether the distance between each operating data in the operating data set and the preset healthy centroid does not exceed the preset healthy maximum distance. If yes, the voltage regulator to be tested is determined to be in a healthy state. If no, proceed to step S2. S2: Remove data from the running dataset whose distance from the preset healthy centroid does not exceed the maximum healthy distance. When the distance between the remaining data is less than the first threshold, calculate the current centroid of the current dataset composed of the remaining data and the first offset between the current centroid and the healthy centroid, and then proceed to step S3. S3: After a preset time interval, collect the operating data of the voltage regulator under test again to obtain a new operating dataset, and calculate the new centroid of the new operating dataset. Determine whether the difference between the new centroid and the current centroid is not greater than the second threshold. If yes, return to step S2. If the difference between the new centroid and the current centroid is greater than the second threshold and the data in the new operating dataset meets the clustering conditions, determine whether the first offset vector and the second offset vector are collinear, where the second offset vector is the offset vector between the new centroid and the current centroid. If they are not collinear, return to step S3. If they are collinear, proceed to step S4. S4: Calculate the distance between the new centroid and each preset fault centroid, and use the fault type corresponding to the minimum distance as the diagnosis result.

2. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, Removing data from the running dataset whose distance from the preset healthy centroid does not exceed the maximum healthy distance includes: traversing each running data in the running dataset, calculating the Euclidean distance between the running data and the healthy centroid, and if the Euclidean distance is less than or equal to the maximum healthy distance, then marking the running data as healthy data and removing it from the running dataset.

3. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, The distance between the remaining data is the Euclidean distance between any two data points in the remaining data.

4. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, The process of obtaining the current centroid includes: calculating the arithmetic mean of each dimension of all data in the current dataset composed of the remaining data, and using the arithmetic mean of each dimension as the coordinate component of the current centroid.

5. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, The running dataset consists of several feature vector groups, which include: pressure features, vibration features, noise features, signal kurtosis features, spectral centroid features, spectral variance features, signal margin factor features, and signal supplementary statistical features.

6. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, The data in the new running dataset meet the clustering conditions as follows: the distance between any two sets of feature vectors in the new running dataset is less than a preset third threshold.

7. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, After obtaining the first offset, a multi-level anomaly alarm determination process is also set up, specifically including: The system continuously collects running data and updates the centroid at preset time intervals to form multiple rounds of continuous cyclic detection; it records the offset distance of the centroid in each round of cyclic detection relative to the centroid in the previous round; it summarizes the offset distances and calculates the average offset. When the average offset is less than the first set value, a level 1 abnormal alarm is triggered; when the average offset is between the first set value and the second set value, a level 2 abnormal alarm is triggered; when the average offset is greater than the third set value, a level 3 abnormal alarm is triggered.

8. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, The running dataset consists of multiple sets of feature vectors, and the method further includes a warning determination step based on distance ratio: Select a single feature vector from the running data, determine the distance between the feature vector and the healthy centroid, and for each type of preset fault centroid, determine the distance between the fault centroid and the healthy centroid, and calculate the ratio of the two corresponding distances in turn. When the ratio is less than a first preset value, a level one anomaly warning is triggered; When the ratio is within the range of the first preset value to the second preset value, a level two abnormality warning is triggered; When the ratio is within the range of the second preset value to the third preset value, a level three abnormality warning is triggered.

9. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, The method further includes: Before acquiring the running dataset, a basic operating condition verification is performed on the pressure regulator to be tested: the effective value of the inlet pressure of the pressure regulator to be tested is checked in sequence to see if it is greater than the preset minimum allowable value, the degree of inlet pressure fluctuation is less than the preset stable threshold, and the outlet temperature is higher than the preset temperature value. If any one of the checks fails, the data is collected again. If a preset number of checks fail, the voltage regulator under test is determined to have a basic operating condition fault.

10. The method for diagnosing the operating status of a voltage regulator as described in claim 1, characterized in that, Determining whether the first offset vector and the second offset vector are collinear includes: if there exists a real number such that the second offset vector, after scaling, coincides with the first offset vector, then the first offset vector and the second offset vector are collinear.

Images