Instrument fault detection method based on multi-method fusion

Abstract

The invention discloses an instrument fault detection method based on multi-method fusion, and relates to the technical field of equipment detection, and the method comprises the five steps of detection tool preparation, data acquisition and analysis, fusion detection, fault maintenance and verification, and detection reporting. The dynamic threshold value is obtained through the dynamic threshold value adjusting unit, then the abnormal score is obtained through the abnormal condition scoring unit, whether the instrument breaks down or not is judged, the dynamic threshold value is obtained through calculation by collecting real-time data of a production site, and therefore the system adapts to dynamic changes of the equipment state and automatically adjusts the dynamic threshold value; and whether an alarm is given is judged according to an abnormal scoring result, so that the accuracy and the real-time performance of fault detection are improved, the automatic processing capability is improved, and finally, the fault probability of the instrument and meter is predicted, so that the effect of timely and accurately processing the fault of the instrument and meter can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment detection, and in particular to an instrument fault detection method based on multi-method fusion. Background Art

[0002] my country's economic construction has achieved significant development, and the corresponding power industry is also making continuous progress. The power industry has entered the fast lane of development, its overall strength is constantly increasing, and the annual power supply has maintained a high growth rate. Therefore, this has continuously increased the requirements for the safe operation of substation operation and maintenance instruments and meters. Many fields use instrument cabinets for equipment installation and control. It is one of the indispensable central control components of subway, power, railway, construction and other systems. Therefore, the detection of instrument cabinet faults is also a necessary requirement.

[0003] Traditional instrument fault detection methods mostly use static thresholds for detection. This method cannot adapt to the dynamic changes in equipment status, resulting in high false alarm and missed alarm rates. Existing fault detection systems still rely on manual intervention and expert experience, lack automated processing capabilities, and are unable to handle instrument faults in a timely and accurate manner and take countermeasures. Summary of the Invention

[0004] The purpose of the present invention is to provide an instrument fault detection method based on multi-method fusion, which solves the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for detecting instrument faults based on multi-method fusion, comprising the following steps:

[0006] Prepare testing tools;

[0007] Data collection and analysis: Use sensors to collect real-time data from instruments and meters, adjust the fault detection threshold of the real-time data through the dynamic threshold adjustment unit, obtain the dynamic threshold DH, and then input the dynamic threshold DH into the abnormal condition scoring unit to obtain the abnormality score S, that is, to determine whether the instrument and meter has a fault;

[0008] Fusion detection: Use the knocking and hand pressure method to perform fusion detection on the instrument to determine the cause and specific location of the instrument failure;

[0009] Fault maintenance and verification: After determining the cause and specific location of the instrument failure, the faulty component of the instrument is replaced, and the dynamic threshold DH and anomaly score S are input into the failure probability prediction unit to obtain the failure probability P, thereby planning maintenance tasks;

[0010] Test report, write instrument fault detection and maintenance report.

[0011] Optionally, the adjustment process of the dynamic threshold adjustment unit is as follows:

[0012] DH t =BH t +α*(Mean t-w +StdDev t-w )

[0013] Where DH is the dynamic threshold;

[0014] DH t is the dynamic threshold at time t;

[0015] BH t is the basic threshold at time t;

[0016] Real-time data includes temperature, conductivity, voltage, and current;

[0017] Mean t-w is the average value of any data in the real-time data in the past w time units;

[0018] StdDev t-w is the standard deviation of any data in the real-time data in the past w time units;

[0019] α is Mean t-w with StdDev t-w The influence coefficient of the sum, the initial value of α is 5;

[0020] The dynamic threshold DH is obtained through real-time data to capture changes in data distribution, identify abnormal data points, and reduce false positives and false negatives caused by data fluctuations and abnormal pattern changes.

[0021] Optionally, the scoring process of the abnormality scoring unit is as follows:

[0022]

[0023] Where S is the abnormality score;

[0024] S t is the anomaly score at time t;

[0025] X t is the measurement value of any instrument at time t;

[0026] DH t is the dynamic threshold at time t;

[0027] Z is a positive number used to prevent the denominator from being zero;

[0028] β is the impact factor, which is adjusted according to the type of anomaly;

[0029] Exception types are divided into runtime exceptions and compile-time exceptions. Runtime exceptions are of lower severity and are caused by code errors. In this case, the β value ranges from 1 to 5.

[0030] Compile-time exceptions are of higher severity and are related to the program running environment and external conditions. In this case, the β value ranges from 6 to 10.

[0031] Set the abnormality score threshold, if S t If the threshold is exceeded, an abnormality occurs at time t and the alarm mechanism is triggered.

[0032] Optionally, the prediction process of the fault probability prediction unit is as follows:

[0033]

[0034] Where P is the probability of failure;

[0035] P t+1 To predict the failure probability at the future time point t+1;

[0036] S i is the abnormality score at time i;

[0037] DH i is the dynamic threshold at time i;

[0038] (S i -DH i ) + Indicates a positive operation. If the value in the brackets is negative, the result is 0. If the value in the brackets is positive, the result is the value itself.

[0039] n is the length of the time period;

[0040] M is a positive number used to prevent the denominator from being zero;

[0041] When P t+1 When the probability of failure is greater than 50%, it is a high probability of failure. t+1 When the probability of failure is less than 50% and greater than 20%, it is a medium probability of failure. t+1 When it is less than 20%, it is a low failure probability;

[0042] Respond according to different failure probabilities. t+1 When the value of α in the dynamic threshold calculation formula is greater than 50%, the value of α changes. The calculation formula of α is as follows:

[0043] α=10-((P t+1 )*5].

[0044] Optionally, the tapping and hand pressing method in the multi-method fusion detection step is to tap or press the plug-in board or component that may be faulty to observe whether the fault recurs.

[0045] Optionally, an appearance inspection, an olfactory inspection and a tactile inspection are performed before the data collection and analysis steps.

[0046] Optionally, the appearance inspection uses a visual inspection to check the appearance of the instrument to look for burnt, discolored, blistering, cracked and damaged components, the olfactory inspection is to smell the instrument for burnt and other abnormal odors, and the tactile inspection is to touch the key components of the instrument to check for overheating, abnormal vibration and looseness.

[0047] Optionally, the P t+1 When the probability of failure is greater than 50%, the risk of equipment failure is high. At this time, the equipment should be shut down immediately, and a comprehensive inspection should be carried out on the equipment to find potential problem points. The prediction results should be reported to the production, equipment and safety departments. Wearing parts should be prepared in advance. Based on the inspection results, a detailed repair plan should be formulated to clarify the repair steps, required time and responsible persons, etc. to ensure that the repair work is carried out in an orderly manner. t+1 When the probability of failure is less than 50% and greater than 20%, it is a medium probability. Strengthen the monitoring and inspection frequency. t+1 When the failure probability is less than 20%, it is considered a low probability and management and maintenance should be carried out according to the normal monitoring and maintenance plan.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. The present invention automatically adjusts the threshold according to the change of the instrument status through the dynamic threshold adjustment unit, ensuring that the anomaly scoring algorithm can accurately reflect the actual status of the instrument. The abnormal condition scoring unit evaluates the degree of abnormality of the system based on real-time data and dynamic thresholds, and determines whether to issue an alarm based on the set anomaly scoring threshold. The change of the dynamic threshold reduces false positives and missed negatives, thereby improving the real-time performance and accuracy of the instrument fault detection method.

[0050] 2. The fault probability prediction unit of the present invention combines with the dynamic threshold adjustment unit and the abnormal condition scoring unit to predict the probability of instrument failure, and prepares corresponding countermeasures according to different failure probabilities, thereby reducing maintenance time and reducing the impact of instrument failure on production. The failure probability results will affect the subsequent dynamic threshold, so that the instrument failure probability prediction algorithm, dynamic threshold algorithm and abnormality scoring algorithm will influence each other in the instrument failure detection method, thereby achieving more accurate fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The figure is a flow chart of the instrument fault detection method of the present invention. DETAILED DESCRIPTION

[0052] Regarding this instrument fault detection method based on multi-method fusion, it is different from existing instrument fault detection methods. Existing instrument fault detection methods often rely on manually set rules and thresholds, and cannot automatically identify abnormal conditions of equipment in real time, resulting in slow response speed and inability to take timely measures, which amplifies losses. This method calculates dynamic thresholds by collecting real-time data from instruments and meters. It can automatically adjust the dynamic thresholds according to changes in the real-time operating conditions of the instruments and meters, avoiding the limitations of static thresholds, and calculating anomaly scores based on the dynamic thresholds to quantitatively evaluate the abnormal conditions of the instruments and meters, thereby improving the accuracy and real-time performance of fault detection.

[0053] 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.

[0054] For examples, see Figure 1 ,This implementation provides an instrument fault detection method based on multi-method fusion as follows:

[0055] Understand the plant control system architecture, instrument functions, wiring diagrams, circuit diagrams, and historical fault records; prepare testing tools, including multimeters, oscilloscopes, signal generators, thermal imagers, and tapping tools, as well as spare parts and replacement components;

[0056] Sensors are used to collect real-time data from instruments and meters, including temperature, conductivity, voltage, and current. A dynamic threshold adjustment unit is used to adjust the fault detection threshold of the real-time data to obtain a dynamic threshold DH. This dynamic threshold DH is then input into an abnormality scoring unit to obtain an abnormality score S, which determines whether the instrument or meter has a fault. If a fault occurs, a corresponding alarm mechanism is triggered.

[0057] The instrument is then tested using the knocking and hand pressure methods to determine the cause and specific location of the instrument failure;

[0058] After determining the cause and specific location of the instrument failure, the faulty component of the instrument is replaced, and the dynamic threshold DH and anomaly score S are input into the failure probability prediction unit to obtain the failure probability P, thereby planning the maintenance task;

[0059] Test report, write a detailed instrument fault detection and maintenance report, including the fault phenomenon, detection process, fault cause, troubleshooting measures and verification results, store the instrument test data and maintenance records to provide reference for subsequent fault analysis and prevention.

[0060] More specifically, in this embodiment: by collecting real-time data from the production site for calculation, a dynamic threshold is obtained, thereby adapting to the dynamic changes in the equipment status and automatically adjusting the dynamic threshold, avoiding the limitations of the static threshold, and then calculating the anomaly score based on the dynamic threshold, quantitatively evaluating the abnormal status of the instrument, and judging whether to issue an alarm based on the anomaly score result, thereby improving the accuracy and real-time performance of fault detection and improving the automation processing capability. At the same time, during the detection process, a comparison method, a fault isolation method, a tapping method, and a hand pressure method are used for multi-method fusion detection to improve the accuracy of fault detection, and the cause and specific location of the fault can be determined, thereby facilitating subsequent maintenance. Finally, by combining the dynamic threshold and the anomaly score to predict the probability of instrument failure, preventive measures can be taken before the instrument failure causes serious consequences, reducing downtime and maintenance costs. By predicting failures in advance, maintenance tasks can be planned, and countermeasures can be taken in advance to reduce the risk of production interruptions or safety accidents caused by sudden failures, and instrument failures can be handled and countermeasures can be taken in a timely and accurate manner.

[0061] Furthermore, the adjustment process of the dynamic threshold adjustment unit is as follows:

[0062] DH t =BH t +α*(Mean t-w +StdDev t-w )

[0063] Where DH is the dynamic threshold;

[0064] DH t is the dynamic threshold at time t;

[0065] BH t is the basic threshold at time t;

[0066] Real-time data includes temperature, conductivity, voltage, and current;

[0067] Mean t-w is the average value of any data in the real-time data in the past w time units;

[0068] StdDev t-w is the standard deviation of any data in the real-time data in the past w time units;

[0069] α is Mean t-w with StdDev t-w The influence coefficient of the sum, the initial value of α is 5;

[0070] The dynamic threshold DH is obtained through real-time data to capture changes in data distribution, identify abnormal data points, and reduce false positives and false negatives caused by data fluctuations and abnormal pattern changes.

[0071] Specifically, by calculating dynamic thresholds and automatically adjusting them according to changes in real-time data, changes in data distribution can be captured, abnormal data points can be identified, and false alarms and missed alarms caused by data fluctuations and abnormal pattern changes can be reduced. In the power operation environment, real-time data may be affected by many factors, such as equipment aging, environmental changes, etc. By calculating dynamic thresholds, real-time adjustments can be made according to different situations to adapt to data changes, reducing system instability problems caused by improper human intervention, thereby improving the stability and reliability of the entire system.

[0072] Furthermore, the calculation formula for the abnormality score in the data collection and analysis step is as follows:

[0073]

[0074] Among them S t is the anomaly score at time t;

[0075] X t is the measurement value of any instrument at time t;

[0076] DH t is the dynamic threshold at time t;

[0077] Z is a positive number used to prevent the denominator from being zero;

[0078] β is the impact factor, which is adjusted according to the type of anomaly;

[0079] Exception types are divided into runtime exceptions and compile-time exceptions. Runtime exceptions are of lower severity and are caused by code errors. In this case, the β value is 3 (any value from 1 to 5). Compile-time exceptions are of higher severity and are related to the program running environment and external conditions. In this case, the β value is 8 (any value from 6 to 10). The value is generally an integer.

[0080] Set the anomaly score threshold to 10. If S t If the threshold is exceeded, an abnormality occurs at time t and the alarm mechanism is triggered.

[0081] Specifically, anomaly scoring quantitatively evaluates real-time data and gives the degree of abnormality of each real-time data relative to the normal state, allowing managers to intuitively understand the abnormal situation of data points. Through real-time processing and giving corresponding anomaly scores, they can promptly discover and issue alarms, reminding staff to take countermeasures, avoid potential risks and losses, and improve the real-time and accuracy of the fault detection system.

[0082] Furthermore, the prediction process of the fault probability prediction unit is as follows:

[0083]

[0084] Where P is the probability of failure;

[0085] P t+1 To predict the failure probability at the future time point t+1;

[0086] S i is the abnormality score at time i;

[0087] DH i is the dynamic threshold at time i;

[0088] (S i -DH i ) + Indicates a positive operation. If the value in the brackets is negative, the result is 0. If the value in the brackets is positive, the result is the value itself.

[0089] n is the length of the time period;

[0090] M is a positive number used to prevent the denominator from being zero;

[0091] When P t+1 When the probability of failure is greater than 50%, it is a high probability of failure. t+1 When the probability of failure is less than 50% and greater than 20%, it is a medium probability of failure. t+1 When it is less than 20%, it is a low failure probability;

[0092] Respond according to different failure probabilities. t+1 When the value of α in the dynamic threshold calculation formula is greater than 50%, the value of α changes. The calculation formula of α is as follows:

[0093] α=10-((P t+1 )*5].

[0094] Specifically, through P t+1 The value of affects the value of α in the dynamic threshold calculation formula. When the predicted probability of instrument failure is higher, it means that the risk of using this equipment is greater. As the probability of instrument failure increases, the α value becomes smaller. For example, when P t+1 When the value is 60%:

[0095] α=10-((60%*5]=10-3=7.

[0096] More specifically, the formula for calculating the probability of instrument failure combines dynamic thresholds and anomaly scores to predict the probability of failure, making it easier to plan maintenance tasks in advance, reduce downtime and repair costs caused by failures, and improve the utilization rate and production efficiency of power equipment. By adjusting the α value, it is easier to adjust the dynamic threshold according to the real-time situation of the instrument and change the anomaly score at the same time, so as to timely detect instrument failures, issue alarms, and remind staff to take countermeasures, so that the instrument failure detection system can handle instrument failures in a timely and accurate manner.

[0097] Furthermore, P t+1 When the probability of failure is greater than 50%, the risk of equipment failure is high. At this time, the equipment should be shut down immediately, and a comprehensive inspection should be carried out on the equipment to find potential problem points. The prediction results should be reported to the production, equipment and safety departments. Wearing parts should be prepared in advance. Based on the inspection results, a detailed repair plan should be formulated to clarify the repair steps, required time and responsible persons, etc. to ensure that the repair work is carried out in an orderly manner. t+1 When the probability of failure is less than 50% and greater than 20%, it is a medium probability. Strengthen the monitoring and inspection frequency. t+1 When the failure probability is less than 20%, it is considered a low probability and management and maintenance should be carried out according to the normal monitoring and maintenance plan.

[0098] Specifically, by formulating corresponding response measures in advance based on the different probabilities of instrument failure, we can respond quickly when a failure is predicted, which greatly saves preparation time and reduces the impact of instrument failure on production.

[0099] See also Figure 1 ,The knocking and hand pressing methods in the fusion detection steps are to tap or press the ,possible faulty plug-in board or component to observe whether the fault ,recurs.

[0100] Specifically, the dynamic threshold adjustment method and abnormal condition scoring method can be used to determine whether the instrument is faulty. By combining the tapping and hand pressure methods, the fault location can be accurately found and the cause of the fault can be determined, facilitating subsequent maintenance work.

[0101] Furthermore, visual inspection, olfactory inspection and tactile inspection were performed before the data collection and analysis steps.

[0102] Specifically, by first performing an olfactory and tactile inspection, it is possible to more quickly determine whether the instrument is faulty, so as to facilitate subsequent fault detection. For example, if a burning odor is smelled before the instrument is operated, wiring processing will no longer be performed, otherwise a safety accident may easily occur.

[0103] Furthermore, the appearance inspection uses visual inspection to check the appearance of the instrument to find burnt, discolored, blistering, cracked and damaged components. The olfactory inspection is to smell the burnt and other abnormal odors of the instrument. The tactile inspection is to touch the key components of the instrument to check for overheating, abnormal vibration and looseness of the instrument.

[0104] Specifically, by inspecting instruments from different aspects, we can directly understand the actual operating status of equipment, systems or processes. Some problems may not be discovered through normal testing, but can be discovered through observation, such as abnormal sounds, odors, color changes, etc. Through appearance inspection, olfactory inspection and tactile inspection, maintenance and repair work can be carried out more specifically, avoiding blind operation and ineffective labor, and improving maintenance efficiency and quality.

[0105] 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 instrument fault detection method based on multi-method fusion, characterized in that: The following steps are involved: Step S1: preparing detection tools; Step S2: Data collection and analysis: Using sensors to collect real-time data from instruments and meters, the dynamic threshold adjustment unit adjusts the fault detection threshold of the real-time data to obtain a dynamic threshold DH. The dynamic threshold DH is then input into the abnormality scoring unit to obtain an abnormality score S, i.e., to determine whether the instrument and meter has a fault; Step S3: Fusion detection, using the knocking and hand pressure method to perform fusion detection on the instrument to determine the cause and specific location of the instrument failure; Step S4: Fault maintenance and verification: After determining the cause and specific location of the instrument failure, the faulty component of the instrument is replaced, and the dynamic threshold DH and anomaly score S are input into the failure probability prediction unit to obtain the failure probability P, thereby planning the maintenance task; Step S5: Test report, write instrument fault detection and maintenance report.

2. The instrumentation fault detection method based on multi-method fusion according to claim 1 is characterized in that: The adjustment process of the dynamic threshold adjustment unit is as follows: DH t =BH t +α*(Mean t-w +StdDev t-w ) Where DH is the dynamic threshold; DH t is the dynamic threshold at time t; BH t is the basic threshold at time t; Real-time data includes temperature, conductivity, voltage, and current; Mean t-w is the average value of any data in the real-time data in the past w time units; StdDev t-w is the standard deviation of any data in the real-time data in the past w time units; α is Mean t-w with StdDev t-w The influence coefficient of the sum, the initial value of α is 5; The dynamic threshold DH is obtained through real-time data to capture changes in data distribution, identify abnormal data points, and reduce false positives and false negatives caused by data fluctuations and abnormal pattern changes.

3. The instrumentation fault detection method based on multi-method fusion according to claim 2 is characterized in that: The scoring process of the abnormal condition scoring unit is as follows: Where S is the abnormality score; S t is the anomaly score at time t; X t is the measurement value of any instrument at time t; DH t is the dynamic threshold at time t; Z is a positive number used to prevent the denominator from being zero; β is the impact factor, which is adjusted according to the type of anomaly; Exception types are divided into runtime exceptions and compile-time exceptions. Runtime exceptions are of lower severity and are caused by code errors. In this case, the β value ranges from 1 to 5. Compile-time exceptions are of higher severity and are related to the program running environment and external conditions. In this case, the β value ranges from 6 to 10. Set the abnormality score threshold, if S t If the threshold is exceeded, an abnormality occurs at time t and the alarm mechanism is triggered.

4. The instrument fault detection method based on multi-method fusion according to claim 3 is characterized by: The prediction process of the fault probability prediction unit is as follows: Where P is the probability of failure; P t+1 To predict the failure probability at the future time point t+1; S i is the abnormality score at time i; DH i is the dynamic threshold at time i; (S i -DH i ) + Indicates a positive operation. If the value in the brackets is negative, the result is 0. If the value in the brackets is positive, the result is the value itself. n is the length of the time period; M is a positive number used to prevent the denominator from being zero; When P t+1 When the probability of failure is greater than 50%, it is a high probability of failure. t+1 When the probability of failure is less than 50% and greater than 20%, it is a medium probability of failure. t+1 When it is less than 20%, it is a low failure probability; Respond according to different failure probabilities. t+1 When the value of α in the dynamic threshold calculation formula is greater than 50%, the value of α changes. The calculation formula of α is as follows: α＝10-((P t+1 )*5]。 5. The instrument fault detection method based on multi-method fusion according to claim 1 is characterized in that: The tapping and hand pressing method in the multi-method fusion detection step is to tap or press the plug-in board or component that may have a fault to observe whether the fault recurs.

6. The instrumentation fault detection method based on multi-method fusion according to claim 1 is characterized in that: The data collection and analysis steps are preceded by visual inspection, olfactory inspection, and tactile inspection.

7. The instrumentation fault detection method based on multi-method fusion according to claim 6 is characterized in that: The appearance inspection uses visual inspection to check the appearance of the instrument to find burnt, discolored, blistering, cracked and damaged components; The olfactory examination is to smell the burnt odors and other abnormal odors of instruments; The tactile inspection is to touch the key components of the instrument to check for overheating, abnormal vibration and looseness of the instrument.

8. The instrumentation fault detection method based on multi-method fusion according to claim 4 is characterized in that: The P t+1 When the probability of failure is greater than 50%, the risk of equipment failure is high. In this case, the equipment should be shut down immediately and a comprehensive inspection should be conducted to identify potential problems. The predicted results should be reported to the production, equipment, and safety departments. Consumable parts should be prepared in advance. Based on the inspection results, a detailed repair plan should be developed, clearly defining the repair steps, required time, and responsible persons to ensure that the repair work is carried out in an orderly manner. When P t+1 When the probability of failure is less than 50% and greater than 20%, which is a medium probability, the monitoring and inspection frequency should be strengthened; When P t+1 When the failure probability is less than 20%, it is considered a low probability and management and maintenance should be carried out according to the normal monitoring and maintenance plan.

Images