A circuit breaker fault self-diagnosis method and system based on electrical signal characteristics and fault tree

Abstract

This invention discloses a circuit breaker fault self-diagnosis method and system based on electrical signal characteristics and fault trees. The method acquires the original current signal during the circuit breaker's operation, employs an adaptive bidirectional threshold triggering algorithm to search from both ends of the sampling time axis towards the center, determines the starting trigger point and the attenuation termination point to extract the effective action sequence signal; based on this signal, it calculates the action time, effective current value, statistical peak current, and Joule integral to construct a four-dimensional state feature vector; based on typical circuit breaker fault modes, it constructs a fault tree model, mapping its logical relationships to production-based diagnostic rules in an expert system knowledge base; the four-dimensional state feature vector is input into the expert system inference engine, and the rules are matched using a forward inference mechanism to output the fault diagnosis result. This invention reduces the personnel costs of on-site inspection and the potential damage caused by disassembly through online fault identification and location, significantly reducing operation and maintenance costs, and is applicable to online fault self-diagnosis of universal circuit breakers.

Description

Technical Field

[0001] This invention belongs to the technical field of intelligent power distribution switchgear, specifically relating to a circuit breaker fault self-diagnosis method and system based on electrical signal characteristics and fault trees. Background Technology

[0002] As the core control and protection equipment of low-voltage power distribution systems, the operating status of universal circuit breakers directly affects the safety and stability of the entire power distribution network. Due to their highly complex internal mechanical structure and numerous components, the operation and maintenance of circuit breakers has traditionally relied mainly on manual periodic inspections. This approach suffers from problems such as high maintenance costs, indiscriminate disassembly of easily damaged equipment, and existing monitoring methods being limited to basic physical quantities such as voltage and current, which cannot provide in-depth information about the mechanical condition.

[0003] To improve the above situation, technologies utilizing operating current signals for circuit breaker fault diagnosis have emerged in recent years. For example, Chinese patent application CN104502837A discloses a method for diagnosing mechanical faults in circuit breakers. This method detects the current waveform of an electromagnet coil, extracts the instantaneous current amplitude at preset characteristic time points and corresponding moments, and matches it with a typical waveform database to determine the fault type. However, electromagnetic interference is severe in actual industrial settings, and the operating current signal of a circuit breaker is often superimposed with high-frequency noise and random pulse spikes. Existing current signal-based diagnostic methods generally suffer from insufficient accuracy in extracting the effective range of the operating current waveform, easily introducing static zero drift or losing effective data, thus affecting subsequent analysis. The extracted current features are highly sensitive to noise interference, resulting in poor feature stability. Furthermore, the diagnostic logic relies on empirical matching or data-driven models, lacking physical interpretability and struggling to handle complex or gradual faults. Therefore, there is an urgent need to improve existing technologies to address the technical problems of inaccurate processing of operating current signals in complex electromagnetic environments, weak anti-interference capability of feature extraction, and poor interpretability of diagnostic reasoning. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a circuit breaker fault self-diagnosis method and system based on electrical signal characteristics and fault trees, so as to solve the problems of current feature extraction being susceptible to interference, inaccurate truncation intervals, and lack of interpretability of diagnostic logic in complex electromagnetic environments.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault trees, comprising the following steps: S100: Acquire the original current signal during the circuit breaker operation process, and process the original current signal using an adaptive bidirectional threshold triggering algorithm to extract the valid operation sequence signal; S200: Calculate and extract action time, effective current value, statistical peak current and Joule integral based on the effective action sequence signal, and construct a four-dimensional state feature vector; S300: Construct a fault tree model based on typical fault modes of circuit breakers, and map the logical relationship of the fault tree to the production diagnostic rules in the expert system knowledge base. The production diagnostic rules use the threshold criteria of each feature in the four-dimensional state feature vector as conditions. S400: Input the four-dimensional state feature vector into the expert system inference engine, use the forward inference mechanism to match the production rule, and output the fault diagnosis result.

[0006] The present invention further proposes that the adaptive bidirectional threshold triggering algorithm is specifically as follows: Set preset current threshold For the total number of discrete sampling points M The original current sequence A bidirectional search is performed from both ends of the sampling time axis toward the center to determine the starting trigger point of the original current sequence. and attenuation termination point Specifically: According to the starting trigger point and attenuation termination point The effective length obtained is N The effective action sequence signal .

[0007] The present invention further proposes that the extraction of action time, effective current value, statistical peak current, and Joule integral in step S200 specifically includes: The discrete current sequence within the effective action sequence signal The total number of sampling points is N The sampling interval is ; The action time according to calculate; The effective value of the current according to calculate; The statistical peak current The cumulative distribution function is calculated from the absolute value sequence of the discrete current sequence. Extracting the required information The value is taken as the statistical peak current, where This is a preset percentage threshold; Joule integral according to Calculated.

[0008] The present invention further proposes that the preset percentage threshold The value ranges from 0.97 to 0.995.

[0009] The present invention further proposes that the circuit breaker includes an energy storage motor, and the base events of the fault tree model include multiple failure modes of the energy storage motor; the mapping of the logical relationships of the fault tree to production rules in the expert system knowledge base specifically involves: treating each logic gate of the fault tree as a unit, converting each unit into a corresponding production rule, and all production rules constituting the expert system knowledge base. The production rules, based on the relationship between each feature quantity in the four-dimensional state feature vector and a preset threshold, correspond to different failure mode determination conditions.

[0010] The present invention further proposes to establish the following production-based diagnostic rules for energy storage motors: When the action time is greater than the first preset time threshold, or the statistical peak current is greater than the first preset current amplitude threshold, or the effective current value is greater than the second preset current amplitude threshold, the motor is determined to be aging. When the statistical peak current is less than the third preset current amplitude threshold and the effective current value is greater than the fourth preset current amplitude threshold, the rotor is determined to be stuck. When the statistical peak current is greater than the fifth preset current amplitude threshold and the effective current value is greater than the fourth preset current amplitude threshold, it is determined that the gear is stuck. When the effective value of the current is less than the sixth preset current amplitude threshold, it is determined to be a voltage abnormality.

[0011] The present invention further proposes that the forward reasoning mechanism in step S400 includes: using the four-dimensional state feature vector as the initial fact and matching it with the condition part of the production rule; when the feature quantity satisfies the threshold criterion set in the rule, the rule is triggered and the corresponding fault type is output as the conclusion fact.

[0012] Secondly, the present invention also provides a circuit breaker fault self-diagnosis system based on electrical signal characteristics and fault trees, comprising: The signal preprocessing module is used to acquire the raw current signal during the circuit breaker operation process and to process the raw current signal using an adaptive bidirectional threshold triggering algorithm. Specifically, a preset current threshold is set, and a bidirectional search is performed from both ends of the sampling time axis toward the center to determine the starting point where the absolute value of the current first exceeds the preset current threshold and the attenuation termination point where the absolute value of the current last exceeds the preset current threshold. The waveform between the starting point and the attenuation termination point is extracted as the effective action sequence signal. The feature extraction module is used to calculate and extract the action time, effective current value, statistical peak current and Joule integral based on the effective action sequence signal, and construct a four-dimensional state feature vector. The knowledge base module is used to construct a fault tree model based on typical fault modes of circuit breakers, and to map the logical relationship of the fault tree to the production diagnostic rules in the expert system knowledge base. The production diagnostic rules use the threshold criteria of each feature in the four-dimensional state feature vector as conditions. The reasoning and diagnosis module is used to input the four-dimensional state feature vector into the expert system inference engine, match the production rule using a forward reasoning mechanism, and output the fault diagnosis result.

[0013] This invention discloses a circuit breaker fault self-diagnosis method and system based on electrical signal characteristics and fault trees. Its core lies in the organic combination of adaptive signal preprocessing, multi-dimensional statistical energy feature extraction, and interpretable reasoning based on fault trees. Compared to existing technologies, this invention offers the following advantages: It accurately locates the effective operating range through a bidirectional threshold triggering algorithm, eliminating background noise and static zero-drift interference in industrial settings; it uses statistical peak current instead of traditional absolute maximum value, improving the anti-interference capability of amplitude characteristics; it introduces Joule integral to represent the total energy consumption of the operating process, helping to comprehensively reflect the mechanical state of the circuit breaker; the production-based diagnostic rules constructed based on fault trees have clear physical mechanism support, resulting in highly interpretable diagnostic results without requiring a large number of training samples; and the overall algorithm has low computational complexity, enabling direct deployment on resource-constrained embedded trip unit terminals to achieve real-time online diagnosis and precise location of circuit breaker faults. More importantly, this invention achieves accurate fault identification and location by realizing online fault self-diagnosis, changing the traditional model of relying on manual periodic inspections and significantly reducing the personnel costs of on-site testing and inspection. At the same time, accurate fault location avoids secondary damage to equipment that may be caused by blind disassembly and repair, effectively and significantly reducing the overall operation and maintenance costs of circuit breakers. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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.

[0015] Figure 1 This is a flowchart illustrating the circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault trees provided by the present invention. Figure 2 The waveform of the original current before truncation; Figure 3 The waveform of the effective operating current after truncation; Figure 4 The overall architecture diagram of the expert system provided for this invention; Figure 5 This is a schematic diagram of the circuit breaker fault self-diagnosis system based on electrical signal characteristics and fault tree provided by the present invention.

[0016] The following are the markings in the attached diagram: 10. Signal preprocessing module; 20. Feature extraction module; 30. Knowledge base module; 40. Reasoning and diagnosis module. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0019] Existing circuit breaker fault diagnosis technologies suffer from problems such as inaccurate signal interception, poor feature anti-interference capability, and lack of interpretability of diagnostic results in complex industrial environments, making it difficult to meet the high reliability requirements of smart grid equipment operation and maintenance. Therefore, to improve upon these issues, this invention provides a circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault trees, such as... Figure 1 As shown: Firstly, this invention proposes a circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault trees, such as... Figure 1 As shown, it includes the following steps: S100: Acquire the original current signal during the circuit breaker operation process, and process the original current signal using an adaptive bidirectional threshold triggering algorithm to extract the valid operation sequence signal; S200: Calculate and extract action time, effective current value, statistical peak current and Joule integral based on the effective action sequence signal, and construct a four-dimensional state feature vector; S300: Construct a fault tree model based on typical fault modes of circuit breakers, and map the logical relationship of the fault tree to the production diagnostic rules in the expert system knowledge base. The production diagnostic rules use the threshold criteria of each feature in the four-dimensional state feature vector as conditions. S400: Input the four-dimensional state feature vector into the expert system inference engine, use the forward inference mechanism to match the production rule, and output the fault diagnosis result.

[0020] The circuit breaker operating current signal is acquired by a Hall current sensor installed in the circuit breaker's energy storage motor or electromagnet coil circuit. The sampling frequency is set according to the circuit breaker's operating speed, typically not lower than 1kHz. The raw current signal includes the current change throughout the entire circuit breaker operation process, superimposed with electromagnetic interference noise and static zero drift from the industrial environment. The adaptive bidirectional threshold triggering algorithm, by searching simultaneously from both ends of the time axis, can accurately identify the start and end times of the operation, avoiding the interval truncation errors caused by noise interference in unidirectional triggering. The four-dimensional state feature vector reflects the mechanical state of the circuit breaker from different dimensions, where the operating time characterizes the mechanism's movement speed, the effective current value reflects the average load, the statistical peak current reflects the instantaneous impact, and the Joule integral reflects the total energy consumption. The fault tree model is established based on the circuit breaker's failure mechanism, clarifying the causal relationship between each fault type and electrical signal characteristics. After being transformed into production rules, it can be directly called by the expert system's inference engine. The forward inference mechanism uses the extracted feature vectors as factual basis, matching each diagnostic rule in the rule base, triggering rules that meet the conditions, and outputting the corresponding fault type.

[0021] Specifically, the adaptive bidirectional threshold triggering algorithm in step S100 is based on searching from both ends of the time axis towards the center to determine the starting point when the absolute value of the current first exceeds the preset threshold and the attenuation termination point when it last exceeds the preset threshold. This bidirectional search method avoids the problem of premature or missed triggering caused by small fluctuations in the initial segment of the signal in unidirectional threshold triggering. It ensures that the captured effective action sequence signal completely contains the operating current waveform of the circuit breaker and removes static zero drift and redundant data before and after, providing a clean data foundation for subsequent feature extraction.

[0022] In step S200, based on the captured effective action sequence signal, four characteristic quantities with clear physical meaning are calculated: action time, reflecting the speed of the circuit breaker's mechanical action; effective current value, reflecting the average work intensity throughout the entire action process; statistical peak current, extracted by calculating the cumulative distribution function from the absolute value sequence of the discrete current sequence; and Joule integral, reflecting the total energy consumption during the action process. These four characteristic quantities characterize the electromechanical characteristics of the circuit breaker's action from three dimensions: time, amplitude, and energy. They are all independent of instantaneous values ​​and possess good anti-interference capabilities.

[0023] In step S300, a fault tree model is first established based on typical circuit breaker failure modes (such as motor aging, rotor jamming, gear jamming, and voltage anomalies) to determine the logical relationship between the bottom event (fault cause) and the top event (fault manifestation). Then, each logic gate (such as an AND gate or an OR gate) in the fault tree is treated as a unit, and each unit is converted into a corresponding production rule. All production rules constitute the expert system knowledge base. Each rule uses the relationship between each feature quantity in the four-dimensional state feature vector and a preset threshold as a condition, and the fault type as the conclusion. For example, for an energy storage motor, a rule can be established: if the operating time is greater than a first preset time threshold, or the statistical peak current is greater than a first preset current amplitude threshold, or the effective current value is greater than a second preset current amplitude threshold, then the motor is determined to be aging.

[0024] In step S400, the forward reasoning mechanism works as follows: the four-dimensional state feature vector extracted in step S200 is used as the initial fact and matched sequentially with the conditional parts of the production rules in the knowledge base. When all conditions of a rule are met, the rule is triggered and the corresponding fault type is output as the conclusion fact; if multiple rules match, the most likely fault type is output according to priority or confidence. The entire reasoning process is transparent and traceable, and each conclusion can be traced back to a specific feature threshold criterion.

[0025] The above technical solution improves the noise resistance of feature extraction by combining bidirectional search truncation with statistical peak current; it provides clear physical interpretability to the diagnostic process through the mapping from fault trees to production rules; and all calculations are simple time-domain operations that can run in real time on embedded microprocessors. Therefore, this solution effectively solves the problems of feature extraction being susceptible to interference, inaccurate truncation intervals, and lack of interpretability in diagnostic logic in existing technologies.

[0026] The present invention further proposes that the adaptive bidirectional threshold triggering algorithm is specifically as follows: Set preset current threshold For the total number of discrete sampling points M The original current sequence A bidirectional search is performed from both ends of the sampling time axis toward the center to determine the starting trigger point of the original current sequence. and attenuation termination point Specifically: (1) (2) According to the starting trigger point and attenuation termination point The effective length obtained is N The effective action sequence signal .

[0027] Among them, the preset current threshold The setting is based on the circuit breaker's rated operating current and the site noise level, typically ranging from 1% to 5% of the rated current. The process involves searching from the beginning of the sampling time axis towards the center to find the first current with an absolute value greater than... The sampling point is used as the starting trigger point; simultaneously, the search proceeds from the end of the sampling time axis towards the center to find the last current with an absolute value greater than [missing value]. The sampling point is used as the attenuation termination point. All sampling points between these two points are extracted to obtain the effective current sequence containing only the circuit breaker's operating process, eliminating static zero drift and background noise before and after the operation. This bidirectional search method ensures the integrity of the effective operating range, avoids false triggering caused by noise in the initial segment during unidirectional triggering, and does not miss the attenuation process in the later stage of the operation, providing a reliable guarantee for the accuracy of subsequent feature extraction.

[0028] As one specific implementation method, the adaptive truncation threshold set in this embodiment ,like Figure 2 The original waveform before truncation showed obvious zero drift, which was obtained after bidirectional search. Figure 3 The clean waveform shown is an example. Compared to the conventional unidirectional rising edge triggering, this method reduces the truncation error by approximately an order of magnitude.

[0029] The present invention further proposes that the extraction of action time, effective current value, statistical peak current, and Joule integral in step S200 specifically includes: The discrete current sequence within the effective action sequence signal The total number of sampling points is N The sampling interval is ; The action time according to calculate; The effective value of the current according to calculate; The statistical peak current The cumulative distribution function is calculated from the absolute value sequence of the discrete current sequence. Extracting the required information The value is given as the statistical peak current, where This is a preset percentage threshold; Joule integral according to Calculated.

[0030] Wherein, cumulative distribution function Defined as F ( i )= P ( X ≤ i ), that is, random variables X (Absolute value of current) The value shall not exceed i The probability of the circuit breaker's operation is as follows: The operating time reflects the total time required for the circuit breaker to complete its entire operation; wear and jamming of mechanical components can prolong the operating time. The RMS current characterizes the average current during operation, reflecting the average load on the motor. Statistical peak current is extracted using probabilistic statistical methods, effectively filtering out the influence of random pulse noise and more accurately reflecting the true peak current level. The Joule integral is the integral of the square of the current over time, directly reflecting the total energy consumed during operation; increased resistance due to mechanical faults will increase the Joule integral. These four features reflect the mechanical state of the circuit breaker from different perspectives, complementing each other to form a complete state characterization system, effectively distinguishing different types of fault modes. The above feature extraction process does not involve complex frequency domain transformations or iterative optimizations, requiring only one numerical sorting and several multiplication and addition operations, which can be completed within milliseconds, making it very suitable for the real-time execution of embedded trip units.

[0031] The present invention further proposes that the preset percentage threshold The value ranges from 0.97 to 0.995.

[0032] Understandable, percentage threshold α This determines the probability level represented by the statistical peak current. α If the value is too small, it will excessively remove valid data, causing the peak characteristic to be underestimated; if α If the value is too large, it will fail to effectively filter noise spikes, thus losing the significance of statistical peak values. Those skilled in the art can adjust this value according to the severity of the electromagnetic environment: a smaller value should be selected in noisy environments. α A value of 0.97 would be more robust, but might sacrifice minute sensitivity; a larger value could be chosen in low-noise environments. α(e.g., 0.995) to obtain an estimate closer to the true peak value. This range of values ​​has been verified through simulation and can cover the needs of most industrial sites. For example, when α When the value is 0.990, the statistical peak current indicates that the absolute value of 99% of the current sampling points is less than this value. Only the top 1% of abnormal peak data is removed, which can effectively filter out random pulse noise without losing the true peak information.

[0033] The present invention further proposes that the circuit breaker includes an energy storage motor, and the base events of the fault tree model include multiple failure modes of the energy storage motor; the mapping of the logical relationships of the fault tree to production rules in the expert system knowledge base specifically involves: treating each logic gate of the fault tree as a unit, converting each unit into a corresponding production rule, and all production rules constituting the expert system knowledge base. The production rules, based on the relationship between each feature quantity in the four-dimensional state feature vector and a preset threshold, correspond to different failure mode determination conditions.

[0034] The energy storage motor is the core component of the closing energy storage mechanism in a universal circuit breaker. Typical failure modes include motor aging, rotor jamming, gear jamming, and voltage anomalies. Each of these faults produces unique characteristic changes in the current waveform: motor aging leads to prolonged operating time and increased effective current; rotor jamming causes the disappearance of the starting impact peak but a larger steady-state current; gear jamming causes an abnormally large starting impact peak; and voltage anomalies proportionally reduce the overall current amplitude. The fault tree model uses "energy storage motor failure" as the top event, decomposing it downwards into intermediate events such as "motor aging," "rotor jamming," "gear jamming," and "voltage anomalies." Each intermediate event is further decomposed into corresponding characteristic quantity anomalies as bottom events. The AND gates in the fault tree correspond to the logical AND relationship of multiple conditions in the production rules, and the OR gates correspond to the logical OR relationship of multiple conditions in the production rules. Through this mapping method, the logical structure of the fault tree is directly transformed into rules executable by the expert system, ensuring that the diagnostic process fully follows the physical mechanism of the fault occurrence.

[0035] The matching process between the above production rules and feature data is as follows: Figure 4The expert system architecture shown is implemented in this paper. The system consists of six core components: feature extraction, real-time feature database, inference engine, knowledge base, interpreter, and output report. After calculating the four-dimensional state feature vector, the feature extraction module 20 stores the standardized feature data in the real-time feature database as the initial facts for inference. The inference engine retrieves the current feature facts from the real-time feature database and loads all pre-stored production rules from the knowledge base, performing rule matching one by one according to the forward inference mechanism. During the matching process, the triggered rules and their corresponding fault conclusions are synchronously transmitted to the interpreter. The interpreter records the triggering conditions and logical basis of each rule, generating an interpretable diagnostic report containing the fault cause, feature manifestations, and mechanism explanations, which is finally presented to maintenance personnel through the output report module. The knowledge base and inference engine employ a bidirectional interactive design, supporting dynamic updates and expansions of rules. When adding a fault mode or optimizing threshold parameters, only the production rules in the knowledge base need to be modified; there is no need to adjust the core algorithm of the inference engine, improving the maintainability and versatility of the system.

[0036] The present invention further proposes to establish the following production-based diagnostic rules for energy storage motors: When the action time is greater than the first preset time threshold, or the statistical peak current is greater than the first preset current amplitude threshold, or the effective current value is greater than the second preset current amplitude threshold, the motor is determined to be aging. When the statistical peak current is less than the third preset current amplitude threshold and the effective current value is greater than the fourth preset current amplitude threshold, the rotor is determined to be stuck. When the statistical peak current is greater than the fifth preset current amplitude threshold and the effective current value is greater than the fourth preset current amplitude threshold, it is determined that the gear is stuck. When the effective value of the current is less than the sixth preset current amplitude threshold, it is determined to be a voltage abnormality.

[0037] The preset thresholds are determined based on the rated parameters of the corresponding circuit breaker model, statistical data under normal operating conditions, and experimental data of typical fault modes. Different rated power and different models of energy storage motors correspond to different threshold sets. The first preset time threshold is the upper limit of the time required for the circuit breaker's energy storage motor to complete the energy storage action normally. When the motor windings age, the bearings wear, leading to increased internal resistance or mechanical resistance, the action time will exceed this threshold. The first and second preset current amplitude thresholds are the upper limits of the statistical peak current and the effective current value under normal conditions, respectively. Motor aging will cause the overall operating current to increase, and meeting either condition can be judged as an aging fault. The third preset current amplitude threshold is the lower limit of the normal starting impact peak value. When the rotor is stuck, the motor cannot rotate, there is no starting impact process, and the statistical peak current will be lower than this threshold. The fourth preset current amplitude threshold is the lower limit of the motor stall current. Both rotor jamming and gear jamming will cause the motor to stall, and the effective current value will be higher than this threshold. The fifth preset current amplitude threshold is the upper limit of the normal starting surge peak. When gears jam, the motor must overcome enormous mechanical resistance at startup, generating a surge current far exceeding normal levels. The statistical peak current exceeds this threshold. The sixth preset current amplitude threshold is the lower limit of the motor's normal operating current. When the power supply voltage is abnormally low, the motor's operating current will decrease proportionally. If it falls below this threshold, it can be considered a voltage anomaly. These thresholds can be adaptively calibrated based on on-site operating data or manually updated through the expert system's knowledge base module 30 to adapt to different operating environments and equipment aging levels.

[0038] The present invention further proposes that the forward reasoning mechanism in step S400 includes: using the four-dimensional state feature vector as the initial fact and matching it with the condition part of the production rule; when the feature quantity satisfies the threshold criterion set in the rule, the rule is triggered and the corresponding fault type is output as the conclusion fact.

[0039] Specifically, firstly, all known facts (i.e., four-dimensional feature values) are stored in the fact base; then, each rule in the rule base is traversed, and its condition is checked to see if it is satisfied by the current fact base; if satisfied, the conclusion (fault type) of the rule is added to the fact base as a new fact; the above process is repeated until no new facts are generated or the preset reasoning depth is reached. Because the rule base of this invention is relatively small (usually no more than a few dozen rules) and the reasoning direction is forward data-driven, the entire matching process can be completed in milliseconds, making it suitable for real-time fault diagnosis.

[0040] Secondly, the present invention also provides a circuit breaker fault self-diagnosis system based on electrical signal characteristics and fault trees, such as... Figure 5 As shown, it includes: The signal preprocessing module 10 is used to acquire the original current signal during the circuit breaker operation process and process the original current signal using an adaptive bidirectional threshold triggering algorithm. Specifically, it sets a preset current threshold, performs a bidirectional search from both ends of the sampling time axis towards the center, determines the starting point where the absolute value of the current first exceeds the preset current threshold and the attenuation termination point where the absolute value of the current last exceeds the preset current threshold, and extracts the waveform between the starting point and the attenuation termination point as the effective action sequence signal. Feature extraction module 20 is used to calculate and extract action time, effective current value, statistical peak current and Joule integral based on the effective action sequence signal, and construct a four-dimensional state feature vector; The knowledge base module 30 is used to construct a fault tree model based on typical fault modes of circuit breakers, and to map the logical relationship of the fault tree to the production diagnosis rules in the expert system knowledge base. The production diagnosis rules use the threshold criteria of each feature in the four-dimensional state feature vector as conditions. The reasoning and diagnosis module 40 is used to input the four-dimensional state feature vector into the expert system inference engine, match the production rule using the forward reasoning mechanism, and output the fault diagnosis result.

[0041] The above modules can be integrated into the intelligent trip unit of the circuit breaker and implemented using an embedded microprocessor. The signal preprocessing module 10 typically consists of a current sensor (such as a Hall element) and an analog-to-digital converter; the feature extraction module 20, the knowledge base module 30, and the reasoning and diagnosis module 40 can run on the MCU through software algorithms. The entire system can independently complete fault diagnosis without an external host computer, resulting in low cost and high reliability.

[0042] Verification of Examples This embodiment uses a certain model of intelligent universal circuit breaker (ACB) with a rated current of 2000A as the experimental and protection object. This circuit breaker is equipped with an embedded intelligent trip unit, a Hall current sampling module, and an energy storage motor. The energy storage motor has a rated operating voltage of AC220V and a rated power of 100W. The experimental platform simulates abnormal voltage fluctuations using an adjustable voltage source, and simulates typical failure modes such as mechanism jamming and rotor / gear jamming by artificially setting frictional resistance and mechanical limits in the transmission mechanism.

[0043] This method relies on the electromechanical energy conversion physical model of a universal circuit breaker. The specific diagnostic implementation steps are as follows: The first step is to perform signal preprocessing and effective interval truncation. Because the raw operating current signal collected by the Hall current sensor in industrial settings is superimposed with high-frequency harmonics and background noise, an adaptive truncation threshold is set. For discrete sequences The algorithm searches for trigger points bidirectionally. (3) max (4) The truncation process eliminates the cumulative contamination of subsequent feature integrals caused by static zero drift.

[0044] The second step is to extract a multi-dimensional state feature vector within the effective action range. In this embodiment, after capturing the effective action range, the microprocessor calculates and extracts the action time, effective current value, statistical peak current, and Joule integral to construct a four-dimensional state feature vector. Action Time Based on the total number of valid interval sampling points N and the sampling interval Calculation, formula: (5) RMS value of current : Represents the average work intensity throughout the entire action, formula: (6) in It is a discrete current sequence within the effective operating range.

[0045] Statistical peak current To eliminate the distortion of peak characteristics by extreme impact noise, a preset percentage coefficient is used in this embodiment. The preferred value is 0.99. The cumulative distribution function is calculated from the absolute values ​​of the current sequence. Extracting the required information The value is used as the statistical peak current (i.e., the 99th percentile peak) to replace the traditional absolute maximum value.

[0046] Joule integral : Reflects the total energy consumption during the action process, formula: (7) The third step is to build the expert system knowledge base.

[0047] This paper analyzes the working principle of the energy storage motor and trip unit of the universal circuit breaker and the physical transmission path of closing / opening faults, and establishes a fault tree model. The base events of the fault tree are mapped to the production diagnostic rules in the expert system knowledge base, and specific hard threshold criteria are set for the feature quantities of each dimension.

[0048] The fourth step is the design and diagnostic output of the expert system inference engine.

[0049] Using the four-dimensional state feature vector as data input, a forward data-driven reasoning mechanism is employed to match rules in the knowledge base. In this embodiment, the specific diagnostic rules and preferred thresholds for the energy storage motor are set as follows: like or or If the action time is too long or the peak or effective value is too high, the first rule is triggered, and the motor is judged to be aging.

[0050] like and (i.e., lack of starting impact peak, but with a large steady-state current), then the second rule is triggered, and the rotor is judged to be stuck.

[0051] like and (That is, if the starting impact peak is abnormally large and the steady-state current is large), then the third rule is triggered, and the gear is judged to be stuck.

[0052] like If the overall operating current is too low, the fourth rule is triggered, and the voltage is determined to be abnormal.

[0053] To verify the effectiveness and robustness of this invention, this embodiment collected 300 sets of circuit breaker operation data (including normal state and various typical fault states) in an industrial environment, and compared the performance of the method of this invention with that of traditional threshold diagnosis methods. The results are as follows: Robustness analysis of feature extraction: In the presence of high-frequency electromagnetic interference in industrial environments, traditional methods extract peak current using an absolute maximum function. However, due to noise spikes, the characteristic value deviates from the true value by more than 15%, leading to severe false triggering. This invention, however, uses statistical peak current... ( The top 1% of abnormal spikes were eliminated using probability distribution. Experimental data show that the extracted feature values ​​have a good fit of over 98% to the actual waveform envelope, effectively shielding against environmental noise interference.

[0054] Fault classification and diagnosis accuracy analysis: The collected samples were automatically diagnosed using an expert system inference engine. The identification results for various states are shown in Table 1. Table 1. Statistical Table of Fault Diagnosis Test Results for Universal Circuit Breakers under Typical Conditions Experimental results show that the diagnostic method based on multidimensional feature vectors and expert systems provided in this invention can not only achieve real-time deployment at resource-constrained embedded trip units, but also achieve an average classification accuracy of 98.4% for complex electromechanical faults, which is superior to traditional single-threshold monitoring schemes. The mapping system established by this method from physical features to logical rules provides a scientific and intuitive mechanistic basis for predictive maintenance of circuit breakers. Furthermore, this system pre-emptively and online performs fault diagnosis, effectively eliminating the cumbersome on-site manual inspection process through high-precision online fault identification and location, saving significant personnel costs, and fundamentally avoiding the equipment damage risk caused by traditional blind disassembly. It has extremely high engineering application value and a significant advantage in reducing operation and maintenance costs.

[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault tree, characterized in that, Includes the following steps: Step S100: Obtain the original current signal during the circuit breaker operation process, and process the original current signal using an adaptive bidirectional threshold triggering algorithm to extract the effective operation sequence signal; Step S200: Calculate and extract the action time, effective current value, statistical peak current and Joule integral based on the effective action sequence signal, and construct a four-dimensional state feature vector; Step S300: Construct a fault tree model based on typical fault modes of circuit breakers, and map the logical relationship of the fault tree to the production diagnostic rules in the expert system knowledge base. The production diagnostic rules use the threshold criteria of each feature in the four-dimensional state feature vector as conditions. Step S400: Input the four-dimensional state feature vector into the expert system inference engine, use the forward inference mechanism to match the production rule, and output the fault diagnosis result.

2. The circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault tree according to claim 1, characterized in that, The adaptive bidirectional threshold triggering algorithm is specifically as follows: Set preset current threshold For the total number of discrete sampling points M The original current sequence The original current sequence is determined by performing a bidirectional search from both ends of the sampling time axis toward the center. The starting trigger point and attenuation termination point Specifically: According to the starting trigger point and attenuation termination point The effective length obtained is N The effective action sequence signal .

3. The circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault tree according to claim 1, characterized in that, The extraction of action time, RMS current, statistical peak current, and Joule integral in step S200 specifically includes: The discrete current sequence within the effective action sequence signal The total number of sampling points is N The sampling interval is ; The action time according to calculate; The effective value of the current according to calculate; The statistical peak current The cumulative distribution function is calculated from the absolute value sequence of the discrete current sequence. Extracting the required information The value is used as the statistical peak current, where This is a preset percentage threshold; Joule integral according to Calculated.

4. The circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault tree according to claim 3, characterized in that, The preset percentage threshold The value ranges from 0.97 to 0.

995.

5. The circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault tree according to claim 1, characterized in that, The circuit breaker includes an energy storage motor, and the base events of the fault tree model include multiple failure modes of the energy storage motor. The mapping of the logical relationships of the fault tree to the production rules in the expert system knowledge base specifically involves: treating each logic gate of the fault tree as a unit, converting each unit into a corresponding production rule, and all production rules constituting the expert system knowledge base. The production rules correspond to different failure mode judgment conditions based on the relationship between each feature quantity in the four-dimensional state feature vector and a preset threshold.

6. The circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault tree according to claim 5, characterized in that, The following production-based diagnostic rules are established for energy storage motors: When the action time is greater than the first preset time threshold, or the statistical peak current is greater than the first preset current amplitude threshold, or the effective current value is greater than the second preset current amplitude threshold, the motor is determined to be aging. When the statistical peak current is less than the third preset current amplitude threshold and the effective current value is greater than the fourth preset current amplitude threshold, the rotor is determined to be stuck. When the statistical peak current is greater than the fifth preset current amplitude threshold and the effective current value is greater than the fourth preset current amplitude threshold, it is determined that the gear is stuck. When the effective value of the current is less than the sixth preset current amplitude threshold, it is determined to be a voltage abnormality.

7. The circuit breaker fault self-diagnosis method based on electrical signal characteristics and fault tree according to claim 1, characterized in that, The forward reasoning mechanism in step S400 includes: using the four-dimensional state feature vector as initial facts and matching it with the condition part of the production rule; when the feature quantity meets the threshold criterion set in the rule, the rule is triggered and the corresponding fault type is output as the conclusion fact.

8. A circuit breaker fault self-diagnosis system based on electrical signal characteristics and fault tree, characterized in that, include: The signal preprocessing module is used to acquire the raw current signal during the circuit breaker operation process and to process the raw current signal using an adaptive bidirectional threshold triggering algorithm. Specifically, a preset current threshold is set, and a bidirectional search is performed from both ends of the sampling time axis toward the center to determine the starting point where the absolute value of the current first exceeds the preset current threshold and the attenuation termination point where the absolute value of the current last exceeds the preset current threshold. The waveform between the starting point and the attenuation termination point is extracted as the effective action sequence signal. The feature extraction module is used to calculate and extract the action time, effective current value, statistical peak current and Joule integral based on the effective action sequence signal, and construct a four-dimensional state feature vector. The knowledge base module is used to construct a fault tree model based on typical fault modes of circuit breakers, and to map the logical relationship of the fault tree to the production diagnostic rules in the expert system knowledge base. The production diagnostic rules use the threshold criteria of each feature in the four-dimensional state feature vector as conditions. The reasoning and diagnosis module is used to input the four-dimensional state feature vector into the expert system inference engine, match the production rule using a forward reasoning mechanism, and output the fault diagnosis result.

Images