A fault diagnosis method and system for a voltage regulation module

Abstract

This invention discloses a fault diagnosis method and system for a voltage regulator module, belonging to the field of fault detection technology. The method specifically includes: periodically collecting key operating parameters of the voltage regulator module; performing preliminary anomaly detection on the collected key operating parameters using a multi-dimensional dynamic threshold model; the preliminary anomaly detection is based on the deviation of the current operating state from the historical normal trajectory; when a preliminary anomaly is determined, an anomaly trend analysis is performed within a time window; if the anomaly continues to exceed a preset time threshold, the current anomaly state is compared with a preset fault feature matrix to identify whether the voltage regulator module has a fault; and a corresponding fault label is generated based on the identified fault type. This application can not only accurately capture latent, nonlinear, or multi-dimensional coupled faults, but also has strong adaptive capabilities, enabling early warning and significantly improving the operational reliability and maintainability of the voltage regulator module under complex operating conditions.

Description

Technical Field

[0001] This invention belongs to the field of fault detection technology, specifically a fault diagnosis method and system for a voltage regulation module. Background Technology

[0002] As a core component of power supply systems, voltage regulation modules are widely used in key fields such as industrial control, electric vehicles, communication base stations, and server power supplies.

[0003] In practical applications, voltage regulation modules need to maintain stable output under various uncertainties such as dynamic load changes, input fluctuations, electromagnetic interference, and thermal coupling effects. These modules typically include multiple functional subsystems such as feedback control loops, voltage sampling, error amplification, PWM modulation, inductor filtering, and thermal protection. If any component fails or degrades, it can easily lead to abnormal output, and in severe cases, damage to downstream circuits or system downtime.

[0004] Current fault detection methods for voltage regulation modules mainly rely on passive fault response based on hardware protection and fault determination methods based on rule thresholds and logical judgments. These methods have the following drawbacks: fixed thresholds, poor adaptability, lack of fault semantic modeling, and difficulty in detecting early fault trends. Therefore, there is an urgent need for a fault diagnosis method for voltage regulation modules. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a fault diagnosis method and system for voltage regulation modules.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A fault diagnosis method for a voltage regulating module includes:

[0008] Periodically collect key operating parameters of the voltage regulation module;

[0009] The collected key operating parameters are subjected to preliminary anomaly detection using a multidimensional dynamic threshold model. The preliminary anomaly detection is based on the deviation of the current operating state from the historical normal trajectory.

[0010] When a preliminary anomaly is identified, an anomaly trend analysis is performed within a time window. If the anomaly continues to exceed a preset time threshold, the current anomaly state is compared with a preset fault feature matrix to identify whether the voltage regulation module is faulty.

[0011] Based on the identified fault type, a corresponding fault label is generated.

[0012] Specifically, the preliminary anomaly detection of the collected key operating parameters using a multidimensional dynamic threshold model includes:

[0013] The key operating parameters of the collected voltage regulation module are preprocessed, including moving average and noise reduction;

[0014] The preprocessed key operating parameters are categorized and mapped to the corresponding parameter nodes in the control-parameter semantic graph, and their position level and controlled relationship in the control logic chain are recorded.

[0015] A time-related vector is constructed for the mapped key operating parameters, and the time-related vector is embedded into the path node sequence of the control-parameter semantic graph using a similarity function;

[0016] Calculate the evolution rate of the embedded trajectory and the cross-node transition distance, assess the deviation of the current operating state from the historical normal trajectory, and judge the preliminary anomaly based on the deviation.

[0017] Specifically, the process of classifying and mapping the preprocessed key operating parameters to corresponding parameter nodes in the control-parameter semantic graph, and recording their position hierarchy and controlled relationship in the control logic chain, includes:

[0018] Based on the control topology of the voltage regulation module, a mapping table of the interaction between various operating parameters and corresponding control units is predefined, and a control-parameter semantic graph is constructed. The operating parameters include voltage, current, temperature, duty cycle and feedback error signal.

[0019] The preprocessed key operating parameters are matched to functional nodes in the control-parameter semantic map, including adjustment nodes, feedback nodes, driving nodes, and controlled nodes;

[0020] Each key operating parameter node is labeled with its hierarchical identifier and causal direction in the control link. The hierarchical identifier is automatically numbered based on the signal flow path from the control input to the final response output, and the causal direction describes the upstream and downstream dependencies of the key operating parameter in the control loop.

[0021] Specifically, the step of constructing a time-related vector for the mapped key operating parameters and embedding the time-related vector into the path node sequence of the control-parameter semantic graph using a similarity function includes:

[0022] The mapped key operating parameters are combined into time series vector segments according to the control-response time sequence, and then causally ordered according to the control flow topology to form a state vector sequence with semantic labels.

[0023] The similarity between the state vector sequence and the historical trajectory template is calculated to generate the index position of the most matching path in the control-parameter semantic graph;

[0024] Embed the state vector sequence into the corresponding path node sequence.

[0025] Specifically, the calculation of the evolution rate of the embedded trajectory and the cross-node transition distance, and the assessment of the deviation of the current operating state from the historical normal trajectory, include:

[0026] Based on the timestamps and topological positions of each node in the embedded trajectory, the evolution velocity vector between nodes is calculated, where the velocity vector is the unit time displacement of the embedded state in the control-parameter semantic graph.

[0027] Based on the jump frequency and path connectivity between nodes in the state trajectory, a cross-node jump distance index is defined.

[0028] A set of baseline evolution channels is constructed using a historical normal state trajectory library, and the velocity direction and transition distance of the current state trajectory in the evolution space are compared with the baseline channels using cosine similarity and amplitude offset.

[0029] If the current trajectory deviates significantly from the historical evolutionary channel across all indicators, it is marked as an abnormal evolutionary state; otherwise, it is marked as a normal evolutionary state.

[0030] Specifically, when an initial anomaly is identified, an anomaly trend analysis is performed within a time window. If the anomaly persists beyond a preset time threshold, the current anomaly state is compared with a preset fault feature matrix to identify whether the voltage regulation module is faulty, including:

[0031] After being identified as a preliminary anomaly, the anomaly trend monitoring window is used to monitor the anomaly trend.

[0032] Based on the abnormal trend monitoring results, a parameter abnormal evolution vector is generated and compared with a preset multi-fault state prototype as a comparative input.

[0033] The parameter anomaly evolution vector is matched with the fault vector template in the preset fault feature matrix by similarity, and the preset fault feature matrix is ​​constructed according to the fault type.

[0034] When the similarity is higher than the set recognition threshold and the matching result remains stable for n sampling cycles, the voltage regulation module is determined to have malfunctioned, and the identified fault type is output.

[0035] Specifically, the abnormal trend monitoring window records parameter offset, rate of change, and duration of abnormality through a sliding time window.

[0036] Specifically, the step of performing similarity matching between the parameter anomaly evolution vector and the fault vector template in the preset fault feature matrix includes:

[0037] The parameter anomaly evolution vector is divided into multi-dimensional sub-vector groups according to the parameter category, including voltage characteristic sub-vectors, current characteristic sub-vectors, thermal characteristic sub-vectors, and control response sub-vectors;

[0038] Based on the feature space dimension of each sub-vector group, the predefined fault template vector in the fault feature matrix is ​​deconstructed in the same structure so that it has the corresponding sub-vector structure and metric semantics.

[0039] Using a composite similarity calculation method, the distance scores between the parameter anomaly evolution vector and the sub-vectors in the fault feature matrix are calculated separately, and then fused to obtain the global similarity.

[0040] A fault diagnosis system for a voltage regulator module, used to implement the fault diagnosis method for the voltage regulator module, includes: a parameter acquisition module, a preliminary detection module, a fault diagnosis module, and a tag generation module;

[0041] The parameter acquisition module is used to periodically acquire key operating parameters of the voltage regulation module;

[0042] The preliminary detection module is used to perform preliminary anomaly detection on the collected key operating parameters using a multi-dimensional dynamic threshold model.

[0043] The fault diagnosis module is used to perform an abnormal trend analysis within a time window when the fault is determined to be a preliminary abnormality. If the abnormality continues to exceed a preset time threshold, the current abnormal state is compared with a preset fault feature matrix to identify whether the voltage regulation module has a fault.

[0044] The tag generation module is used to generate corresponding fault tags based on the identified fault type.

[0045] Specifically, the fault diagnosis module includes: a trend monitoring unit, a similarity matching unit, and a fault diagnosis unit;

[0046] The trend monitoring unit is used to monitor abnormal trends using the abnormal trend monitoring window after being identified as a preliminary abnormality.

[0047] The similarity matching unit is used to perform similarity matching between the parameter anomaly evolution vector and the fault vector template in the preset fault feature matrix;

[0048] The fault diagnosis unit is used to determine that the voltage regulation module has failed when the similarity is higher than the set recognition threshold and the matching result is stable for n sampling cycles, and outputs the identified fault type.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] 1. This invention proposes a fault diagnosis method and system for voltage regulation modules. By constructing a control-parameter semantic graph of the voltage regulation module, the collected multi-source operating data is mapped to a structure with causal relationships. Combined with state trajectory embedding, abnormal evolution trend analysis, and fault feature matrix comparison, dynamic identification and judgment of fault states are achieved. This method can not only accurately capture latent, nonlinear, or multidimensional coupled faults, but also has strong adaptive capabilities and model interpretability, enabling early warning and significantly improving the operational reliability and maintainability of the voltage regulation module under complex operating conditions. It overcomes the problems of high misjudgment rate and inability to identify early faults in traditional methods based on fixed thresholds or single parameters. Attached Figure Description

[0051] Figure 1 A flowchart of a fault diagnosis method for a voltage regulating module provided by the present invention;

[0052] Figure 2 This invention provides a fault diagnosis system architecture diagram for a voltage regulation module. Detailed Implementation

[0053] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] It should be noted that, unless there is conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be performed in a different order than the module division in the device or the order in the flowchart. In addition, the "

[0056] The terms "first," "second," and "third" do not limit the data or execution order; they are merely used to distinguish identical or similar items with essentially the same function and purpose.

[0057] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0058] Example 1

[0059] Please see Figure 1 The present invention provides an embodiment of a fault diagnosis method for a voltage regulating module, comprising the following specific steps:

[0060] Step S1: Periodically collect key operating parameters of the voltage regulation module.

[0061] In this embodiment, the parameters periodically collected include input voltage, output voltage, output current, module internal temperature, control loop reference voltage, and PWM drive signal duty cycle. The sampling period can be set to 100μs, and the sampled data is refreshed in real time using a sliding window.

[0062] Step S2: Perform preliminary anomaly detection on the collected key operating parameters using a multidimensional dynamic threshold model. The preliminary anomaly detection is based on the deviation of the current operating state from the historical normal trajectory.

[0063] The specific steps of step S2 are as follows:

[0064] Step S201: Preprocess the key operating parameters of the collected voltage regulation module, including moving average and noise reduction.

[0065] Step S202: Classify and map the preprocessed key operating parameters to the corresponding parameter nodes in the control-parameter semantic graph, and record their position level and controlled relationship in the control logic chain.

[0066] The specific steps of step S202 are as follows:

[0067] Step S2021: Based on the control topology of the voltage regulation module, predefine the mapping relationship table between various operating parameters and corresponding control units, and construct a control-parameter semantic graph. The operating parameters include voltage, current, temperature, duty cycle and feedback error signal.

[0068] In this embodiment, the internal control structure of the voltage regulation module follows a closed control chain of target setting, error calculation, modulation response, and electrical behavior feedback. Each physical parameter in this chain has both measurement attributes and control meaning. Its control semantics depends on its functional level and transmission path. By analyzing the circuit schematic and control logic flowchart, the input-output relationship of each control unit is identified, and the operating parameters are represented as nodes in the graph. A one-to-one mapping relationship between control and physical quantities is established. Based on the control behavior dependence and direction of action, the mapping is extended into a directed graph with structural semantics, thus forming a control-parameter semantic graph.

[0069] For example, taking a closed-loop Buck-type voltage regulation module as an example, its control topology includes: a voltage feedback network, an error amplifier, a PWM controller, a power MOS driver stage, an inductor filter, and an output sampling circuit. In this structure, the operating parameters are mapped according to their position and control dependence. For example, the output voltage Vout is the negative feedback input of the error amplifier, and its stability depends on the real-time adjustment of the PWM duty cycle D. The duty cycle is calculated by the controller based on the deviation between the reference voltage Vref and the feedback voltage. Although the output current Iout does not directly participate in the control closed loop, it is related to the load boundary conditions. The module temperature T is affected by the operating state of the power devices and is fed back to the PWM modulator to trigger the current limiting or frequency reduction mechanism when the MOS on-resistance increases. The above mapping forms a directed graph structure with the control unit, the controlled parameters, and the feedback path as the directions.

[0070] Step S2022: Match the preprocessed key operating parameters to the functional nodes in the control-parameter semantic graph, wherein the functional nodes include adjustment nodes, feedback nodes, driving nodes and controlled nodes.

[0071] In this embodiment, the matching process involves establishing a control semantic mapping relationship between operating parameters and functions. By analyzing the signal flow and control causal chain in the electrical control path, and combining the control level where the node is located, the structural function is assigned. Specifically, the regulating node has input characteristics and participates in the setting of the desired state; the feedback node provides control state feedback and has a direct impact on the closed-loop stability; the driving node embodies the modulator response behavior and acts unidirectionally towards the power stage; the controlled node is located at the tail of the control chain and passively changes in response to other control behaviors.

[0072] In specific implementation, taking Vref as an example, it serves as the control target input and directly forms an adjustment path with the output of the error amplifier, thus being classified as an adjustment node; Vout is fed back to the error amplifier through a resistor voltage divider, forming a closed-loop comparison relationship with Vref, thus being classified as a feedback node; the duty cycle D output by the PWM driver is a control quantity that directly acts on the power MOSFET, and its control behavior has forced drive characteristics, thus being classified as a drive node; although Vout and Iout participate in feedback regulation, they are at the controlled end in the control topology, thus being classified as controlled nodes.

[0073] Step S2023: Label each key operating parameter node with its hierarchical identifier and causal direction in the control link. The hierarchical identifier is automatically numbered based on the signal flow path from the control input to the final response output, and the causal direction describes the upstream and downstream dependencies of the key operating parameter in the control loop.

[0074] In this embodiment, the control chain of the voltage regulation module is essentially a self-consistent signal flow network. Each parameter node has its own definite position and functional path. The essence of the hierarchical identification is to establish a transmission depth sequence of control information, so that the upstream and downstream control relationships between parameters are sortable. This sequence is constructed by traversing the directed edges in the control semantic graph, automatically numbering from the source node to the terminal node, generating a hierarchical mapping table on the control path topology. On this basis, the definition of causal direction is automatically generated according to the action path in the control logic.

[0075] For example, in a closed-loop Buck-type voltage regulation module, the control link starts from the control reference input (such as Vref), generates a control deviation signal (Verr) through an error amplifier, and then the PWM controller adjusts the duty cycle D, which acts on the drive circuit to control the conduction state of the power MOSFET, thereby affecting the output voltage Vout and the output current Iout. By analyzing the above signal flow path, each key operating parameter node is numbered and layered according to the control conduction direction. For example, Vref is marked as level 0 as the control starting point, Verr is level 1, D is level 2, Vout and Iout are level 3, and Tj is level 4. Vref, Verr, D, and Vout constitute a complete unidirectional control link.

[0076] Step S203: Construct a time-related vector for the mapped key operating parameters, and embed the time-related vector into the path node sequence of the control-parameter semantic graph using a similarity function.

[0077] The specific steps of step S203 are as follows:

[0078] Step S2031: Combine the mapped key operating parameters into time series vector segments according to the control-response time sequence, and sort them causally according to the control flow topology to form a state vector sequence with semantic labels.

[0079] In this embodiment, the operating state of the voltage regulation module is not a superposition of isolated variables, but a state evolution process composed of parameters with temporal and structural dependencies. It is ordered semantically based on the control flow topology. If parameter A has an impact on the control path before parameter B, then A should be arranged before B in the state vector. This principle ensures that the state vector has directionality and interpretability, that is, the evolution trend of the state can be quantitatively analyzed through the structural changes of the vector segments.

[0080] In practical applications, after time alignment of the key operating parameters collected by the voltage regulation module, a state vector with structural semantics needs to be constructed according to its control response relationship. Taking each 50 sampling periods as a data window, the parameter nodes (such as Vref, Verr, D, Vout, Iout, T) that have been mapped to the semantic graph are combined according to the control logic, where Vref precedes Verr and D precedes Vout. The parameter arrangement order in each time step is: [Vref, Verr, D, Vout, Iout, T].

[0081] Step S2032: Calculate the similarity between the state vector sequence and the historical trajectory template to generate the index position of the most matching path in the control-parameter semantic graph.

[0082] In this embodiment, traditional single-parameter matching methods cannot identify such cross-parameter coupling dynamics. Therefore, similarity calculation needs to consider three dimensions: the relative morphological differences in the trend of parameter sequences; the structural consistency of time delay and coupling rhythm between different parameters; whether the control flow semantic ordering constraints within the alignment window are preserved; and by jointly comparing the current state vector with the template trajectory in the structural space and time domain, and based on the definition of each path in the control-parameter semantic graph, finding the best-fitting trajectory segment, the position of the running state on the control topology graph can be obtained.

[0083] Step S2033: Embed the state vector sequence into the corresponding path node sequence.

[0084] In this embodiment, the embedding of the state vector sequence is essentially a mapping process from behavioral fragments to structural nodes. It aims to identify whether the current dynamic state is reasonably projected to a certain control logic stage. This process constructs an embedding mapping function to encode the temporal characteristics, signal coupling degree, and control path sequence position of the state vector into a node embedding vector. This vector is then matched with the historical embedding representation of the node in the graph to ensure the semantic rationality of the embedding.

[0085] Step S204: Calculate the evolution rate of the embedded trajectory and the cross-node transition distance, assess the deviation of the current running state from the historical normal trajectory, and judge the preliminary abnormal situation based on the deviation.

[0086] The specific steps of step S204 are as follows:

[0087] Step S2041: Calculate the evolution velocity vector between nodes based on the timestamps and topological positions of each node in the embedded trajectory. The velocity vector is the unit time displacement of the embedded state in the control-parameter semantic graph.

[0088] In this embodiment, each node in the control-parameter semantic graph has structural coordinates, reflecting its positional attributes in the control logic chain, such as hierarchical depth, control role type, and adjacency relationship. The trajectory of the voltage regulation module's operating state is embedded in this graph, which is essentially the motion track of the state point in a structured space. By dividing the topological spacing between consecutive embedded points by the corresponding sampling time interval, a unit-time displacement vector is constructed, which can quantify the motion speed and direction change rate of the trajectory. The velocity vector reveals the migration rhythm of the voltage regulation module's state in the control graph space. When the control system is running stably, its evolution speed should show a low amplitude and smooth direction trend; while when it is disturbed or driven abnormally, the state transition will show abrupt changes or jumps across layers, which will be reflected in the velocity vector as instantaneous high amplitude or sudden structural direction changes.

[0089] Step S2042: Define the cross-node jump distance index based on the jump frequency and path connectivity between nodes in the state trajectory.

[0090] In this embodiment, the topological connectivity between nodes in the control-parameter semantic graph reflects the reachability and structural coupling degree of the control logic. The normal state trajectory should evolve sequentially along the logical order of the control chain or the path adjacent to the feedback loop. Its structural jumps should be limited to the local subgraph of the graph. When the state trajectory frequently crosses nodes with no direct control causal relationship, such as jumping from the pre-regulation stage to the output stage or directly from the drive unit to the thermal feedback path, it indicates that its evolution process no longer follows the established control order. The jump distance index is used to measure the structural discontinuity and logical spacing in the node jump event, and to evaluate the structural stability of the trajectory in combination with the jump frequency. The longer the path, the farther the level, and the greater the functional difference in the graph structure, the higher the risk of the jump. When such jumps are frequently concentrated in a short period of time, they can be regarded as early abnormal signals.

[0091] Step S2043: Construct a set of baseline evolution channels using the historical normal state trajectory library, and compare the velocity direction and transition distance of the current state trajectory in the evolution space with the baseline channels using cosine similarity and amplitude offset.

[0092] In this embodiment, during operation, the voltage regulation module forms repeatable, structurally closed evolutionary trajectories. These trajectories are represented in the graph as state embedding sequences on fixed paths, exhibiting characteristics such as consistent direction, smooth transitions, and stable rates. By clustering and extracting vectors from these trajectories, a standard set of baseline evolutionary channels can be formed. The velocity direction of the current trajectory can be compared with the channel velocity vector to obtain cosine similarity, reflecting whether its evolutionary direction is consistent. The magnitude offset of the transition distance measures whether the trajectory exhibits structural transitions beyond historical channels in the graph structure. This comparison method avoids sensitivity to absolute numerical values ​​and focuses more on evolutionary trends and path patterns, making it suitable for identifying non-mutational, slow-onset, and structural anomalies.

[0093] Step S2044: If the current trajectory deviates significantly from the historical evolution channel in all indicators, it is marked as an abnormal evolution state; otherwise, it is marked as a normal evolution state.

[0094] Step S3: When a preliminary abnormality is identified, an abnormality trend analysis is performed within the time window. If the abnormality continues to exceed the preset time threshold, the current abnormality status is compared with the preset fault feature matrix to identify whether the voltage regulation module has a fault.

[0095] The specific steps of step S3 are as follows:

[0096] Step S301: After being identified as a preliminary anomaly, anomaly trend monitoring is performed using an anomaly trend monitoring window. The anomaly trend monitoring window records parameter offset, rate of change, and duration of anomaly through a sliding time window.

[0097] In this embodiment, abnormal trends and instantaneous anomalies are fundamentally different. The latter is caused by sudden interference, sensor jitter, or load disturbance, while the former is manifested as a continuous, slow, or non-periodic deviation of the operating state from the stable range, exhibiting significant time-cumulative characteristics. The trend monitoring window aims to capture this continuous evolutionary characteristic, treating the abnormal state as a time-coupled pattern in a time series, rather than an isolated event. The offset reflects the overall deviation of the current state from the target or steady state. The rate of change characterizes the trend direction and the acceleration of rise / fall. The duration of the anomaly serves as a cumulative indicator to determine whether the deviation constitutes an early manifestation of fault evolution. The sliding window structure enables real-time, low-latency monitoring of these time-related behaviors.

[0098] Step S302: Generate a parameter anomaly evolution vector based on the anomaly trend monitoring results, and compare it with the preset multi-fault state prototype as a comparison input.

[0099] In this embodiment, an anomaly in a single electrical parameter is insufficient to determine whether the voltage regulation module has entered a fault state. A comprehensive analysis combining its evolutionary characteristics and the correlation patterns between multiple parameters is necessary. The parameter anomaly evolution vector is a structured encoding of this behavioral process. Since anomalies in control systems typically manifest as both persistent trends and abrupt changes in coupling characteristics, each component of the vector extracts deviation attributes of the current state in amplitude, rate, duration, and multi-parameter coupling, transforming them into a standard input format convenient for numerical comparison. Fault type vectors in the prototype library are extracted from historical labeled data or simulation conditions. The comparison process does not rely on absolute deviations in parameter values ​​but uses the similarity of behavioral patterns as the core judgment criterion.

[0100] Step S303: Perform similarity matching between the parameter anomaly evolution vector and the fault vector template in the preset fault feature matrix, wherein the preset fault feature matrix is ​​constructed according to the fault type.

[0101] The specific steps of step S303 are as follows:

[0102] Step S3031: Divide the parameter anomaly evolution vector into multi-dimensional sub-vector groups according to the parameter category, including voltage characteristic sub-vectors, current characteristic sub-vectors, thermal characteristic sub-vectors, and control response sub-vectors.

[0103] In this embodiment, the principle of the classification method is as follows: voltage and current types are usually significantly affected by load disturbances, thermal characteristics reflect the accumulation of energy loss, and control response parameters dominate the dynamic adjustment process of the entire closed loop.

[0104] Step S3032: Based on the feature space dimension of each sub-vector group, perform homostructure deconstruction on the predefined fault template vector in the fault feature matrix so that it has the corresponding sub-vector structure and metric semantics.

[0105] In this embodiment, the dominant physical domain affected by each fault type during its evolution is analyzed, and then the template vector is structurally mapped based on the dimensional label system of the observation vector to ensure that subsequent comparisons can be performed locally in their respective subspaces.

[0106] Step S3033: Using the composite similarity calculation method, calculate the distance scores between the parameter anomaly evolution vector and the sub-vectors in the fault feature matrix, and fuse them to obtain the global similarity.

[0107] In this embodiment, if only a single distance metric, such as Euclidean distance, is used for similarity calculation, the correlation between parameters may be easily overlooked; while using Mahalanobis distance alone will amplify the covariance noise in high-dimensional data. The similarity evaluation adopts a composite similarity calculation method with dual constraints of structure and magnitude. At the sub-vector level, Mahalanobis distance is used to reveal the distributional synergy of parameter variations and to determine whether a behavioral cointegration pattern consistent with a certain type of fault is presented; the Euclidean part supplements the reflection of the degree of deviation of the absolute value. The similarity scores of each dimension are weighted and fused to obtain a comprehensive score that reflects the overall state pattern matching degree while retaining the sensitivity of each sub-vector.

[0108] Step S304: When the similarity is higher than the set recognition threshold and the matching result is stable for n sampling cycles, the voltage regulation module is determined to have failed, and the identified fault type is output.

[0109] Step S4: Generate the corresponding fault label based on the identified fault type.

[0110] Example 2

[0111] Please see Figure 2 Another embodiment of the present invention provides: a fault diagnosis system for a voltage regulating module, comprising: a parameter acquisition module, a preliminary detection module, a fault diagnosis module, and a tag generation module;

[0112] The parameter acquisition module is used to periodically acquire key operating parameters of the voltage regulation module;

[0113] The preliminary detection module is used to perform preliminary anomaly detection on the collected key operating parameters using a multi-dimensional dynamic threshold model.

[0114] The fault diagnosis module is used to perform an abnormal trend analysis within a time window when the fault is determined to be a preliminary abnormality. If the abnormality continues to exceed a preset time threshold, the current abnormal state is compared with a preset fault feature matrix to identify whether the voltage regulation module has a fault.

[0115] The tag generation module is used to generate corresponding fault tags based on the identified fault type.

[0116] The fault diagnosis module includes: a trend monitoring unit, a similarity matching unit, and a fault diagnosis unit;

[0117] The trend monitoring unit is used to monitor abnormal trends using the abnormal trend monitoring window after being identified as a preliminary abnormality.

[0118] The similarity matching unit is used to perform similarity matching between the parameter anomaly evolution vector and the fault vector template in the preset fault feature matrix;

[0119] The fault diagnosis unit is used to determine that the voltage regulation module has failed when the similarity is higher than the set recognition threshold and the matching result is stable for n sampling cycles, and outputs the identified fault type.

[0120] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0121] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for diagnosing a failure of a pressure regulating module, characterized by, The method comprises the following steps: Periodically collecting key operating parameters of the voltage regulating module; Performing preliminary anomaly detection on the collected key operating parameters through a multi-dimensional dynamic threshold model, wherein the preliminary anomaly detection is based on the deviation of the current operating state from the historical normal trajectory; When a preliminary anomaly is detected, performing anomaly trend analysis within a time window, and if the anomaly lasts for more than a preset time threshold, comparing the current anomaly state with a preset fault feature matrix to identify whether the voltage regulating module has a fault; Generating a corresponding fault label according to the identified fault type; The preliminary anomaly detection on the collected key operating parameters through the multi-dimensional dynamic threshold model comprises the following steps: Performing preprocessing on the collected key operating parameters of the voltage regulating module, including moving average and denoising; According to the control topology structure of the voltage regulating module, predefining an action mapping relationship table between various operating parameters and corresponding control units, constructing a control-parameter semantic graph, wherein the operating parameters include voltage, current, temperature, duty cycle and feedback error signal, and each node in the control-parameter semantic graph has a structure coordinate, reflecting its position attribute in the control logic chain, including hierarchical depth, control role type and adjacency relationship; Categorizing and mapping the preprocessed key operating parameters to the corresponding parameter nodes in the control-parameter semantic graph, and recording the position level and controlled relationship in the control logic chain; Constructing a time correlation vector for the mapped key operating parameters, and embedding the time correlation vector into the path node sequence of the control-parameter semantic graph using a similarity function; According to the timestamps and topological positions of the nodes in the embedded trajectory, calculating the evolution speed vector between the nodes, wherein the evolution speed vector is the unit time displacement of the embedded state in the control-parameter semantic graph; Defining the cross-node transition distance based on the jump frequency and path connectivity between the nodes in the embedded trajectory; Calculating the evolution speed vector and the cross-node transition distance of the embedded trajectory, evaluating the deviation of the current operating state from the historical normal trajectory, and determining the preliminary anomaly according to the deviation.

2. The fault diagnosis method for a voltage regulating module as described in claim 1, characterized in that, The categorizing and mapping of the preprocessed key operating parameters to the corresponding parameter nodes in the control-parameter semantic graph, and the recording of the position level and controlled relationship in the control logic chain, comprises the following steps: According to the control topology structure of the voltage regulating module, predefining an action mapping relationship table between various operating parameters and corresponding control units, constructing a control-parameter semantic graph, wherein the operating parameters include voltage, current, temperature, duty cycle and feedback error signal; Matching the preprocessed key operating parameters to the functional nodes in the control-parameter semantic graph, wherein the functional nodes include adjustment nodes, feedback nodes, driving nodes and controlled nodes; Labeling each key operating parameter node with its hierarchical identifier and causal direction in the control link, wherein the hierarchical identifier is automatically numbered based on the signal flow path from the control input to the final response output, and the causal direction describes the upstream and downstream dependency relationship of the key operating parameter in the control loop.

3. The fault diagnosis method for a voltage regulating module as described in claim 2, characterized in that, The mapped key operation parameters construct a time correlation vector, and the time correlation vector is embedded into a path node sequence of a control-parameter semantic graph by using a similarity function, including: The mapped key operation parameters are combined into a time sequence vector segment according to a control-response time sequence, and are causally sorted according to a control flow topology to form a state vector sequence with a semantic label; Similarity calculation is performed on the state vector sequence and a historical trajectory template to generate an index position of a most matched path in the control-parameter semantic graph; The state vector sequence is embedded into a corresponding path node sequence.

4. The fault diagnosis method for a voltage regulating module as described in claim 3, characterized in that, The evolution speed vector and the cross-node transition distance of the embedded trajectory are calculated to evaluate the deviation degree of the current operation state relative to the historical normal trajectory, including: According to the timestamps and the topological positions of the nodes in the embedded trajectory, the evolution speed vector between the nodes is calculated, and the evolution speed vector is the unit time displacement of the embedded state in the control-parameter semantic graph; Based on the jump frequency and the path connectivity between the nodes in the embedded trajectory, the cross-node transition distance is defined, and the evolution speed vector and the cross-node transition distance of the current state trajectory are obtained based on the evolution speed vector and the cross-node transition distance of the embedded trajectory; A set of reference evolution channels is constructed by using a historical normal state trajectory library, and the cosine similarity is obtained by comparing the evolution speed vector of the current state trajectory in the evolution space with the speed vector of the reference evolution channel, and the amplitude offset is obtained by comparing the cross-node transition distance of the current state trajectory in the evolution space with the cross-node transition distance of the reference evolution channel; If the current state trajectory significantly deviates from the reference evolution channel in all indicators, the state is marked as an abnormal evolution state, otherwise, the state is marked as a normal evolution state.

5. The fault diagnosis method for a voltage regulating module as described in claim 4, characterized in that, When the state is determined to be preliminary abnormal, abnormal trend analysis in a time window is performed, and if the abnormality lasts more than a preset time threshold, the current abnormal state is compared with a preset fault feature matrix to identify whether the voltage regulating module has a fault, including: After being determined to be preliminary abnormal, abnormal trend monitoring is performed by using an abnormal trend monitoring window; A parameter abnormal evolution vector is generated according to the abnormal trend monitoring result, and is compared with a preset multi-fault state prototype; Similarity matching is performed on the parameter abnormal evolution vector and a fault vector template in a preset fault feature matrix, and the preset fault feature matrix is constructed according to fault types; When the similarity is higher than a set identification threshold and the matching result is stable and lasts for n sampling periods, it is determined that the voltage regulating module has a fault, and the identified fault type is output.

6. The fault diagnosis method for a voltage regulating module as described in claim 5, characterized in that, The abnormal trend monitoring window records the parameter offset, the change rate and the abnormal duration by using a sliding time window.

7. A system for diagnosing a fault of a voltage regulating module for implementing the method for diagnosing a fault of a voltage regulating module according to any one of claims 1 to 6, characterized in that, Including: A parameter acquisition module, a preliminary detection module, a fault diagnosis module and a label generation module; The parameter acquisition module is configured to periodically acquire key operation parameters of the voltage regulating module; The preliminary detection module is configured to perform preliminary abnormal detection on the acquired key operation parameters by using a multi-dimensional dynamic threshold model; The fault diagnosis module is configured to, when the preliminary abnormality is determined, perform abnormality trend analysis in a time window, and if the abnormality lasts for more than a preset time threshold, compare the current abnormality state with a preset fault feature matrix to identify whether the voltage regulating module has a fault. The label generation module is configured to generate a corresponding fault label according to the identified fault type.

8. The fault diagnostic system of a voltage regulating module according to claim 7, wherein The fault diagnosis module comprises a trend monitoring unit, a similarity matching unit and a fault diagnosis unit. The trend monitoring unit is configured to, after the preliminary abnormality is determined, perform abnormality trend monitoring by using an abnormality trend monitoring window. The similarity matching unit is configured to perform similarity matching between the parameter abnormality evolution vector and a fault vector template in the preset fault feature matrix. The fault diagnosis unit is configured to, when the similarity is higher than a set identification threshold and the matching result is stable and lasts for n sampling periods, determine that the voltage regulating module has a fault, and output the identified fault type.

Images