Fault Chain-Based Fault Location and Early Warning Method for Equipment Structure
By constructing a fault chain function, real-time fault location and early warning of equipment structure are realized based on strain monitoring data. This solves the problems of insufficient real-time performance and interpretability in existing technologies, improves fault identification accuracy and predictive response timeliness, and is suitable for edge computing platforms.
Patent Information
- Application Number
- CN202511209613.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-27
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2045-08-27
AI Technical Summary
Existing technologies have not yet achieved an effective balance between fault identification accuracy, real-time processing capabilities, and system engineering feasibility. In particular, they have significant shortcomings in real-time location and early warning capabilities in the early stages of fault development, making it difficult to meet the requirements of rapid response, strong interpretability, and strong embedded deployment capabilities in engineering projects.
By acquiring strain monitoring data of the equipment structure, dividing it into sub-regions, setting strain monitoring points, constructing a fault chain function, establishing a mapping relationship between strain data and crack length, and combining it with real-time strain data for inversion processing, fault location and early warning can be achieved.
It enables real-time location and early warning of equipment structural faults, reduces the algorithm's dependence on computing power, enhances the model's interpretability and adaptability to edge computing platforms, has proactive early warning capabilities, and improves the response efficiency and safety protection capabilities of the structural health management system.
Smart Images

Figure CN121118521B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of fault chains, and particularly relates to a method for fault location and early warning of equipment structures based on fault chains. Background Technology
[0002] With the widespread application of various high-performance equipment in complex service environments such as aerospace, wind power, new energy, and marine engineering, the safety and stability of their structures are becoming increasingly prominent. These equipment structures typically possess characteristics such as lightweight, high strength, and complex curved surfaces. During long-term operation, they are inevitably subjected to the continuous effects of various coupled loads, such as thermal loads, vibration loads, aerodynamic impacts, and mechanical stresses. This makes them highly susceptible to microcracks or material fatigue damage in areas of localized stress concentration, ultimately leading to irreversible structural failure. Traditional structural health assurance methods mostly rely on manual inspections, periodic offline testing, or alarm mechanisms based on static thresholds. These methods cannot achieve dynamic identification and response to the entire process of damage from its inception to failure, often exhibiting fatal shortcomings such as delayed response and inability to predict failure risks. While some structural anomaly identification methods based on image coding and deep learning models have emerged in recent years—converting monitoring data into images and training convolutional neural network models for anomaly detection—these methods suffer from poor real-time performance, high computational requirements, and physical uninterpretability, making it difficult to meet the engineering requirements for rapid response, strong interpretability, and strong embedded deployment capabilities.
[0003] Another type of structural health assessment technology that uses embedded CAE models has a strong physical basis and predictive ability, but its simulation calculation is large and its real-time performance is insufficient, especially in edge computing scenarios where it cannot meet the needs of online applications.
[0004] In summary, current mainstream structural health monitoring methods have not yet achieved an effective balance between fault identification accuracy, real-time processing capability, and system engineering feasibility. In particular, there are still significant shortcomings in real-time location and early warning capabilities in the early stages of fault development. There is an urgent need to propose a structural fault monitoring and early warning method that can be deployed in the field operating environment, has high real-time performance and engineering practicality, and is physically interpretable. Summary of the Invention
[0005] The purpose of this invention is to propose a fault location and early warning method for equipment structures based on fault chains, which solves the prominent problems of existing technologies in terms of real-time performance, deployability, and physical interpretability.
[0006] To achieve the above objectives, this invention provides a method for fault location and early warning of equipment structures based on fault chains, comprising the following steps: Obtain the equipment structure to be monitored, and divide the equipment structure to be monitored into N non-overlapping sub-regions; Multiple strain monitoring points are set up in each sub-region to capture local load response characteristics as strain measurements for each strain monitoring point. Based on the strain measurement at each strain monitoring point, strain response datasets for different crack lengths are obtained through experiments or simulations. Strain monitoring points in each sub-region that exhibit monotonic characteristics with crack length in the strain response dataset are selected as valid monitoring points. Based on the effective monitoring points, a fault chain function that integrates multi-point responses is constructed to establish a mapping relationship between strain data and crack length. For the equipment structure under real-time monitoring, the real-time strain data is inverted to generate a crack length sequence for each sub-region. By combining the crack length sequence of each sub-region, the fault location of the real-time monitored equipment structure is performed, and a fault warning is triggered based on a preset static threshold and dynamic trend prediction.
[0007] Furthermore, the number of strain monitoring points is preferably 3 ≤ M ≤ 8, where M is the number.
[0008] Furthermore, the process of obtaining strain response datasets for different crack lengths based on strain measurements at each strain monitoring point through experiments or simulations specifically involves: For each sub-region, tensile crack propagation tests were conducted on a series of crack lengths at the strain monitoring points to collect strain response datasets; the data acquisition was performed using high-precision resistance strain gauges or fiber Bragg grating sensors. or For a series of crack lengths, extended finite element method is used to model the sub-region, and after controlling the crack to extend to a set series of crack lengths, the strain values corresponding to the strain monitoring points are extracted to obtain a strain response dataset of numerical simulation version.
[0009] Furthermore, after obtaining the strain response dataset for different crack lengths through experiments or simulations, the strain response dataset is input into a strain response model with nonlinear compensation and noise regularization characteristics to fit the response path of each monitoring point during crack propagation, specifically as follows: ; in, Indicates the region The Middle Each monitoring point at crack length The strain monitoring data below, It is the crack sensitivity coefficient at that point. This represents the nonlinear growth exponent of the crack response within this region. This is the overall nonlinear compensation factor for the region. It is the geometric nonlinear amplification factor corresponding to the point location. This indicates a correction term caused by low-frequency disturbances such as background strain or temperature drift. According to the strain response Generate strain response dataset .
[0010] Furthermore, the step of selecting strain monitoring points in each sub-region whose strain response dataset exhibits monotonic characteristics with crack length as valid monitoring points specifically involves: At each strain monitoring point, a crack strain increment sequence is constructed based on the strain response dataset, and a trend score for each strain monitoring point is generated by combining a structural similarity regularization term with a trend evaluation function. When the trend score exceeds a certain empirical threshold, the current strain monitoring point is considered to have the potential to construct a stable crack length function, and its strain response data will be retained for subsequent fitting processes; otherwise, it is determined to be a trend unstable point, and the operation is suspended.
[0011] Furthermore, the trend evaluation function is expressed as: ; in, Rate the trend. It is a location-related weighting factor used to enhance the sensitivity to data on the mid-to-late stage of crack development. X is the total number of crack states, and x is the current crack. For the first Crack strain increment sequence at each strain monitoring point This is a structural similarity regularization term, representing the deviation of the strain trend of the current monitoring point from that of other points in the same area, used to penalize local anomalous behavior. This is the regularization intensity coefficient; Among them, structural similarity regularization term The calculation is as follows: ; in, Indicates and Different indexes of monitoring points in the same area This is the strain increment sequence for the corresponding point; For the first The crack strain increment sequence of strain monitoring points; M is the total number of strain monitoring points.
[0012] Furthermore, the fault chain function is expressed as: ; in, This is a sequence of crack length data; k is the number of cracks. There are 1 strain monitoring points; K is the total number of strain monitoring points; First item This is a power-law growth term used to capture the slow, nonlinear evolution stage in the early stages of crack propagation; Second item To suppress the decay term, a model is developed for the slowing behavior after the rapid rise in strain response in the later stage; The weight of strain monitoring point k reflects the contribution of different points to the overall crack assessment, and is set by combining historical fault data and experience in the distribution of point locations. The regional offset compensation amount takes into account the system error caused by boundary effects and uneven loading.
[0013] Furthermore, for the equipment structure under real-time monitoring, the real-time strain data is inverted to generate a crack length sequence for each sub-region, specifically as follows: At each time step Select window length as Historical data sequence By combining the time-series regularization adjustment term, a weighted fusion expression is constructed. The time-series regularization term is used to suppress outliers where strain abrupt changes do not conform to physical laws. Based on the weighted fusion expression Reassembled into region input vector Substitute the values into the fault chain function to calculate the crack length at the current moment. ; Among them, the time-series regularization adjustment term , is represented as: ; in, To adjust the strength coefficient, This represents the second derivative of the strain at that point over time, used to capture high-frequency abrupt changes in signals; It is a threshold activation function, which is applied only when the mutation strength exceeds the physically permissible range. This item will only be activated and participate in the correction at that time.
[0014] Furthermore, the step of combining the crack length sequence of each sub-region to locate the fault in the real-time monitored equipment structure, and triggering a fault warning based on a preset static threshold and dynamic trend prediction, specifically involves: Based on the crack length sequence and combined with the regional weight parameters, fault location is performed, and the fault region number is obtained by extreme value judgment. Design a continuous consistency verification mechanism to verify the currently initially identified fault area number and generate fault location results; Based on the fault location results and the real-time monitored equipment structural strain data, a fault early warning is issued.
[0015] Furthermore, the continuous consistency verification mechanism is specifically as follows: Only when the same area has been consecutive in the past Only when the number of times a region is repeatedly identified as a fault area in a given time period reaches a certain threshold is the current region considered to be the final fault location result. The fault early warning design incorporates a dual-threshold early warning mechanism: One is the static strain threshold, which is directly based on whether the current crack length exceeds the limit; The second is the dynamic trend threshold, which uses a multinomial trend prediction model to assess whether the critical value will be reached within a foreseeable time window in the future.
[0016] The beneficial technical effects of the present invention are at least as follows: To address the prominent issues of existing technologies in terms of real-time performance, deployability, and physical interpretability, this invention proposes a fault identification technology approach that integrates engineering practicality with mechanical mechanisms. Based on the stress-strain response law during structural damage evolution, this method establishes a direct link between monitoring data and structural degradation state by constructing a function mapping relationship that expresses the crack propagation trend, thereby enabling the identification of structural fault locations and the prediction of their development trends. Compared to traditional black-box models or purely numerical simulation methods, this invention is driven by actual monitoring data, supplemented by physically based degradation evolution process modeling. This significantly reduces the algorithm's dependence on computing power and enhances the model's interpretability and adaptability to edge computing platforms. Simultaneously, while ensuring detection accuracy, this method can output the current damage level and development trend of the structure in real time, possessing proactive early warning capabilities. It can issue risk warnings before significant damage occurs to equipment, greatly improving the response efficiency and safety protection capabilities of the structural health management system. Therefore, this invention achieves an effective balance between fault location accuracy, predictive response timeliness, and engineering application feasibility, possessing significant technical promotion value and practical engineering significance. Attached Figure Description
[0017] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of the steps of the equipment structure fault location and early warning method based on fault chain of the present invention. Detailed Implementation
[0019] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.
[0020] In one or more embodiments, such as Figure 1 As shown, a method for fault location and early warning of equipment structure based on fault chains is disclosed. The method includes the following steps: S1. Obtain the equipment structure to be monitored, and divide the equipment structure to be monitored into N non-overlapping sub-regions.
[0021] Specifically, the equipment structure to be monitored Divided into Each region is denoted as [number]. ,in .
[0022] To achieve spatial location and area attribution identification of faults, it is first necessary to identify the structure of the equipment to be monitored. Divided into several independent structural sub-regions Each sub-region has a clearly defined boundary and can be individually identified and tracked during monitoring and analysis. This division is based on engineering drawings or 3D modeling data of the equipment structure, and comprehensively considers the following factors: structural geometry, material distribution characteristics, stress concentration areas, common crack locations, and the feasibility of sensor placement. The structural division should ensure that each sub-region has a certain degree of mechanical independence and, as far as possible, encompass potential locations of local damage to avoid cross-regional interference.
[0023] Number of regions The value can be flexibly set according to the structural complexity and monitoring resolution requirements, and a value between 3 and 10 is generally recommended. During the specific partitioning process, if there are multiple obvious high-stress concentration areas, it is advisable to prioritize treating them as independent monitoring units and partitioning them separately. For symmetrical structures, symmetrical partitioning can be implemented through geometric mapping to reduce computational redundancy. The partitioning results can be used to extract the boundary node information of each region from the structural finite element mesh or CAD model as input, which will then be used for sensor deployment and strain data attribution processing.
[0024] Once the regional boundaries are determined, the integrity constraint of the overall structure, which consists of these non-overlapping sub-regions, must be satisfied, i.e., the equipment structure. Should be all sub-regions The union of regions, where no two regions overlap: ; in, This refers to the overall structural unit of the equipment. The total number of regions after division. For the first Sub-regions This indicates that there is no overlap or intersection between any two subregions. This indicates a set union operation.
[0025] This partitioning strategy ensures that a unique spatial region can be clearly identified during fault location, eliminating the ambiguity caused by overlap. The structural partitioning process does not involve complex numerical calculations, making it friendly to on-site deployment and data structure processing. At the same time, it provides a basic spatial framework for subsequent steps such as strain response collection, fault chain construction, and trend prediction.
[0026] S2. Multiple strain monitoring points are set up in each sub-region to capture the local load response characteristics as the strain measurement value of each strain monitoring point.
[0027] Specifically, in each region set up There are 1 strain monitoring point, denoted as ,in ; To obtain the structural local response characteristics and support the subsequent fault identification process, it is necessary to perform analysis in each defined region. Several strain monitoring points are set up inside. Number of monitoring points The settings should be determined based on the local structural complexity, stress gradient variations, and sensor layout density requirements. A recommended setting range is [range to be specified]. This is to balance the monitoring coverage with the constraints of data collection resources.
[0028] The layout of monitoring points should follow these design principles: Priority should be given to locations with a high incidence of historical cracks, stress concentration boundaries, or near connection nodes within the area; the arrangement of points should ideally cover the principal stress directions and potential crack propagation paths; and the physical feasibility of sensor placement should be considered, including surface fit conditions, wiring channels, and fixed locations. Point placement can refer to the regional grid boundary information, using the boundary centroid method, equidistant projection method, or in-plane distribution optimization algorithm to generate initial monitoring point positions, ensuring a uniform overall layout and mechanical representativeness.
[0029] Each monitoring point A strain sensor is installed for each monitoring point. Common types include resistance strain gauges and fiber Bragg gratings (FBGs). The acquired data forms a strain response time series, which serves as the raw input data for the structural state. The specific spatial coordinates of each monitoring point can be directly extracted from the structural 3D model and CAD annotations. The data format is as follows: The triplet can be directly associated with the simulation mesh number or the actual installation number.
[0030] This step of the deployment process ensures that each structural area... Each device possesses independent local strain sensing capabilities, laying the spatial sampling foundation for subsequently establishing the "strain-crack" mapping relationship. Its overall layout should satisfy the following spatial definition relationships: ; in, Indicates the first The first region The coordinates of each strain monitoring point For the structure of the first Divide into regions, For the total number of regions, The number of monitoring points deployed in each region.
[0031] This lattice-based deployment strategy can effectively capture the disturbance characteristics of local cracks on the strain field, while improving the overall system's perception resolution of crack propagation direction and velocity changes, demonstrating clear engineering feasibility and physical adaptability.
[0032] S3. Based on the strain measurement value of each strain monitoring point, obtain the strain response dataset under different crack lengths through experiments or simulations.
[0033] Specifically, tensile tests or numerical simulations are used to obtain data on equipment structures under different crack lengths. ( Strain monitoring data under) ; Having completed the structural region division and strain monitoring point deployment, this step is used to construct a mapping dataset between regional crack propagation states and local strain responses, which forms the basis of the entire failure chain formation mechanism. Each sub-region... Monitoring points in It is necessary to consider a series of crack lengths Under the given conditions, obtain the corresponding static strain response value. This is to support subsequent function relationship fitting.
[0034] Data acquisition can be achieved through two technical approaches. One is to conduct crack propagation tensile tests. On components with materials, structural forms, and installation environments consistent with actual application scenarios, a crack initiation notch design combined with displacement-controlled loading is used to induce crack propagation along a predetermined path. At each crack length stage... Next, pause loading and collect all points. The strain values were obtained. Strain acquisition was performed using high-precision resistance strain gauges or fiber Bragg grating sensors, and the sampling frequency, time window, and channel synchronization accuracy all met the experimental accuracy requirements.
[0035] Another approach is to use simulation techniques with crack simulation capabilities, such as Extended Finite Element Method (XFEM), to simulate cracks in the region. Modeling is performed, and crack propagation is controlled to a set length. Then, the strain values corresponding to the nodes are extracted to obtain the numerical simulation version. Data. The two paths mentioned above can be used in combination to enhance validation and increase data diversity.
[0036] To express the multi-mechanism strain response characteristics of a specific region of a structure, this invention proposes a strain response model with nonlinear compensation and noise regularization characteristics during the data generation stage, used to fit the response path of each monitoring point during crack propagation. The model is as follows: ; in, Indicates the region The Middle Each monitoring point at crack length Strain response under the following conditions It is the crack sensitivity coefficient at that point. This represents the nonlinear growth exponent of the crack response within this region. This is the overall nonlinear compensation factor for the region. It is the geometric nonlinear amplification factor corresponding to the point location. This indicates a correction term caused by low-frequency disturbances such as background strain or temperature drift.
[0037] This model exhibits significant engineering innovations in structural health monitoring scenarios: First, it introduces a power-function-based master response path, suitable for describing the regular evolution trend of stable crack propagation stages; second, it incorporates a logarithmic enhancement term, accurately expressing the local response amplification effect caused by stress gradient concentration regions or material heterogeneity; and third, it designs an empirical compensation term. It is used to shield the baseline offset caused by factors such as environmental temperature drift, load coupling or test system errors, and improves the model's adaptability and anti-interference ability in field data.
[0038] To ensure consistency in the magnitude and physical meaning of the terms in the model, crack length Before modeling, scale normalization is required. This process can be embedded in the data acquisition or simulation preprocessing stage and executed uniformly without affecting the function structure of the main model and subsequent fitting.
[0039] The final strain response data matrix It will serve as the direct input for constructing the fault chain function relationship in subsequent steps, and its accuracy and interpretability determine the stability and predictive ability of the entire early warning system.
[0040] S4. Select strain monitoring points in each sub-region whose strain response dataset exhibits monotonic characteristics with crack length as valid monitoring points.
[0041] Specifically, determine each region Internal strain monitoring data With crack length Monotonicity; After obtaining the structural regions Internal monitoring point In the state of multiple crack propagation Strain response data The core task of this step is to identify valid data points that meet the conditions for constructing the fault chain. This process serves as a screening step for the previous "data generation" step and is also a prerequisite for the next step, "function fitting." Its technical significance lies in improving the stability, physical rationality, and reversibility of the subsequent mapping model through trend constraints, thereby laying a mathematical and engineering foundation for the online inversion and early warning of crack states.
[0042] Considering that actual equipment structures often encounter complex operating conditions such as temperature drift, loading disturbances, and sensor adhesion errors in service environments, simply using absolute monotonicity judgment methods (such as increasing or decreasing throughout the process) can easily lead to excessive rejection of valid points or misidentification of noise responses. This invention proposes a trend screening method with a "nonlinear weighting function" and "structural similarity regularization" tailored to actual monitoring scenarios, ensuring tolerance for local disturbances while retaining key trend characteristics.
[0043] At each monitoring point Above, construct its crack strain increment sequence. Then, the following trend evaluation function is introduced: ; In this expression, the first term is the main trend integral term. The first term is a location-related weighting factor used to enhance the sensitivity to data from the mid-to-late stages of crack development; the second term is an innovative structural similarity regularization term. It indicates the degree of deviation of the current monitoring point from the strain trend of other points in the same area, and is used to punish local abnormal behavior; The regularization strength coefficient is adjustable. This design differs from traditional single-point trend analysis methods, enabling robust trend identification driven by group behavior under complex structural conditions.
[0044] Structural similarity regularization term The definition is as follows: ; in, Indicates and Different indexes of monitoring points in the same area This represents the strain increment sequence at the corresponding point. Essentially, this assesses whether the strain growth trend at a monitoring point is consistent with that of other points in the same region, thereby enhancing the ability to suppress point deviations and reflecting the data consistency judgment concept within the context of structural coupling.
[0045] When trend score Exceeding a certain empirical threshold (Suggested value range is) When the parameters (which can be set according to regional characteristics) are considered to indicate that the monitoring point has the potential to construct a stable crack length function, its strain data... It will be retained for subsequent fitting process; otherwise, it will be judged as an unstable point and will not be included in function modeling for the time being.
[0046] This step not only achieves the extraction of directional features from strain data, but also innovatively introduces inter-point cooperative behavior within the structural region as a regularization control, significantly improving the screening robustness under multi-source disturbances. This design aligns with the core objective of this invention, "construction of multi-point fault chains," ensuring the physical foundation and numerical stability of the subsequent model training.
[0047] S5. Based on the effective monitoring points, construct a fault chain function that integrates multi-point responses and establish a mapping relationship between strain data and crack length.
[0048] Specifically, strain monitoring points in the steps where the data exhibits monotonic characteristics with crack length are selected, and a functional relationship between strain data and crack length is established to form a fault chain; After completing the structural regions After screening the monotonicity of internal strain monitoring points, this step focuses on constructing a mathematical mapping function that expresses the relationship between "damage degree and strain response" in this region, namely the fault chain function. This function is the core component of the early warning method of this invention, determining whether real-time strain data can be effectively converted into crack length estimation, thereby supporting fault location and trend prediction.
[0049] To improve the accuracy and engineering stability of the function expression, this invention does not employ traditional univariate interpolation or low-order polynomial fitting methods. Instead, it designs a multi-point mapping model with a trend adjustment term and a monitoring point weight fusion mechanism. Let the region... The CCP For each valid monitoring point that passes the monotonicity test, its strain response vector is constructed. Each item corresponds to the crack length. The monitored strain value.
[0050] The fault chain function expression designed in this invention is as follows: ; The function consists of two parts: the first term The first term is a power-law increasing term, used to capture the slow, nonlinear evolution stage in the early stages of crack propagation; the second term... To suppress the decay term, a model is developed for the slowing behavior after the rapid rise in strain response in the later stage; The weights of the monitoring points reflect the contribution of different points to the overall crack assessment, and are set based on historical fault data and experience in the distribution of point locations. The regional offset compensation amount takes into account the system error caused by boundary effects and uneven loading.
[0051] The function structure demonstrates innovation in three aspects: First, it adopts a weighted fusion of multi-point responses, avoiding the problem of single-point functions being susceptible to noise; second, it simultaneously models the behavior of crack propagation in the early-to-late stages, which conforms to the physical evolution trend of actual structures during service; and third, it introduces a suppression mechanism through an exponential decay term, avoiding oscillations in the predicted value in the high-response region and improving the numerical stability of the function.
[0052] To avoid overfitting during the parameter fitting process, the parameters in the function... The initial values can be estimated from simulation data. The fitting adopts a residual sum of squares minimization strategy, and a regularization term can be introduced to control the parameter amplitude. The entire fitting process does not rely on neural networks or deep learning models, ensuring the physical interpretability and engineering deployability of the model, which is especially suitable for low-power, high-response application scenarios in edge computing devices.
[0053] The final fault chain function This constitutes a regional monitoring and response mechanism. To the crack propagation state The mapping path will be used in subsequent steps for online inversion and dynamic prediction to achieve proactive risk identification and zonal health assessment of the structure.
[0054] S6. For the equipment structure under real-time monitoring, the real-time strain data of the real-time strain data is inverted to generate a crack length sequence for each sub-region.
[0055] Specifically, this refers to the real-time monitoring of equipment structural strain data sequences. Calculate each region using fault chains Crack length data sequence ; After completing the fault chain function After the construction is completed, this step applies it to the online operation scenario during the service of the structure to realize the dynamic inversion of real-time monitoring data to the structural crack state. Unlike step (5), which mainly focuses on fitting static historical samples, this step emphasizes the time-by-time evolution tracking based on online strain input. It is the key interface for the entire invention to transform from modeling to practical engineering deployment, and determines the responsiveness and adaptability of the method in edge computing terminals or on-site embedded systems.
[0056] In actual service, deployed in the region Strain monitoring points It will generate time-series data streams at a fixed sampling frequency. Since actual monitoring signals are often accompanied by various types of signal contamination, such as low-frequency drift, high-frequency noise, and load disturbances, directly substituting the raw data into the fault chain function will lead to jumps or misjudgments in the prediction results. Therefore, this invention proposes an innovative fusion mechanism that balances response stability and abrupt change sensitivity, performing structural filtering and physical consistency constraint processing on the input strain sequence.
[0057] Specifically, at each time step Select window length as Historical data sequence Construct its weighted fusion expression : ; in, This is a historical weighted coefficient sequence, which can be designed based on exponential decay or relative risk factors to meet the following requirements. This ensures that energy is conserved. The time-series regularization term, introduced for innovation, is used to suppress outliers where strain abrupt changes do not conform to physical laws. This term does not directly smooth the data; instead, it intervenes in the historical window weighting through trend deviation to improve the ability to identify abrupt changes rather than simply filtering them out.
[0058] The expression is as follows: ; in, To adjust the strength coefficient, This represents the second derivative of the strain at that point over time, used to capture high-frequency abrupt changes in signals; It is a threshold activation function, which is applied only when the mutation strength exceeds the physically permissible range. This feature is only activated and participates in the correction process at that time. This effectively avoids false anomalies caused by equipment shocks, communication interruptions, or human interference.
[0059] After completing the fusion strain sequence After construction, it is reassembled into a region input vector. Substitute into the fault chain function Calculate the crack length at the current moment: ; The inversion operation is performed once per monitoring cycle, and the cycle can be set according to the actual system sampling rate (e.g., 10Hz, 50Hz), providing high controllability. In terms of hardware implementation, the entire process relies only on finite sliding windows, vector weighting, and function substitution, resulting in significantly lower resource consumption compared to traditional real-time simulation or neural network prediction modes, making it suitable for deployment requirements of low-power edge devices and embedded platforms.
[0060] This step essentially transforms the raw monitoring data into stable and reliable model input through physical enhancement and data preprocessing mechanisms, demonstrating the fusion capability of the invention's dual objectives of "lightweight modeling + robust inversion." Through the combined design of time windows, trend regularization, and mutation activation mechanisms, not only is stable system operation ensured, but the system also possesses real-time adaptability to cope with complex dynamic working conditions on-site, providing high-quality input for subsequent fault location and early warning.
[0061] S7. Combining the crack length sequence of each sub-region, fault location is performed on the real-time monitored equipment structure, and a fault warning is triggered based on a preset static threshold and dynamic trend prediction.
[0062] Specifically, fault location is performed based on real-time monitored equipment structural strain data.
[0063] This step is based on the crack length sequence obtained from the inversion in the previous step. By combining the structural region division results, spatial location of the fault can be achieved. Specifically, the goal is to determine the region most likely to experience crack propagation or structural damage at the current moment. This provides spatial boundaries for early warning and maintenance decisions. As one of the core outputs of this invention, the positioning process not only affects the trigger range of subsequent early warning areas, but also determines the reliability and response efficiency of the system on the engineering site.
[0064] Because different regions have varying physical properties, historical fault risks, and sensor signal-to-noise ratios, relying solely on the absolute value of crack length for "maximum value determination" can lead to sensitivity of the location results to single-point anomaly responses, especially when the structure experiences large-area load disturbances or local signal fluctuations, easily resulting in misjudgments. Therefore, this invention designs a fault location index based on the fusion of "regional crack length + weighting coefficient," and obtains the suspected fault area at the current moment through simple extreme value judgment: ; in, This indicates the initial fault area number. For the first Each region at time The estimated value of the crack, This is a region weight parameter used to reflect the physical importance of the region, historical failure probability, or sensor stability. This approach is more robust to disturbances than traditional methods and also incorporates prior structural knowledge.
[0065] However, in real-world environments, transient outliers can lead to misjudgments. Therefore, this invention further proposes a continuous consistency verification mechanism: only when the same region has been continuously... The number of times a region is repeatedly identified as a fault area at a given time point reaches a certain threshold. Only after this process is completed is the region considered the final fault location result. This judgment is achieved through a simple sliding window voting mechanism, which does not increase the system's computational burden but significantly improves timing stability and anti-interference capability.
[0066] This positioning strategy allows the system to not only output the current risk concentration area of the structure in real time, but also to naturally shield against typical field problems such as short-term noise and sensor offset, meeting the robustness requirements for long-term equipment operation. The final positioning result will serve as the input condition for the next fault early warning module, realizing an alarm strategy under spatiotemporal constraints.
[0067] Then, based on the real-time monitored equipment structural strain data, fault warnings are issued. After completing the structural regions Crack length After real-time inversion and spatial positioning, this step aims to make a forward-looking judgment on the structural health status, that is, to trigger fault warning signals through trend prediction and risk threshold management before failure is reached, so as to realize the early response and maintenance preparation of the equipment system.
[0068] Fault early warning differs from traditional alarm triggering; its core lies in identifying the time point when a "high-risk state" is about to be entered based on the structural evolution trend. Therefore, this invention designs a dual-threshold early warning mechanism: the first is a static strain threshold, which is directly based on the current crack length exceeding the limit; the second is a dynamic trend threshold, which uses a polynomial trend prediction model to assess whether a critical value will be reached within a foreseeable time window in the future.
[0069] In each region Internally, set alarm thresholds. and failure threshold When the crack length at the current moment satisfies When this happens, the area is identified as a potential risk area and enters a warning state. If it is necessary to predict in advance whether the area will be at risk in the future... Within a time step Then, within the time window, a crack evolution trend curve is fitted, and the predicted arrival time is calculated: ; in, Indicates based on history The fitted trend prediction function (lightweight models such as second-order polynomials and exponential functions can be used). This represents the critical failure crack length of the structural region. If this condition is met, the system will issue a "predictive risk warning" signal at the current moment, providing response time for engineering decisions such as maintenance, load reduction, and isolation.
[0070] To avoid allergy warnings, this invention introduces a time-stationary strategy: requiring at least An actual early warning output can only be triggered if all of the above threshold conditions are met in each consecutive time step. This strategy can effectively filter out short-term false alarms caused by environmental fluctuations, noise interference, or instability of individual sensors, thereby improving the actual reliability of the system's early warning.
[0071] This step, as the output of the entire invention process, has the key value of realizing the leap from "diagnosis" to "prediction" by combining the current structural state with future evolution trends. This enables the structural health monitoring system to no longer respond passively, but to have the ability to actively prevent and control, significantly improving the operational safety and engineering reliability of equipment in complex environments.
[0072] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.
[0073] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0074] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
[0075] The various illustrative logic blocks, modules, and circuits described in conjunction with the embodiments disclosed herein can be implemented or performed using a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general-purpose processor may be a microprocessor, but in alternatives, it may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors cooperating with a DSP core, or any other such configuration.
[0076] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of both. The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor such that the processor can read and write information to / from the storage medium. In an alternative, the storage medium may be integrated into the processor. The processor and storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and storage medium may reside as discrete components in the user terminal.
[0077] In one or more exemplary embodiments, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functionality may be stored or transmitted as one or more instructions or code on or through a computer-readable medium. A computer-readable medium includes both computer storage media and communication media, encompassing any medium that facilitates the transfer of a computer program from one location to another. A storage medium may be any available medium accessible to a computer. By way of example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage, disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and is accessible to a computer. Any connection is also legitimately referred to as a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of a medium. As used in this article, disk and disc include compact discs (CDs), laser discs, optical discs, digital multi-purpose discs (DVDs), floppy disks, and Blu-ray discs. Disks typically reproduce data magnetically, while discs reproduce data optically using lasers. Combinations of these should also be included within the scope of computer-readable media.
[0078] The prior description of this disclosure is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not intended to be limited to the examples and designs described herein, but should be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for fault location and early warning of equipment structure based on fault chains, characterized in that, Includes the following steps: Obtain the equipment structure to be monitored, and divide the equipment structure to be monitored into N non-overlapping sub-regions; Multiple strain monitoring points are set up in each sub-region to capture local load response characteristics as strain measurements for each strain monitoring point. Based on the strain measurement at each strain monitoring point, strain response datasets for different crack lengths are obtained through experiments or simulations. Strain monitoring points in each sub-region that exhibit monotonic characteristics with crack length in the strain response dataset are selected as valid monitoring points. Based on the effective monitoring points, a fault chain function that integrates multi-point responses is constructed to establish a mapping relationship between strain data and crack length. For the equipment structure under real-time monitoring, the real-time strain data is inverted to generate crack length sequences for each sub-region. By combining the crack length sequence of each sub-region, the fault location of the real-time monitored equipment structure is performed, and a fault warning is triggered based on a preset static threshold and dynamic trend prediction.
2. The equipment structure fault location and early warning method based on fault chain according to claim 1, characterized in that, The number of strain monitoring points is 3 ≤ M ≤ 8, where M is the number.
3. The equipment structure fault location and early warning method based on fault chain according to claim 1, characterized in that, The strain response dataset for different crack lengths is obtained through experiments or simulations based on the strain measurements at each strain monitoring point. For each sub-region, tensile crack propagation tests were conducted on a series of crack lengths at the strain monitoring points to collect strain response datasets; the data acquisition was performed using high-precision resistance strain gauges or fiber Bragg grating sensors. or For a series of crack lengths, extended finite element method is used to model the sub-region, and after controlling the crack to extend to a set series of crack lengths, the strain values corresponding to the strain monitoring points are extracted to obtain a strain response dataset of numerical simulation version.
4. The equipment structure fault location and early warning method based on fault chain according to claim 3, characterized in that, After obtaining strain response datasets for different crack lengths through experiments or simulations, the strain response datasets are then input into a strain response model with nonlinear compensation and noise regularization characteristics to fit the response path of each monitoring point during crack propagation, specifically as follows: ; in, Indicates the region The Middle Each monitoring point at crack length The strain monitoring data below, It is the crack sensitivity coefficient at that point. This represents the nonlinear growth exponent of the crack response within this region. This is the overall nonlinear compensation factor for the region. It is the geometric nonlinear amplification factor corresponding to the point location. This indicates a correction term caused by low-frequency disturbances such as background strain or temperature drift. According to the strain response Generate strain response dataset .
5. The equipment structure fault location and early warning method based on fault chain according to claim 1, characterized in that, The process of selecting strain monitoring points in each sub-region that exhibit monotonic characteristics with crack length from the strain response dataset is specifically as follows: At each strain monitoring point, a crack strain increment sequence is constructed based on the strain response dataset, and a trend score for each strain monitoring point is generated by combining a structural similarity regularization term with a trend evaluation function. When the trend score exceeds a certain empirical threshold, the current strain monitoring point is considered to have the potential to construct a stable crack length function, and its strain response data will be retained for subsequent fitting processes; otherwise, it is determined to be a trend unstable point, and the operation is suspended.
6. The equipment structure fault location and early warning method based on fault chain according to claim 5, characterized in that, The trend evaluation function is expressed as follows: ; in, Rate the trend. It is a location-related weighting factor used to enhance the sensitivity to data on the mid-to-late stage of crack development. X is the total number of crack states, and x is the current crack. For the first Crack strain increment sequence at each strain monitoring point This is a structural similarity regularization term, representing the deviation of the strain trend of the current monitoring point from that of other points in the same area, used to penalize local anomalous behavior. This is the regularization intensity coefficient; Among them, structural similarity regularization term The calculation is as follows: ; in, Indicates and Different indexes of monitoring points in the same area This is the strain increment sequence for the corresponding point; For the first The crack strain increment sequence of strain monitoring points; M is the total number of strain monitoring points.
7. The equipment structure fault location and early warning method based on fault chain according to claim 1, characterized in that, The fault chain function is represented as follows: ; in, This is a sequence of crack length data; k is the number of cracks. There are 1 strain monitoring points; K is the total number of strain monitoring points; First item This is a power-law growth term used to capture the slow, nonlinear evolution stage in the early stages of crack propagation; Second item To suppress the decay term, a model is developed for the slowing behavior after the rapid rise in strain response in the later stage; The weight of strain monitoring point k reflects the contribution of different points to the overall crack assessment, and is set by combining historical fault data and experience in the distribution of point locations. The regional offset compensation amount takes into account the system error caused by boundary effects and uneven loading.
8. The equipment structure fault location and early warning method based on fault chain according to claim 7, characterized in that, For the equipment structure under real-time monitoring, the real-time strain data is inverted to generate a crack length sequence for each sub-region, specifically as follows: At each time step Select window length as Historical data sequence By combining the time-series regularization adjustment term, a weighted fusion expression is constructed. The time-series regularization term is used to suppress anomalous points where strain abrupt changes do not conform to physical laws; and at each time step... Select window length as Historical data sequence Construct its weighted fusion expression : ; in, For the historical weight coefficient sequence, satisfying This ensures that energy is conserved. This is a time-series regularization adjustment term used to suppress outliers where strain abrupt changes do not conform to physical laws; based on the weighted fusion expression... Reassembled into region input vector Substitute the values into the fault chain function to calculate the crack length at the current moment. ; Among them, the time-series regularization adjustment term , represented as: ; in, To adjust the strength coefficient, This represents the second derivative of the strain at the current strain monitoring point in time, used to capture high-frequency abrupt signals; It is a threshold activation function, which is applied only when the mutation strength exceeds the physically permissible range. Only when the timing regularization adjustment term is activated will it participate in the correction.
9. The equipment structure fault location and early warning method based on fault chain according to claim 1, characterized in that, The method involves combining the crack length sequences of each sub-region to locate faults in the real-time monitored equipment structure, and triggering fault warnings based on preset static thresholds and dynamic trend predictions. Specifically: Based on the crack length sequence and combined with the regional weight parameters, fault location is performed, and the fault area number is obtained by extreme value judgment. Design a continuous consistency verification mechanism to verify the currently initially identified fault area number and generate fault location results; Based on the fault location results and the real-time monitored equipment structure strain data, a fault early warning is issued.
10. The equipment structure fault location and early warning method based on fault chain according to claim 9, characterized in that, The continuous consistency verification mechanism is specifically as follows: Only when the same area has been consecutive in the past Only when the number of times a region is repeatedly identified as a fault area in a given time period reaches a certain threshold is the current region considered to be the final fault location result. The fault early warning design incorporates a dual-threshold early warning mechanism: One is the static strain threshold, which is directly based on whether the current crack length exceeds the limit; The second is the dynamic trend threshold, which uses a multinomial trend prediction model to assess whether the critical value will be reached within a foreseeable time window in the future.
Citation Information
Patent Citations
Crack propagation prediction method based on digital twinborn technology and data fusion
CN116384193A
Crack detection method, electronic equipment and computer readable storage medium
CN120176564A
