A Smart Compensation Method for Strain Gauge Bond Thickness Based on Finite Element Analysis

Abstract

This invention discloses an intelligent compensation method for strain gauge adhesive thickness based on finite element analysis, belonging to the field of intelligent compensation technology. The method includes: conducting a preliminary assessment of the health status of the strain gauge adhesive layer based on a thermomechanical microcirculation data package, generating a health assessment report; filtering a set of candidate anchor points using a temperature-sensitive drift peeling algorithm and removing abnormal points based on the thermomechanical microcirculation data package and the health assessment report to obtain a purified strain flow object; performing short-time-window finite element reference calculation based on the purified strain flow object, and generating a finite element reference response set through similarity matching; comparing the finite element reference response set with the purified strain flow object segment by segment, and obtaining an intelligent compensation strain result report for adhesive thickness through online consistency compensation verification. This invention, through the temperature-sensitive drift peeling algorithm and online consistency compensation verification, achieves a leap from static calibration to dynamic sensing in strain measurement, improving the accuracy and reliability of long-term monitoring.

Description

Technical Field

[0001] This invention relates to the field of intelligent compensation technology, and in particular to an intelligent compensation method for strain gauge bonding thickness based on finite element analysis. Background Technology

[0002] In recent years, resistance strain gauges have been widely used for surface strain measurement in mechanical components, aerospace structures, and civil engineering due to their high sensitivity and ease of deployment. Traditional strain measurement methods have evolved over decades from early static single-point measurements to dynamic distributed monitoring, primarily focusing on improving hardware accuracy, enhancing adaptability to measurement environments, and developing post-processing algorithms based on calibration data. With advancements in computer simulation, finite element analysis (FEM) has been introduced into strain measurement error analysis to study strain transfer mechanisms under ideal conditions. For example, by establishing a "strain gauge-adhesive layer-matrix" model, the influence of adhesive layer shear modulus and thickness on strain transfer efficiency can be analyzed. Currently, FEM is largely limited to offline, static sensitivity studies or parameter calibration, gradually leading to the formation of a calibration framework that is initially integrated with measurement data.

[0003] Existing strain measurement methods have limitations, primarily in two aspects: insufficient ability to sense time-varying parameters of the adhesive layer and difficulty in effectively decoupling temperature-mechanical-strain coupling errors. Traditional calibration methods are usually based on the initial calibration values ​​of the adhesive layer parameters, failing to track their dynamic changes under curing, aging, and temperature and humidity cycling in real time, leading to significant cumulative errors in long-term monitoring. Furthermore, existing methods struggle to separate the complex errors caused by temperature changes from a mechanistic perspective, are mostly limited to offline parameter studies, and fail to form a closed-loop linkage with real-time acquired strain, temperature, and load data. They also lack online adaptability and compensation verification mechanisms for complex field conditions, which restricts the reliability of strain measurement accuracy in long-term service environments. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a strain gauge bonding thickness intelligent compensation method based on finite element analysis to solve the problem of not being able to perceive the time-varying parameters of the adhesive layer in real time and to effectively decouple thermomechanical coupling errors.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for intelligent compensation of strain gauge adhesive thickness based on finite element analysis. The method includes: collecting and preprocessing raw data from measuring points to obtain a thermodynamic microcirculation data package; performing a preliminary assessment of the strain gauge adhesive layer health status based on the thermodynamic microcirculation data package and generating a health assessment report; filtering a set of candidate anchor points using a temperature-sensitive drift peeling algorithm based on the thermodynamic microcirculation data package and the health assessment report, and removing abnormal points to obtain a purified strain flow object; performing short-time-window finite element reference calculation based on the purified strain flow object, and generating a finite element reference response set through similarity matching; comparing the finite element reference response set with the purified strain flow object segment by segment, and obtaining an intelligent compensation strain result report for adhesive thickness through online consistency verification.

[0008] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the original data of the measuring point working condition includes timestamp, original strain, strain gauge adhesive layer temperature, working condition range code and mechanical load value.

[0009] The preprocessing includes noise filtering, standardization of physical quantity units, outlier removal, and decoupling of characteristic signals.

[0010] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the thermomechanical microcirculation data package includes short time window start and end timestamps, original strain sequence, adhesive layer temperature sequence, and same window mechanical load sequence.

[0011] As a preferred embodiment of the intelligent compensation method for strain gauge adhesive thickness based on finite element analysis described in this invention, the step of performing a preliminary assessment of the health status of the strain gauge adhesive layer based on a thermomechanical micro-circulation data package and generating a health assessment report includes the following specific steps.

[0012] Short-window monotonic decomposition is performed on the heat engine microcirculation data package to obtain the basic quantization table, and the directed area of ​​the loop is calculated by the phase loop geometry algorithm to generate the phase loop feature table.

[0013] Based on the phase loop feature table and the same window mechanical load sequence, the strain rate is obtained by the steady segment differential back-back method and the candidate anomaly point index is marked to form a rate back-back consistency table.

[0014] Perform sequence relation analysis on the phase loop characteristic table and the rate return consistency table to generate a health assessment report.

[0015] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the step of screening the candidate anchor point set using a temperature-sensitive drift peeling algorithm based on the thermodynamic microcirculation data package and health assessment report is as follows:

[0016] Extract candidate anomalies from the health assessment report, locate the corresponding short time window and monotonic segment in the heat engine microcirculation data package through index projection, and obtain the anomaly mask;

[0017] The anomaly mask and the original strain sequence are registered at the same amplitude and temperature. The local temperature-sensitive slope is calculated, and the initial temperature drift baseline is obtained by discrete line integration.

[0018] Based on the initial temperature drift baseline, a sliding window polynomial fitting is used for smooth correction, and a set of candidate deletion anchor points is generated by identifying residual jump points.

[0019] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the method of obtaining the purified strain flow object refers to fusing the anomaly mask and the set of candidate anchor points, performing temperature drift stripping and anomaly point removal, and obtaining the purified strain flow object through linear interpolation.

[0020] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the steps of performing short-time-window finite element reference calculation based on the purified strain flow object are as follows:

[0021] The load spectrum of the purified strain flow object is dynamically decomposed to generate a short time window load component set, and the candidate reference strain field is obtained through viscoelastic time-varying analysis.

[0022] The candidate reference strain field is subjected to differential operation and the gradient modulus field is calculated to identify the region that is sensitive to the change in adhesive thickness and obtain the adhesive thickness sensitive difference gradient field.

[0023] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the specific steps for generating the finite element reference response set are as follows:

[0024] The adhesive thickness sensitive difference is extracted from the gradient field of the adhesive thickness sensitive difference, and similarity matching is performed with the purification strain flow object to generate a labeled reference response;

[0025] All tagged reference responses are aggregated and smoothly stitched together in chronological order to generate a finite element reference response set.

[0026] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the step of comparing the finite element reference response set with the purified strain flow object segment by segment refers to aligning the finite element reference response set with the purified strain flow object in time sequence and comparing them segment by segment to generate a reference measured strain sequence pair.

[0027] As a preferred embodiment of the intelligent compensation method for strain gauge bonding thickness based on finite element analysis described in this invention, the specific steps for obtaining the intelligent compensation strain result report for bonding thickness are as follows:

[0028] The strain difference is calculated based on the reference measured strain sequence, and consistency is verified by combining the adhesive thickness sensitive difference gradient field to obtain the thickness sensitive difference segment.

[0029] Inverse linear interpolation is performed on the thickness-sensitive difference segment, and the equivalent bond thickness sequence is obtained by real-time calculation of the equivalent bond thickness.

[0030] Based on the equivalent bond thickness sequence, online compensation is performed on the purified strain flow object to generate a smart bond thickness compensation strain result report.

[0031] The beneficial effects of this invention are as follows: by using a temperature-sensitive drift stripping algorithm and online consistency compensation verification, a closed-loop intelligent compensation method is constructed, which accurately separates temperature drift from real mechanical strain and dynamically compares real-time data with the finite element model; it realizes the leap from static calibration to dynamic sensing in strain measurement, and improves the accuracy and reliability of long-term monitoring. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart of a smart compensation method for strain gauge bonding thickness based on finite element analysis.

[0034] Figure 2 A flowchart for filtering the set of anchor points to be deleted.

[0035] Figure 3 A flowchart for generating a reference response set.

[0036] Figure 4 A flowchart for obtaining a smart strain compensation result report for bond thickness. Detailed Implementation

[0037] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0038] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0039] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0040] Reference Figures 1-4 This is one embodiment of the present invention, which provides a smart compensation method for strain gauge bonding thickness based on finite element analysis, including the following steps:

[0041] S1: Collect raw data of the measuring point operating conditions and perform preprocessing to obtain the heat engine micro-circulation data package.

[0042] S1.1: The original data of the measuring point includes timestamp, original strain, strain gauge adhesive temperature, condition range code and mechanical load value;

[0043] Specifically, the timestamp is generated by the unified time reference at the acquisition end at the moment of sampling and attached to each sampling record;

[0044] The raw strain is converted from the voltage output of the strain measurement bridge into an uncorrected strain reading by the acquisition hardware and saved in timestamp order;

[0045] The temperature of the strain gauge adhesive layer is synchronously acquired by the temperature sensor channel located at the strain gauge adhesive layer position and recorded with time stamps;

[0046] The operating condition gear code is updated by the gear signal of the field controller or the gear input of the operation panel when the gear is switched, and a one-to-one correspondence is established with the timestamp;

[0047] Mechanical load values ​​are synchronously acquired and timestamped by mechanical sensor channels such as force, pressure, torque, or rotational speed associated with the measuring point.

[0048] S1.2: Preprocessing includes noise filtering, physical quantity unit standardization, outlier removal, and characteristic signal decoupling;

[0049] Specifically, the timestamps are first aligned and marked for integrity. After noise filtering, physical quantity unit standardization and outlier removal are completed, a continuous interval with unchanged working condition code and consistent strain gauge adhesive temperature change direction is taken as a short time window. The start timestamp of the short time window is taken from the first timestamp of the continuous interval, and the end timestamp of the short time window is taken from the last timestamp of the continuous interval, thus generating the start and end timestamps of the short time window.

[0050] After the start and end timestamps of the short time window are determined, all samples within the short time window are extracted in the order of timestamps from the original strain that has been aligned, denoised, standardized in units, and removed from outliers, to obtain the original strain sequence that contains only the effective points within the short time window.

[0051] After noise filtering, physical quantity unit standardization, and outlier removal, the adhesive layer temperature of the strain gauge is extracted point by point according to the short time window range and kept in one-to-one correspondence with the original strain sequence to obtain the adhesive layer temperature sequence.

[0052] After noise filtering, physical quantity unit standardization, and outlier removal, the mechanical load values ​​are extracted point by point according to the start and end timestamps of the short time window and arranged in the order of the timestamps to obtain the mechanical load sequence within the same window range.

[0053] The thermomechanical microcirculation data package includes short time window start and end timestamps, original strain sequence, adhesive layer temperature sequence, and same-window mechanical load sequence.

[0054] S2: Based on the thermomechanical microcirculation data package, perform a preliminary assessment of the health status of the strain gauge adhesive layer and generate a health assessment report.

[0055] S2.1: Perform short-time window monotonic decomposition on the heat engine microcirculation data package to obtain the basic quantization table, and calculate the directed area of ​​the loop through the phase loop geometry algorithm to generate the phase loop feature table;

[0056] It should be noted that the thermomechanical microcirculation data package is processed one by one according to short time windows. Within each short time window, the adhesive layer temperature sequence is linearly scanned along the time axis. The inflection point where the adhesive layer temperature sequence changes from upward to downward or from downward to upward is taken as the landing point. The time range between the previous landing point and the current landing point is used to establish the start and end timestamps of the monotonic segment. Based on the start and end timestamps of the monotonic segment, the original strain sequence, adhesive layer temperature sequence, and mechanical load sequence within the same short time window are synchronized with consistent timestamps to generate a basic quantization table.

[0057] Based on the start and end timestamps of each monotonic segment in the basic quantization table, the corresponding original strain sequence and adhesive temperature sequence are extracted segment by segment from the thermomechanical microcirculation data package. The points are connected in time order on the temperature-strain plane to form a closed trajectory, and the directional area of ​​the loop is calculated. This is then summarized together with the start and end timestamps of the monotonic segments to generate a phase loop feature table.

[0058] The expression for calculating the directed area of ​​a hysteresis loop is:

[0059] ;

[0060] in, The directional area of ​​the lapis latus. This represents the number of discrete points within the monotonic segment. The index value of the discrete point.

[0061] For the first The temperature of the adhesive layer at discrete points For the first Strain at discrete points.

[0062] is the area restoration coefficient, used to restore the doubled directed area generated by the cross product to the true hysteresis directed area.

[0063] Better yet, the directional area of ​​the calculated hysteresis loop can be used to quantify the strength and direction of the temperature-strain hysteresis loop, quickly identify the viscoelastic effect and the degree of energy hysteresis, and provide a stable indicator for subsequent consistency judgment and anomaly location.

[0064] S2.2: Based on the phase loop feature table and the mechanical load sequence in the same window, the strain rate is obtained by the steady segment differential back-back method and the candidate anomaly point index is marked to form a rate back-back consistency table;

[0065] It should be noted that, based on the phase loop feature table and the mechanical load sequence within the same window, corresponding segments are extracted from the mechanical load sequence and the original strain sequence within each short time window according to the start and end timestamps of the monotonic segments given in the phase loop feature table; within each monotonic segment, the stable segment with the smallest mechanical load fluctuation is selected, and the original strain sequence is subjected to adjacent point difference according to time sequence within the stable segment to obtain the strain rate; within the same short time window, based on the adjacent monotonic segments covered by the phase loop feature table, the upward segment and the downward segment are paired one by one to form a back-reversal comparison and the direction of the back-reversal difference is given; the time position where the strain rate is inconsistent with the direction of change of the mechanical load sequence within the same window and the directional area of ​​the loop in the phase loop feature table continuously increases or the sign is flipped is used as the candidate anomaly point index;

[0066] The start and end timestamps of the monotonic segment, the start and end timestamps of the stable segment, the strain rate, the description of the retracement direction, and the index of the candidate anomaly points are compiled and output as a rate retracement consistency table in short time window order.

[0067] S2.3: Perform sequence relationship analysis on the phase loop feature table and the rate return consistency table to generate a health assessment report.

[0068] It should be noted that, according to the short time window, the start and end timestamps of the monotonic segments in the phase loop feature table and the start and end timestamps of the monotonic segments in the rate return consistency table are aligned one by one. Within each aligned monotonic segment, the directed area, directed sign, start and end timestamps of the stable segment, the direction description of the strain rate, and the direction description of the return difference are read in a fixed order. When the directed area of ​​the loop is close to zero and the strain rate direction description is symmetrical between the upward and downward segments, the output is "normal". When the directed area of ​​the loop increases continuously and the strain rate direction description is continuously in the same direction within the stable segment, the output is "suspicious degradation". When the directed sign of the loop is flipped and the direction description of the return difference is continuously inconsistent between adjacent monotonic segments and a candidate anomaly index appears at the same time, the output is "possibly early debonding".

[0069] The textual conclusions and corresponding evidence (directed area of ​​the lap, directional sign, direction of strain rate, direction of backflow difference, and index of candidate anomalies) of all monotonic segments within each short time window are compiled in chronological order to generate a health assessment report that includes the start and end timestamps of the short time window, the start and end timestamps of the monotonic segments, and the health conclusions and the sources of the evidence.

[0070] S3: Based on the thermal microcirculation data package and health assessment report, the set of candidate anchor points is screened using the temperature-sensitive drift stripping algorithm, and abnormal points are cleared to obtain the purified strain flow object.

[0071] S3.1: Extract candidate anomalies from the health assessment report, locate the corresponding short time window and monotonic segment in the heat engine microcirculation data package through index projection, and obtain the anomaly mask;

[0072] It should be noted that each candidate anomaly in the health assessment report (including the start and end timestamps of the short time window, the start and end timestamps of the monotonic segment, the time location of the candidate anomaly and the explanation of the basis) is read one by one. In the heat engine microcirculation data package, a short time window index is built in ascending order of time according to the start and end timestamps of the short time window, and a monotonic segment index is built in each short time window according to the start and end timestamps of the monotonic segment.

[0073] For each candidate anomaly, the location is sequentially completed by monotonic segment index projection, and the timestamp of the candidate anomaly is used to locate it to a short time window interval; then, the short time window, monotonic segment and candidate anomaly are combined to form an anomaly mask entry, and the source of the basis is recorded (including the directional area of ​​the loop, the directional sign, the strain rate direction description, and the retracement direction description); the repeated time positions are deduplicated according to the short time window identifier and the monotonic segment identifier, and all anomaly mask entries are compiled in chronological order to obtain the anomaly mask.

[0074] S3.2: Perform equal-amplitude and same-temperature registration on the anomaly mask and the original strain sequence, calculate the local temperature-sensitive slope, and obtain the initial temperature drift baseline through discrete line integral;

[0075] It should be noted that, based on the anomaly mask and the thermomechanical microcirculation data packet, within each short time window, the time position in the anomaly mask is compared with the short time window timestamp one by one. When the short time window timestamp matches the time position in the anomaly mask, the corresponding row is deleted synchronously in the original strain sequence, adhesive layer temperature sequence and mechanical load sequence of the same window, and a gap mark is recorded on the time axis. The remaining samples are used as a whole as a valid sample set.

[0076] In the effective sample set, the mechanical load sequence of the same window is divided into four amplitude intervals within each short time window. The samples corresponding to each timestamp in the effective sample set are classified according to their intervals, and the sample index set of each interval is output as the equal amplitude candidate point set. In each equal amplitude candidate point set, the pair of times with the closest temperature is found according to the adhesive layer temperature sequence, and an equal amplitude and same temperature registration index is established.

[0077] For each pair of registration indices, the local temperature-sensitive slope value is calculated based on the original strain sequence and the adhesive layer temperature sequence; the local temperature-sensitive slope value and the adjacent temperature change are gradually accumulated along the time sequence and kept in correspondence with the start and end timestamps of the short time window to generate the initial temperature drift baseline;

[0078] The expression for calculating the local temperature-sensitive slope value is:

[0079] ;

[0080] in, This represents the local temperature-sensitive slope value. To assign values ​​to the original strain sequence, For the temperature sequence of the adhesive layer, and A pair of time positions determined by the isotropic registration index.

[0081] The better local temperature-sensitive slope can directly quantify the instantaneous sensitivity of strain to adhesive layer temperature, thereby distinguishing the slow temperature-driven drift from the rapid response caused by mechanical load; at the matching points with equal amplitude and temperature registration, the influence of operating condition fluctuations and noise can be reduced, and it is easy to accumulate over time to obtain a stable initial temperature drift baseline.

[0082] S3.3: Based on the initial temperature drift baseline, a sliding window polynomial fitting is used for smooth correction, and residual jump points are identified to generate a set of candidate anchor points.

[0083] It should be noted that, based solely on the initial temperature drift baseline and the start and end timestamps of the short-time windows and monotonic segments in the thermal engine microcirculation data package, a continuous sliding window is established along the time axis within each short-time window. The time-value pairs of the initial temperature drift baseline within the continuous sliding window are fitted using least squares to obtain the local smoothed value at the center of the window, and the difference between the value of the initial temperature drift baseline at the sampling point covered by the window and the corresponding local smoothed value is used to generate the residual sequence. Within the same short-time window, the residual sequence is segmented and constrained according to the start and end timestamps of the monotonic segments. Within each monotonic segment, the time point where the signs of adjacent differences are reversed and simultaneously fall at the local extreme value of the residual is taken as the residual jump point. Adjacent residual jump points are merged and deduplicated. All constrained and confirmed time points in each short-time window are aggregated into a set of candidate deletion anchor points.

[0084] The superior temperature-sensitive drift stripping algorithm can separate the slow drift caused by the adhesive layer temperature from the strain response, making the purified strain flow object closer to the real strain dominated by mechanical load. With the combination of anomaly mask constraint and initial temperature drift baseline accumulation, it reduces the interference of temperature fluctuation and noise on the judgment, while preserving the one-to-one correspondence between short time windows and monotonic segments.

[0085] S3.4: Merge the anomaly mask and the set of anchor points to be deleted, perform temperature drift stripping and anomaly point removal, and complete data repair through linear interpolation to obtain the purified strain flow object.

[0086] It should be noted that, based on the anomaly mask, the set of anchor points to be deleted, the initial temperature drift baseline and the thermal microcirculation data packet, within each short time window, the anomaly mask and the set of anchor points to be deleted are first merged according to their time positions to maintain a one-to-one correspondence with the start and end timestamps of the monotonic segment, generating a set of union tags.

[0087] Within each short time window, the initial temperature drift baseline is aligned point-by-point with the original strain sequence according to the timestamp, and the difference between the original strain sequence and the initial temperature drift baseline is used as the de-drift strain sequence. Within each short time window, using the time position of the union mark set as the index, corresponding rows are simultaneously deleted in the de-drift strain sequence, the adhesive layer temperature sequence, and the mechanical load sequence within the same window, and missing markers are registered on the time axis, thus forming continuous gap segments. For each gap in the continuous gap segment, the leftmost and rightmost most valid timestamps are located, and linear interpolation samples are inserted into the de-drift strain sequence. At the same time, corresponding linear interpolation samples are inserted into the adhesive layer temperature sequence and the mechanical load sequence within the same window with the same timestamp to complete the repair. Time alignment with the adhesive layer temperature sequence and the mechanical load sequence within the same window is maintained. The de-drifted and repaired strain sequence of each short time window is combined with the start and end timestamps of the short time window to output the purified strain flow object.

[0088] S4: Based on the purified strain flow object, perform short-time window finite element reference solution and generate a finite element reference response set through similarity matching.

[0089] S4.1: Perform dynamic decomposition of the load spectrum on the purified strain flow object to generate a short-time window load component set, and obtain the candidate reference strain field through viscoelastic time-varying analysis.

[0090] It should be noted that within each short time window, the purification strain flow object, the mechanical load sequence and the adhesive layer temperature sequence are aligned to the same time axis according to the start and end timestamps of the short time window, and a time position index is established. The mechanical load sequence of the same window is subjected to dynamic load spectrum decomposition (DC removal, end-point windowing, fixed frame length and fixed step size framing, intra-frame discrete Fourier transform, fixed bandwidth aggregation, and reconstruction of band-limited load time history by overlapping and adding on the original time axis). The load components are then organized into a short time window load component set according to the time position index and frequency band number.

[0091] The short-time-window load component set is used as the external load time history, and the adhesive layer temperature sequence is used as the material modulation input over time. In the viscoelastic time-varying analysis, the load is applied step by step according to the time position index at a fixed time step, and the reference strain field snapshot at the corresponding time position is output. All reference strain field snapshots are aggregated in the same short-time window according to the time position index to form a candidate reference strain field.

[0092] Furthermore, the reference strain field snapshot provides a spatial distribution map of the finite element reference response at discrete time locations, which solidifies the load time history and strain evolution under adhesive layer temperature modulation within a short time window frame by frame. This facilitates similarity matching and differential analysis with the purified strain flow object according to time and spatial locations, thereby providing a comparable and traceable reference benchmark for locating the adhesive thickness-sensitive area and subsequent online consistency compensation verification.

[0093] S4.2: Perform differential operations on the candidate reference strain field and calculate the gradient modulus field to identify regions sensitive to changes in adhesive thickness and obtain the adhesive thickness sensitive difference gradient field.

[0094] It should be noted that within each short time window, candidate reference strain field snapshots are aligned according to grid coordinates. Forward and backward differences are performed on adjacent grid nodes along two coordinate directions of the adhesive layer plane of the strain gauge, and symmetrical combinations are taken to obtain in-plane two-dimensional differences, which are then synthesized into a difference field. The gradient modulus is then calculated, and the expression is as follows:

[0095] ;

[0096] in, For gradient modulus, Scalar values ​​for the candidate reference strain field in the grid coordinates. For the row index of the grid coordinates, The column index for the grid coordinates.

[0097] All gradient moduli in the mesh are integrated into a gradient moduli field. In the neighborhood of the adhesive layer, local extremum connectivity tracking and spatial overlap verification of adjacent frames are used for joint determination. The connected band formed along the continuous extremum ridge is used as the adhesive thickness sensitive difference gradient field.

[0098] Furthermore, local extremum connectivity tracing refers to connecting local extremum points on adjacent grids into continuous ridges according to eight neighborhoods within the gradient modulus field, and recording the temporal and spatial location indices.

[0099] Adjacent frame spatial overlap verification refers to using the neighborhood of the ridge line of the previous frame as the search range in adjacent time frames, retaining the connected segment with the largest overlap area with the ridge line of the current frame, in order to confirm the continuity of the ridge line on the time axis.

[0100] Ideally, the gradient modulus can highlight the intensity of spatial variation of the strain field within the adhesive layer plane. High values ​​correspond to regions of abrupt response changes or stress concentration, making it easier to quickly locate the most sensitive connecting bands to changes in adhesive thickness from candidate reference strain fields. In subsequent steps, using the high ridge of the gradient modulus as a spatial constraint can improve the positioning accuracy of sensitive difference extraction and similarity matching, and reduce interference from irrelevant regions.

[0101] S4.3: Extract the adhesive thickness sensitive difference from the gradient field, combine it with the purification strain flow segment for similarity matching, and generate a labeled reference response;

[0102] It should be noted that, based on the adhesive thickness sensitive difference gradient field and the purification strain flow object, a sequential index is established in each short time window according to the time and space positions given by the adhesive thickness sensitive difference gradient field, and the adhesive thickness sensitive difference is directly read from the adhesive thickness sensitive difference gradient field.

[0103] In the purified strain flow object, a purified strain flow segment centered on the time position is extracted. Simultaneously, a reference strain segment is extracted from the candidate reference strain field at the same time and spatial positions. Normalization is performed on this pair of segments, and the standardized cross-correlation peak value is calculated within a small time lag as a similarity score. The expression for calculating the standardized cross-correlation peak value is:

[0104] ;

[0105] in, To standardize the peak cross-correlation, This represents the number of overlapping sampling points between the two segments. This is the index value of the number of overlapping sampling points between the two segments. To purify the strain flow segment in the first The values ​​of each sampling point For the candidate reference strain segment in the first The values ​​of each sampling point.

[0106] The time location, spatial location, adhesive thickness sensitivity difference, similarity score, and reference strain segment source identifier are merged into a single record and compiled in chronological order to generate a tagged reference response.

[0107] In a better way, the standardized cross-correlation peak value, as a similarity score, can stably characterize the shape consistency between the purified strain flow segment and the reference strain segment even when amplitude differences and DC offsets exist. It can also automatically align the closest time sequence positions within a small time lag, thereby providing an objective, comparable, and sortable matching basis for the selection of labeled reference responses.

[0108] S4.4: Aggregate and smoothly stitch all tagged reference responses in chronological order to generate a finite element reference response set.

[0109] It should be noted that all tagged reference responses are sorted in ascending order of time position within each short time window and merged according to spatial position; multiple records with the same time position and spatial position are weighted by similarity score to synthesize the reference strain segment output values ​​to obtain a single reference strain point; the reference strain trajectories are formed by connecting them sequentially along the time position; a fixed-length sliding smoothing is applied to the reference strain trajectories; and linear interpolation is performed to fill the time gaps caused by the removal of outliers; at the boundaries of adjacent short time windows, smooth splicing is performed according to the time positions on both sides, and all trajectories are aggregated in short time window order to output the finite element reference response set.

[0110] S5: Compare the finite element reference response set with the purified strain flow object segment by segment, and obtain the intelligent compensation strain result report of the bonding thickness through online consistency compensation verification.

[0111] S5.1: Time-align the finite element reference response set with the purified strain flow object to generate a reference measured strain sequence pair;

[0112] It should be noted that, based on the finite element reference response set and the purified strain flow object, a time position index is established in each short time window according to the start and end timestamps of the short time window; on the time position index, the reference strain time points in the finite element reference response set are matched one by one with the strain time points in the purified strain flow object, and for non-overlapping time points, the left and right adjacent effective time points are linearly aligned at the index position to generate the reference strain value and the purified strain flow value.

[0113] The reference strain values ​​and the purified strain flow values ​​are paired up at the same time position according to the time position index. All the paired records are then compiled in short time window order to generate a reference measured strain sequence pair.

[0114] S5.2: Calculate the strain difference based on the reference measured strain sequence, and perform consistency verification by combining the adhesive thickness sensitive difference gradient field to obtain the thickness sensitive difference segment;

[0115] It should be noted that, based on the reference measured strain sequence and the gradient field of the adhesive thickness sensitivity difference, within each short time window, the time position index and spatial position are matched one-to-one, and the difference between the reference strain value and the purification strain value is read as the strain difference sequence; the adhesive thickness sensitivity difference is read at the same time position and spatial position, and the continuous interval of "the strain difference value and the adhesive thickness sensitivity difference value have the same sign and change in the same direction with the time position index" is used as the judgment criterion. The start and end time position index and spatial position are recorded at the beginning and end of each continuous interval, and the thickness sensitivity difference segment is compiled and output.

[0116] S5.3: Perform inverse linear interpolation on the thickness-sensitive difference segment, and obtain the equivalent bond thickness sequence by real-time calculation of the equivalent bond thickness;

[0117] It should be noted that within each thickness-sensitive difference segment, strain difference values ​​are read point by point according to time position index, and adhesive thickness sensitive difference values ​​are read at the same time and spatial positions. Inverse linear interpolation is used to map the strain difference values ​​to equivalent adhesive thickness increments and write them into the sequence in real time. The ratio of the strain difference value to the adhesive thickness sensitive difference value is used as the equivalent adhesive thickness increment. The equivalent adhesive thickness at the previous time position is used as the starting point, and the next starting point is directly connected at the end of the segment. Finally, the equivalent adhesive thickness sequence is compiled and output in short time window order.

[0118] S5.4: Based on the equivalent bond thickness sequence, perform online compensation on the purified strain flow object and generate a report on the intelligent compensation strain results of bond thickness.

[0119] It should be noted that a time position index is established based on the start and end timestamps of the short time window, and the equivalent bond thickness sequence and the bond thickness sensitivity difference gradient field are aligned one by one at the same time and spatial positions. Within the thickness sensitivity difference segment, the equivalent bond thickness increment in the equivalent bond thickness sequence and the bond thickness sensitivity difference in the bond thickness sensitivity difference gradient field are read. The product of the equivalent bond thickness increment and the bond thickness sensitivity difference is used as the compensation value. The equivalent bond thickness increment and the bond thickness sensitivity difference are read at each time and spatial position, and the product of the equivalent bond thickness increment and the bond thickness sensitivity difference is recorded in the result according to "positive if the signs are the same, negative if the signs are opposite" to obtain the point-by-point compensation amount. The point-by-point compensation amount is directly compensated to the purified strain flow object to obtain the compensated strain value.

[0120] In the non-thickness-sensitive difference section, the original values ​​of the purified strain flow object are maintained, the time gaps of the previous abnormal points are cleared, and the alignment and repair are completed by linear interpolation according to the time position index of the reference measured strain sequence. The time position index correspondence between the strain curves before and after compensation, the thickness-sensitive difference section range and the equivalent adhesive thickness sequence is compiled in short time window order, and the intelligent adhesive thickness compensation strain result report is output.

[0121] In summary, this invention constructs a closed-loop intelligent compensation method through a temperature-sensitive drift stripping algorithm and online consistency compensation verification. This method accurately separates temperature drift from actual mechanical strain and dynamically compares real-time data with the finite element model. It achieves a leap from static calibration to dynamic sensing in strain measurement, improving the accuracy and reliability of long-term monitoring.

[0122] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent compensation of strain gauge bonding thickness based on finite element analysis, characterized in that: include, Collect raw data on the operating conditions of the measuring points and perform preprocessing to obtain data on the micro-circulation of the heat engine; Based on the thermomechanical microcirculation data package, a preliminary assessment of the health status of the strain gauge adhesive layer is performed, and a health assessment report is generated. Based on the thermal engine microcirculation data package and health assessment report, the set of candidate anchor points is screened by temperature-sensitive drift stripping algorithm, and abnormal points are removed to obtain the purification strain flow object; Based on the purified strain flow object, a short-time window finite element reference solution is performed, and a finite element reference response set is generated through similarity matching; The finite element reference response set is compared segment by segment with the purified strain flow object, and the bonding thickness intelligent compensation strain result report is obtained through online consistency compensation verification.

2. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 1, characterized in that: The raw data of the measuring point includes timestamp, raw strain, strain gauge adhesive temperature, operating condition range code, and mechanical load value; The preprocessing includes noise filtering, standardization of physical quantity units, outlier removal, and decoupling of characteristic signals.

3. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 2, characterized in that: The thermomechanical microcirculation data package includes short time window start and end timestamps, original strain sequence, adhesive layer temperature sequence, and same window mechanical load sequence.

4. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 3, characterized in that: The preliminary assessment of the strain gauge adhesive layer health status based on the thermomechanical microcirculation data package, and the generation of a health assessment report, are carried out through the following specific steps. Short-window monotonic decomposition is performed on the heat engine microcirculation data package to obtain the basic quantization table, and the directed area of ​​the loop is calculated by the phase loop geometry algorithm to generate the phase loop feature table. Based on the phase loop feature table and the same window mechanical load sequence, the strain rate is obtained by the steady segment differential back-back method and the candidate anomaly point index is marked to form a rate back-back consistency table. Perform sequence relation analysis on the phase loop characteristic table and the rate return consistency table to generate a health assessment report.

5. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 4, characterized in that: The process of filtering candidate anchor points based on the heat engine microcirculation data package and health assessment report using a temperature-sensitive drift stripping algorithm is as follows: Extract candidate anomalies from the health assessment report, locate the corresponding short time window and monotonic segment in the heat engine microcirculation data package through index projection, and obtain the anomaly mask; The anomaly mask and the original strain sequence are registered at the same amplitude and temperature. The local temperature-sensitive slope is calculated, and the initial temperature drift baseline is obtained by discrete line integration. Based on the initial temperature drift baseline, a sliding window polynomial fitting is used for smooth correction, and a set of candidate deletion anchor points is generated by identifying residual jump points.

6. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 5, characterized in that: The process of obtaining the purification strain flow object refers to fusing the anomaly mask and the set of candidate anchor points, performing temperature drift stripping and anomaly point removal, and obtaining the purification strain flow object through linear interpolation.

7. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 6, characterized in that: The specific steps for performing short-time window finite element reference solution based on the purified strain flow object are as follows. The load spectrum of the purified strain flow object is dynamically decomposed to generate a short time window load component set, and the candidate reference strain field is obtained through viscoelastic time-varying analysis. The candidate reference strain field is subjected to differential operation and the gradient modulus field is calculated to identify the region that is sensitive to the change in adhesive thickness and obtain the adhesive thickness sensitive difference gradient field.

8. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 7, characterized in that: The specific steps for generating the finite element reference response set are as follows: The adhesive thickness sensitive difference is extracted from the gradient field of the adhesive thickness sensitive difference, and similarity matching is performed with the purification strain flow object to generate a labeled reference response; All tagged reference responses are aggregated and smoothly stitched together in chronological order to generate a finite element reference response set.

9. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 8, characterized in that: The step-by-step comparison between the finite element reference response set and the purified strain flow object refers to aligning the finite element reference response set with the purified strain flow object in time sequence, and generating a reference measured strain sequence pair through step-by-step comparison.

10. The intelligent compensation method for strain gauge bonding thickness based on finite element analysis as described in claim 9, characterized in that: The specific steps for obtaining the intelligent compensation strain result report for bond thickness are as follows: The strain difference is calculated based on the reference measured strain sequence, and consistency is verified by combining the adhesive thickness sensitive difference gradient field to obtain the thickness sensitive difference segment. Inverse linear interpolation is performed on the thickness-sensitive difference segment, and the equivalent bond thickness sequence is obtained by real-time calculation of the equivalent bond thickness. Based on the equivalent bond thickness sequence, online compensation is performed on the purified strain flow object to generate a smart bond thickness compensation strain result report.

Images