Cryogenic low-temperature tensile test temperature drift compensation method and system

Abstract

The invention provides a cryogenic low-temperature tensile test temperature drift compensation method and system, and the method comprises the steps: applying a standard load at a preset temperature point, and collecting an output value of a load sensor to obtain a sample group; establishing an initial temperature drift model according to the sample group; acquiring an output value of a temperature and load sensor in a tensile test, and inputting the temperature and the output value into the initial temperature excursion model to calculate a temperature excursion compensation value; calculating a compensated net load according to the output value and the temperature drift compensation value; calculating a residual error according to the net load and a theoretical stress-strain curve to obtain a residual error group; and updating the initial temperature drift model according to the residual group to obtain a final temperature drift model, wherein the final temperature drift model performs compensation according to the real-time temperature and the output value.

Description

Technical Field

[0001] The present application relates to the technical field of deep-cold and low-temperature mechanical property testing, and more specifically to a deep-cold and low-temperature tensile test temperature drift compensation method and system thereof. Background Art

[0002] During cryogenic tensile testing, load cells are affected by temperature fluctuations, resulting in temperature drift and nonlinear offset in the output signal. Especially in extremely low temperatures of 4K to 77K, the gauge factor and resistance temperature coefficient of the sensor's sensitive elements can significantly change. Existing compensation methods often use fixed calibration values, which are unable to adapt to real-time temperature changes and drift behavior under complex stress conditions, affecting the accuracy and stability of tensile test data. Summary of the Invention

[0003] The purpose of this application is to provide a method and system for compensating temperature drift in deep-cold and low-temperature tensile testing to correct the temperature drift error of the load sensor, improve the load measurement accuracy, and ensure stable and reliable test data.

[0004] In order to achieve the above-mentioned objectives, the present application provides, on the one hand, a method for compensating temperature drift in a deep-cold and low-temperature tensile test, comprising: applying a standard load at a preset temperature point, collecting the output value of a load sensor to obtain a sample group; establishing an initial temperature drift model based on the sample group; collecting the output values ​​of the temperature and load sensors during the tensile test, and inputting the temperature and output values ​​into the initial temperature drift model to calculate a temperature drift compensation value; calculating a compensated net load based on the output value and the temperature drift compensation value; calculating a residual based on the net load and a theoretical stress-strain curve to obtain a residual group; updating the initial temperature drift model based on the residual group to obtain a final temperature drift model, and the final temperature drift model is compensated based on the real-time temperature and output value.

[0005] Preferably, the sample group includes zero-load output values ​​collected at multiple different preset temperature points and at least two standard load output values ​​with different amplitudes; wherein: the zero-load output value is used to establish a temperature drift offset function; the standard load output value is used to establish a sensitivity change function.

[0006] Preferably, the initial temperature drift model is a second-order polynomial function with temperature as a variable, and the calculation formula of the temperature drift compensation value is: ΔF(T)=a0+a1T+a2T 2 ; Where a0, a1, and a2 are the error correction coefficients obtained by fitting, and T is the temperature.

[0007] Preferably, the calculation formula of the net load is: f(t)=S(t)-ΔF(T); wherein S(t) is the output value of the load sensor, and ΔF(T) is the temperature drift compensation value.

[0008] Preferably, the residual is calculated as follows: R(t)=σ(t)-σ′(t); wherein σ(t) is the currently measured stress, and σ′(t) is the theoretical stress obtained by strain mapping.

[0009] Preferably, the theoretical stress-strain curve is obtained by performing regression fitting on the linear segment data in the initial stage of the tensile test.

[0010] Preferably, the updating method of the initial temperature drift model includes: using the residual group in the sliding window to fit the error correction coefficient by weighted least squares method.

[0011] Preferably, a preset sampling period is set, and the error correction coefficient is updated using the weighted least squares method every preset sampling period.

[0012] Preferably, the preset temperature points cover at least three temperature zones to be measured, and the three temperature zones to be measured are a low temperature section, an intermediate temperature section, and a high temperature section.

[0013] On the other hand, the present application provides a temperature drift compensation system for deep cold and low temperature tensile testing, comprising: a sample acquisition module, configured to apply a standard load at a preset temperature point and collect the output value of the load sensor to obtain a sample group; an initial modeling module, configured to establish an initial temperature drift model based on the sample group; a compensation calculation module, configured to input the temperature value and the sensor output value into the initial temperature drift model to calculate the temperature drift compensation value, and calculate the compensated net load based on the output value and the temperature drift compensation value; a residual analysis module, configured to calculate the residual based on the net load and the theoretical stress-strain curve to obtain a residual group; a model update module, configured to update the initial temperature drift model based on the residual group to obtain a final temperature drift model; and a dynamic compensation module, configured to perform temperature drift compensation on the temperature values ​​and sensor output values ​​collected in real time based on the final temperature drift model.

[0014] It should be understood that the foregoing general description and the following detailed description are merely illustrative and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings, which are incorporated herein and form a part of the specification, illustrate one or more embodiments of the present application and, together with the description, serve to explain the principles of the present application and to enable one of ordinary skill in the relevant art to make and use the present application.

[0016] Figure 1 This is a flow chart of a load error compensation method for a cryogenic tensile test process provided by an embodiment of the present application;

[0017] Figure 2 is a flow chart of the sample group construction method provided in an embodiment of the present application;

[0018] Figure 3 This is a flow chart of a method for constructing and applying a second-order polynomial temperature drift compensation model with temperature as a variable provided in an embodiment of the present application;

[0019] Figure 4 This is a flow chart of the residual calculation method provided in the embodiment of the present application;

[0020] Figure 5 is a flow chart of a method for constructing a theoretical stress-strain curve provided in an embodiment of the present application;

[0021] Figure 6 This is a flow chart of a method for updating a temperature drift model based on a sliding window residual group provided in an embodiment of the present application;

[0022] Figure 7 This is a flow chart of a temperature drift model updating method based on a fixed sampling period provided in an embodiment of the present application;

[0023] Figure 8 This is a flow chart of the temperature drift model initialization method provided in an embodiment of the present application;

[0024] Figure 9 This is a flow chart of the stress-strain model self-identification method provided in an embodiment of the present application;

[0025] Figure 10 This is a structural schematic diagram of a load error compensation system for a deep cold tensile test process provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments may be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, the description of these embodiments is intended to make this application more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to provide a deeper understanding of the embodiments of this application.

[0027] In one embodiment of the present application, a load error compensation method for a cryogenic tensile test process is provided, which is applicable to material mechanical property testing performed within a range of 4K to 77K.

[0028] Specifically, if Figure 1 As shown, the method includes the following steps:

[0029] S10: applying a standard load to the load sensor at multiple preset temperature points and collecting its output value, thereby forming a data sample group containing the relationship between temperature information, load input and load sensor response;

[0030] S20: establishing an initial temperature drift model based on the sample group, so as to reflect the drift characteristics of the load sensor under different temperature conditions.

[0031] S30: During the tensile test, real-time temperature data and load sensor output values ​​are continuously collected and input into the initial temperature drift model to calculate the temperature drift compensation value corresponding to the current moment.

[0032] S40: The compensation value is combined with the original output value of the load sensor to obtain the current net load value, which is used for subsequent stress analysis.

[0033] S50: The stress value calculated based on the net load and the specimen strain data is compared with the theoretical stress-strain curve to form a residual group to reflect the difference between the current model prediction value and the actual test behavior.

[0034] S60: updating the initial temperature drift model according to the residual group to obtain a final temperature drift model that is more suitable for the current test state, and using the final model in real time to perform temperature drift compensation in subsequent tests.

[0035] In one embodiment of the present application, a sample group construction method is provided to support the training of a temperature drift compensation model. The sample group includes zero-load output values ​​and standard load output values ​​of at least two different amplitudes collected at multiple different preset temperature points, and is used to establish a temperature drift offset function and a sensitivity change function, respectively.

[0036] like Figure 2 As shown, S11: before the tensile test, multiple preset temperature points covering the test temperature range are selected, for example: 77K, 50K, 30K, and 10K.

[0037] S12: At each preset temperature point, the load is kept at 0, that is, no sample is loaded, and the output value S0(t) of the load sensor is collected. This output value should theoretically be 0, but due to the influence of temperature drift, there is an offset. Record the zero-load output value at each temperature to form a zero-load sample set: {(T i ,S0(T i ))|i=1,2,…n}; Use this zero-load sample set to fit the temperature drift offset function, and the formula is as follows:

[0038] Δ0(T)=f0(T);

[0039] Where Δ0(T) is the zero offset value of the load sensor at temperature T, and the function form can be a linear or quadratic polynomial.

[0040] S13: At each preset temperature point, further apply at least two standard loads of different amplitudes, for example, 50N and 100N, and record the output values ​​S1 (T, F) of the load sensor respectively.

[0041] After subtracting the zero offset value corresponding to the preset temperature point, the effective response is obtained, and the calculation formula is as follows:

[0042] Δ1(T,F)=S1(T,F)-S0(t);

[0043] Then calculate the unit response value under different load amplitudes to fit the sensitivity change function, the formula is as follows:

[0044]

[0045] Where k(T) is the temperature-related equivalent sensitivity, which is used to indicate the response of the load sensor to unit force and changes with temperature.

[0046] The sensitivity of multiple preset temperature points is fitted into a sensitivity change function, and the formula is as follows:

[0047] k(T)=f1(T);

[0048] S14: Finally, a complete temperature drift compensation model is constructed, combining the offset function and the sensitivity function to express the relationship between the load sensor output and the actual load F at any temperature T. The formula is as follows:

[0049] S1(T,F)=Δ0(T)+k(T)·F;

[0050] This model is used for subsequent real-time compensation calculations, inferring the load value from the load sensor output, or calculating the current temperature drift compensation deviation. This example uses a zero load point and two standard points to independently identify additive offset and proportional response errors.

[0051] In one embodiment of the present application, a method for constructing a temperature drift model based on a second-order polynomial with temperature as a variable is provided, which is used to correct the temperature drift error in the output of a load sensor in real time during a cryogenic tensile test.

[0052] like Figure 3As shown, S21: Before the start of the test, test data are collected at multiple preset temperature points, such as 77K, 50K, 30K, and 10K, to obtain the load response deviation of the load sensor under different temperature conditions, forming a training sample group. Based on the temperature T recorded in the sample group and the corresponding error deviation value ΔF, a polynomial fitting method is used to establish a temperature drift compensation model in the following form:

[0053] ΔF(T)=a0+a1T+a2T 2 ;

[0054] Where T is the temperature of the current test environment, ΔF(T) is the estimated temperature drift error at that temperature, and a0, a1, and a2 are the error correction coefficients obtained by least squares fitting.

[0055] S22: In an example, the fitting results are: a0 = 1.12, a1 = -0.045, a2 = 0.00038. When the current temperature is 42.0K, the compensation value can be calculated according to the temperature drift compensation model:

[0056] ΔF(42.0)=1.12-0.045·42.0+0.00038·42.02=-0.42N;

[0057] This means that the load cell output contains a negative offset of approximately 0.42N.

[0058] In one embodiment of the present application, a net load calculation method is provided, which is applicable to a load error compensation system in a cryogenic tensile test. The method calculates the actual net load based on the real-time output value of the load sensor and a temperature drift compensation model.

[0059] During a tensile test, the load sensor outputs a real-time signal related to the applied load, denoted as S(t), where t represents the current sampling time. Simultaneously, a pre-built temperature drift compensation model is used to calculate the temperature drift compensation value ΔF(T) at that moment, based on the current temperature T(t).

[0060] The calculation formula of the net load is as follows:

[0061] F(t)=S(t)-ΔF(T);

[0062] Where F(t) is the net load at time t, that is, the actual load after removing the temperature drift error, S(t) is the original output value of the load sensor at time t, and ΔF(T) is the temperature drift compensation value calculated based on the current temperature T, which is used to reflect the offset effect of temperature on the sensor output.

[0063] In one example, the current load sensor output value S(t) = 28.4N is collected, the current temperature is T = 45.0K, and ΔF(45.0) = 1.2N is substituted into the temperature drift compensation model. The net load calculation result is: F(t) = 28.4-1.2 = 27.2N, indicating that the actual effective load acting on the specimen at this time is 27.2N.

[0064] Through this example, the system can eliminate temperature errors in the load sensor output throughout the test process, ensuring that the load data is authentic and stable, which can serve as the basis for subsequent stress-strain analysis.

[0065] In one embodiment of the present application, a residual calculation method is provided for evaluating the load measurement accuracy after temperature drift compensation and providing feedback basis for updating the temperature drift model.

[0066] like Figure 4 As shown, S51: During the tensile test, at the current time point, the true stress value is calculated from the compensated net load value and the cross-sectional area of ​​the specimen. The calculation formula is as follows:

[0067]

[0068] Where F(t) is the net load at the current moment, that is, the load value after temperature drift compensation, A is the cross-sectional area of ​​the specimen, which is a known fixed parameter, and σ(t) is the calculated measured stress. This measured stress reflects the actual stress level actually applied to the specimen and can be used to compare with the theoretical stress.

[0069] S52: Furthermore, the theoretical stress is the predicted stress at the current strain in the theoretical stress-strain curve, and its formula is as follows:

[0070] σ′(t)=f(ε(t));

[0071] Where ε(t) is the strain value of the current sample, f() is the theoretical model function, and the difference between the measured stress and the theoretical stress is the residual. The residual can be used to infer whether the current temperature drift model is accurate.

[0072] In one example, the theoretical model function adopts linear segment fitting, and its formula is as follows:

[0073] σ′(t)=E·ε(t);

[0074] Where E is the elastic modulus obtained by fitting the measured stress-strain relationship in the initial elastic stage of the tensile test, which is used to construct the theoretical stress prediction function.

[0075] S53: Calculate the residual at the current moment according to the following formula:

[0076] R(t)=σ(t)-σ′(t);

[0077] The residual R(t) represents the degree of deviation between the measured stress and the theoretical stress and is used to evaluate the accuracy of temperature drift compensation. The smaller the residual, the more accurate the temperature drift compensation model. If the residual shows a continuous deviation trend within a certain time window, the model update mechanism is triggered to further optimize the compensation coefficient.

[0078] In one example, the net load after compensation at one moment F(t) = 30.0 N, and the cross-sectional area of ​​the sample A = 10.0 mm 2 , then the measured stress σ(t)=30.0 / 1.0×10 -5 =3.0MPa. If the strain at this moment is 0.00010 and the elastic modulus E = 28GPa, then the theoretical stress σ′(t) = 28GPa×0.00010 = 2.8MPa, and the residual is: R(t) = 3.0-2.8 = 0.2MPa. This residual is used as one of the bases for updating the temperature drift model and enters the sliding residual window for error trend analysis and compensation model parameter optimization.

[0079] In one example, to obtain a theoretical stress-strain relationship function for residual calculation, a linear fit is performed on the stress and strain data of the specimen in the early stage of the tensile test to determine the elastic modulus E.

[0080] like Figure 5 As shown, S521: select the first several sampling points after the start of the tensile test, such as the first 10 to 20 groups of data. These data are located in the elastic deformation range before the material enters the yield stage, which can ensure a linear relationship between stress and strain.

[0081] S522: For each set of data, calculate the measured stress: The strain value ε(t) is measured at the same time.

[0082] S523: Use the least squares method to perform a linear regression fit on the selected stress-strain data, with strain as the independent variable and stress as the dependent variable. The fitting equation is in the form of: σ = E·ε. The slope of the fitting result is the equivalent elastic modulus E of the sample under the current test conditions.

[0083] S524: Further, the determination coefficient R of the linear fit can be set. 2 Not lower than the preset threshold, where the preset threshold can be set to 0.98 to ensure that the selected segment data has high linearity; if the requirement is not met, the sampling window is expanded or the abnormal points are excluded and refitted.

[0084] In one embodiment of the present application, a temperature drift model updating method for optimizing temperature drift compensation accuracy is provided, which realizes adaptive correction of initial temperature drift model parameters by performing weighted least squares fitting on the residual group in the tensile test process.

[0085] Based on the above examples, it has been obtained that the temperature drift compensation model is in the form of a second-order polynomial:

[0086] ΔF(T)=a0+a1T+a2T 2 ;

[0087] like Figure 6 As shown in S61: During the actual stretching process, the system calculates the measured stress σ(t) based on the net load F(t) and strain ε(t) at each time point t, and compares it with the theoretical stress σ′(t) to obtain the residual:

[0088] R(t)=σ(t)-σ′(t);

[0089] The residual values ​​calculated at the latest N moments, for example, N=10, and the corresponding temperature T(t) are combined into a sliding window residual group: {(T i , R i )|i=t-N+1, ...t};

[0090] S62: To enhance the sensitivity of recent residuals to temperature drift model updates, a weight factor w is introduced i , perform weighted least squares fitting on each set of data in the sliding window and update the temperature drift model coefficients:

[0091]

[0092] Among them, R i is the i-th residual, T i is the corresponding temperature.

[0093] Through the above fitting, the new temperature drift error correction coefficients (a′0, a′1, a′2) are obtained, and the updated temperature drift model is constructed:

[0094] ΔF′(T)=a′0+a′1T+a′2T 2 ;

[0095] S63: The updated temperature drift model ΔF′(T) replaces the original model and is used to calculate the temperature drift compensation value during the subsequent stretching process. By limiting the number of residual samples through a sliding window and controlling the accumulation of historical deviations, the response rate of the temperature drift model can be improved. A weighting strategy is also introduced to enhance the influence of recent data, allowing the temperature drift model to adapt to changes in the load sensor state or environmental interference. Furthermore, a least-squares fitting method ensures update convergence and numerical stability, facilitating efficient computation in embedded devices.

[0096] In one embodiment of the present application, a temperature drift model update mechanism based on a fixed sampling period is provided to periodically optimize the error correction coefficient of the temperature drift compensation model during the deep cold tensile test, thereby ensuring that the sensor compensation accuracy continues to adapt to changes in the environment and equipment status as the test progresses.

[0097] like Figure 7 As shown, S631: set a preset sampling period in the initialization phase, denoted as T s , which means that every time T s The temperature drift model update operation is performed after each data sampling point. The cycle value can be set according to the test duration, material type, and system stability. For example, set T s =500, that is, an update is performed after every 500 sets of data are collected, and the time interval is set, such as updating once every 60 seconds.

[0098] S632: In each sampling period, the continuously collected data include the current temperature T(t), the compensated net load F(T), the measured strain ε(t), and the calculated residual R(t). The T(t) and R(t) collected in the period form a residual sample set:

[0099] S633: After the cycle ends, the system performs weighted least squares fitting on the residual sample set and updates the error correction coefficients (a0, a1, a2) in the temperature drift model. The fitting objective function is:

[0100]

[0101] S634: After fitting is completed, the temperature drift compensation model is automatically replaced with the new model:

[0102] ΔF′(T)=a′0+a′1T+a′2T 2 ;

[0103] The new model is then used for compensation calculations during the next sampling period, forming a cycle of sampling, residual accumulation, and periodic updates. This periodic update prevents system oscillations caused by high-frequency adjustments and ensures that the temperature drift model gradually adapts to environmental changes and sensor drift during the test. The weighted least squares algorithm considers both historical trends and recent errors, resulting in more stable update results.

[0104] In one embodiment of the present application, a sample acquisition method is provided. Before constructing a temperature drift compensation model, temperature points covering at least three temperature zones to be measured are selected for data sampling to enhance the adaptability and fitting accuracy of the temperature drift model within the entire deep-cold test range.

[0105] In this embodiment, the entire test temperature zone is divided into the following three typical temperature sections, namely, low temperature section: 0K–30K, intermediate temperature section: 30K–60K, and high temperature section: 60K–90K. This partitioning method covers the common sensor error drastic change area, transition area, and relatively stable area in deep cryogenic testing.

[0106] When selecting preset temperature points, at least one representative preset temperature point is selected in each temperature zone for sample collection, for example, the low temperature section is selected at 10K, the middle section is selected at 45K, and the high temperature section is selected at 77K.

[0107] It is conceivable that if the accuracy of the temperature drift model needs to be improved, the density of sampling points can be increased in each temperature zone, for example, one point is set every 10K.

[0108] At the preset temperature point selected above, a sample collection operation is performed, including collecting the output value of the load sensor and constructing a data sample group containing temperature and output values. The obtained sample group will be used to fit the temperature drift model. By covering different temperature zones, it can be ensured that the model has good prediction and compensation capabilities throughout the entire operating temperature range, avoiding model distortion or excessive extrapolation error due to an overly narrow sample distribution.

[0109] The three-segment coverage temperature zone in this example can adapt to different thermal response characteristics, improve the breadth of the temperature drift model, ensure that the temperature drift model has training support in the low-temperature nonlinear region, intermediate transition region, and high-temperature stable region, and improve the robustness of the temperature drift model under complex temperature paths.

[0110] In one embodiment of the present application, a method for initializing an initial temperature drift model is provided, which combines experimental sampling data with historical or factory calibration data to construct an initial temperature drift model in a multi-source fusion manner. The method is suitable for occasions where high compensation accuracy is required or the test environment changes significantly.

[0111] like Figure 8 As shown, S101: the training sample data used for initialization includes two types of sources, namely: local field sampling data and historical / calibrated reference data. The first type of source is, during the test preparation stage, applying a standard load to the load sensor at multiple preset temperature points and collecting the load sensor output value; the second type of source includes the factory calibration curve of the load sensor, the residual correction model of the previous round of tests, or the statistical model of multiple specimens.

[0112] S102: Using a weighted least squares fitting method to fuse the two types of data, and obtaining an initial temperature drift model after fusion fitting.

[0113] S103: The model quality judgment index can also be set, such as the residual mean square error, etc. Only when the goodness of fit of the fused model is better than that of the single-source fitting model, the fused model is adopted; otherwise, it falls back to the pure measured sample modeling method to ensure the stability of the initial temperature drift model.

[0114] In an embodiment of the present application, there is provided a model self-identification method for stress–strain relationship, which is used to automatically construct a theoretical stress–strain curve suitable for residual analysis.

[0115] In actual tensile tests, the stress–strain behaviors of different materials (such as metals, polymers, ceramics, composite materials) in the initial loading stage are significantly different. Some materials have a longer elastic section, and the stress–strain is strictly linear. Some materials exhibit yield or non-linear responses early, making it difficult to model with simple linear fitting.

[0116] As Figure 9 shown, S501: In the initial stage of the tensile test, a continuous segment of stress–strain raw data samples is collected, and the number of sampling points is 10–30 to ensure coverage of the elastic stage.

[0117] S502: First, try to use first-order linear regression:

[0118] σ = E·ε + b;

[0119] In the formula, b is the intercept term.

[0120] In an ideal situation, b = 0, and the first-order linear regression expression is: σ = E·ε. However, in actual tests, due to factors such as the instrument not being zero-calibrated, the sensor having a temperature drift, and the initial prestress, all stress values may be shifted by a constant as a whole.

[0121] Calculate the coefficient of determination R 2 , set the coefficient of determination threshold R0 to 0.98. If R0 ≤ R 2 , it is considered that the stress–strain relationship is linear, and a linear model can be adopted, and the fitted E is the equivalent elastic modulus of the material.

[0122] S503: If R 2 < R0, it means that there are significant non-linear deviations in the data, and alternative methods of multi-segment linear fitting or polynomial fitting can be performed. Among them, in the method of multi-segment linear fitting, the initial interval is divided into two or three segments, and different slopes are fitted respectively; in the method of polynomial fitting, a second-order or third-order polynomial model can be adopted.

[0123] The above identification results will be used as the basis for generating the theoretical stress–strain curve for residual calculation. The identification results are recorded as the model type identifier, and subsequent residual calculations are uniformly carried out according to the following formula:

[0124] R(t)=σ(t)-σ′(ε(t));

[0125] Where σ′() is the output of the theoretical model selected for automatic identification. The residual is input into the temperature drift model update module to determine whether to update the correction coefficient.

[0126] This example avoids residual deviation caused by forcibly using a linear model, which inadvertently triggers frequent updates of the compensation model.

[0127] In one embodiment of the present application, a dynamic residual threshold triggering method is provided to determine whether a temperature drift model correction operation needs to be performed to avoid frequent updates of compensation parameters due to slight fluctuations or occasional errors.

[0128] This example counts the residual values ​​within the sliding window, extracts their statistical characteristic parameters (such as mean, standard deviation, variance, maximum value, etc.), and compares them with the dynamically calculated threshold value to determine whether to trigger the temperature drift model update operation.

[0129] Record the residual R(t) in each sampling period and maintain a sliding window of length N: R N ={R(t-N+1),...,R(t)}; N is the window length, which can be selected from 10 to 30 and is set according to the system response period.

[0130] In the current window, the calculated statistical indicators include mean, standard deviation, and maximum residual amplitude; the residual threshold is dynamically generated by combining the model fitting error, historical residual stability, and self-learning error model:

[0131] Y=λ·σ R +δ;

[0132] In the formula, λ is the adjustment factor, which is 2-3 and represents the tolerance to the deviation amplitude, σ R is the mean of the historical residual standard deviations, which can be initialized to the training phase value. δ is a constant offset used to offset small amounts of noise. It is conceivable that this threshold can be automatically adjusted as the system operates.

[0133] Furthermore, if any of the following conditions occurs within the sliding window, the temperature drift model update is triggered, including:

[0134] When the mean value is greater than the threshold, it indicates that the mean shift is significant;

[0135] When standard deviation > σ R When α is too large, it indicates that the residual fluctuation is too large, and the value of α is 1.5 to 2;

[0136] When the maximum residual amplitude>R max When , it indicates that the individual residual is too large, R maxIt is the preset maximum safety deviation (such as 5MPa).

[0137] Once any of the above conditions is met, an update is triggered, and the weighted least squares method is started to refit the temperature drift error correction coefficient. The new model parameters replace the current temperature drift model, and the triggering behavior is recorded in the log for subsequent abnormality tracking and training data accumulation. If the update is not triggered, the current temperature drift model continues to be used to keep the parameters stable.

[0138] After adopting this example, in typical high-sensitivity sensor tests, the model update frequency can be reduced, the temperature difference compensation value is stable, and the system response curve is smooth without frequent jumps.

[0139] In one embodiment of the present application, a temperature drift compensation system for deep-cold and low-temperature tensile testing is provided, which is suitable for material mechanics testing scenarios in low-temperature environments such as 4K to 77K, and is used to improve the stability and accuracy of load measurement by load sensors.

[0140] like Figure 10 As shown, Sample Collection Module 01 is used to apply standard loads to the load sensor and collect signals at multiple preset temperature points before the test. At each preset temperature point, at least two standard loads are applied while the load sensor output signals are recorded, forming a temperature-load-signal sample set. The collected data is used to establish a training dataset for the temperature drift model.

[0141] The initial modeling module 02 is used to establish an initial temperature drift model based on the sample group. It is based on the load sensor output and temperature data, and is modeled by fitting a second-order function polynomial. The initial temperature drift model is:

[0142] ΔF(T)=a0+a1T+a2T 2 ;

[0143] The output of the initial temperature drift model is the error compensation value corresponding to the temperature.

[0144] During the tensile test, the compensation calculation module 03 collects the current temperature value T(t) and the load sensor output value S(t) in real time, inputs the two into the initial temperature drift model, calculates the temperature drift compensation value ΔF(T), and calculates the net load according to the following formula:

[0145] F(t)=S(t)-ΔF(T);

[0146] The net load obtained is the actual load value after the temperature drift error has been eliminated.

[0147] The residual analysis module 04 converts the net load into stress and compares it with the theoretical model. The measured stress σ(t) is obtained from the net load and the cross-sectional area A. Based on the real-time strain ε(t), the theoretical stress value σ′(t) is constructed and the residual is calculated:

[0148] R(t)=σ(t)-σ′(t);

[0149] The residuals of multiple preset time points are combined into a residual group to provide feedback on modeling quality.

[0150] The model update module 05 is started periodically or according to a trigger condition, such as when the cumulative residual deviates from a threshold, and uses the residual group in the sliding window to refit the error correction coefficient by weighted least squares method, update the temperature drift model parameters a0, a1, and a2, and generate a new temperature drift model.

[0151] Dynamic compensation module 06 replaces the old model with an updated temperature drift model. In subsequent tensile tests, the final temperature drift model is continuously used to compensate for temperature and output in real time, forming a closed-loop adaptive system to ensure the accuracy of measurement data.

[0152] In one embodiment of the present application, a graphical interface and a manual correction module are provided to enhance the visual operation capability and human intervention capability of the system, making it suitable for low-temperature load measurement scenarios.

[0153] The graphical interface and manual correction module make it easy for operators to view the load compensation status in real time, debug model parameters, and manually intervene in system operation when necessary. The graphical interface and manual correction module include a real-time data monitoring interface, a model parameter display and adjustment panel, a residual trend graph and error threshold warning, a manual correction input and overwriting mechanism, and a log export and experimental recording function.

[0154] Specifically, the real-time data monitoring interface displays information including: current sampling time, real-time temperature, load sensor original output value, temperature drift compensation value, compensated net load, measured stress and theoretical stress, and residual. The display mode can be a table + line graph overlay, so that users can intuitively identify abnormal fluctuations.

[0155] The model parameter display and adjustment panel displays the current fitting parameters of the temperature drift model in read-only or edit mode. In default mode, a0, a1, and a2 automatically fitted by the current system are displayed. In debug mode, the user can switch to manual mode and modify the parameter value through the input box. After the user confirms, the system replaces the current temperature drift model and records the operation behavior.

[0156] The residual trend chart shows the residual distribution of the temperature drift model within a specific time period; if the residual exceeds the preset threshold, the system pops up a warning prompt and suggests whether the user should allow the system to update the model or intervene manually.

[0157] Manual correction input and override mechanisms prevent the automated model from becoming unsuitable for extreme test conditions. For example, if the user identifies a constant deviation in the system, they can directly enter a correction value for rapid offset calibration. All operations are logged in the background, including the operator, time, and parameter changes.

[0158] The system provides a log export function, including: parameter change records for each temperature drift model update, thresholds and response behaviors for each residual anomaly trigger, each manual intervention or correction operation, and experimental data snapshots (temperature, load, residuals, etc.).

[0159] While several embodiments of the present invention have been described, these embodiments are provided as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms and can be omitted, replaced, or modified without departing from the spirit of the invention. These embodiments and their variations are included within the scope and spirit of the invention and are also included in the invention described in the claims and their equivalents.

Claims

1. A method for compensating temperature drift in a cryogenic tensile test, comprising: Applying a standard load at a preset temperature point and collecting output values ​​of the load sensor to obtain a sample group; establishing an initial temperature drift model based on the sample group; collecting temperature and output values ​​of a load sensor during a tensile test, and inputting the temperature and output values ​​into the initial temperature drift model to calculate a temperature drift compensation value; Calculating a compensated net load based on the output value and the temperature drift compensation value; Calculating residuals based on the net load and the theoretical stress-strain curve to obtain a residual set; An initial temperature drift model is updated according to the residual group to obtain a final temperature drift model, and the final temperature drift model is compensated according to the real-time temperature and the output value.

2. The method according to claim 1, wherein The sample group includes zero-load output values ​​collected at multiple different preset temperature points and at least two standard load output values ​​with different amplitudes; wherein: The zero load output value is used to establish a temperature drift offset function; The standard load output value is used to establish a sensitivity change function.

3. The method according to claim 1, wherein The initial temperature drift model is a second-order polynomial function with temperature as a variable, and the calculation formula of the temperature drift compensation value is: ΔF(T)=a0+a1T+a2T 2 ; Where a0, a1, and a2 are the error correction coefficients obtained by fitting, and T is the temperature.

4. The method according to claim 3, wherein: The calculation formula of the net load is: F(t)=S(t)-ΔF(T); Where S(t) is the output value of the load sensor, and ΔF(T) is the temperature drift compensation value.

5. The method according to claim 1, wherein The calculation formula of the residual is: R(t)=σ(t)-σ ′ (t); Where σ(t) is the actual stress currently measured, σ ′ (t) is the theoretical stress obtained from strain mapping.

6. The method according to claim 1, wherein The theoretical stress-strain curve is obtained by regression fitting based on the linear segment data in the initial stage of the tensile test.

7. The method according to claim 3, wherein: The updating method of the initial temperature drift model includes: using the residual group in the sliding window to fit the error correction coefficient by weighted least square method.

8. The method according to claim 1 or 7, further comprising: A preset sampling period is set, and the error correction coefficient is updated using the weighted least squares method every preset sampling period.

9. The method according to claim 1, wherein The preset temperature points cover at least three temperature zones to be measured, and the three temperature zones to be measured are respectively a low temperature section, an intermediate temperature section and a high temperature section.

10. A temperature drift compensation system for cryogenic tensile testing, using the method according to any one of claims 1 to 9, wherein the system comprises: a sample collection module configured to apply a standard load at a preset temperature point and collect output values ​​of the load sensor to obtain a sample group; an initial modeling module, configured to establish an initial temperature drift model according to the sample group; a compensation calculation module configured to input the temperature value and the sensor output value into an initial temperature drift model to calculate a temperature drift compensation value, and calculate a compensated net load based on the output value and the temperature drift compensation value; a residual analysis module configured to calculate residuals according to the net load and a theoretical stress-strain curve to obtain a residual group; a model updating module configured to update the initial temperature drift model according to the residual group to obtain a final temperature drift model; and The dynamic compensation module is configured to perform temperature drift compensation on the real-time collected temperature value and the sensor output value based on the final temperature drift model.

Images