Array thermocouple multi-mode compensation and DIC stress field space-time coupling fusion method

Abstract

The invention relates to the technical field of high temperature sensing and data fusion, in particular to a method for array thermocouple multi-mode compensation and DIC stress field space-time coupling fusion, which comprises the following steps: step 1, in a thermotechnical signal intelligent processing and compensation unit, completing hardware and algorithm collaborative design of an electronic cold junction compensation module; 2, constructing a multi-algorithm fusion compensation module, integrating nonlinear correction, drift compensation and interference suppression functions, and accurately coping with various interference signals through a dynamic weighting strategy; step 3, adopting a sub-pixel-level matching algorithm and a homography matrix calibration technology to realize high-precision space alignment of the temperature and stress measurement units; and establishing a nonlinear incidence relation between the temperature and the stress based on an improved Gaussian process regression model. According to the invention, based on collaborative design of the thermotechnical signal intelligent processing and compensation unit and the DIC vision and temperature data conjoint analysis module, the core precision problem of temperature and stress detection in a high-temperature environment is solved through hardware optimization and algorithm innovation.

Description

Technical Field

[0001] This invention relates to the field of high-temperature sensing and data fusion technology, and in particular to a method for fusing multimodal compensation of array thermocouples with spatiotemporal coupling of DIC stress field. Background Technology

[0002] In the development of technologies for measuring specimen temperature and stress in high-temperature environments, existing solutions still suffer from several common bottlenecks, making it difficult to meet the measurement requirements for high precision and strong adaptability. Some technologies rely on heat dissipation devices to adjust thermal resistance for temperature compensation, and at the algorithm level, they only use a single frequency domain decomposition method to process interference signals, without building a multi-algorithm collaborative mechanism. This results in significantly insufficient measurement accuracy and response efficiency in high-temperature scenarios with transient temperature changes and multiple superimposed interferences. Other solutions, while capable of measuring temperature fields over a wide temperature range, only focus on the acquisition and reconstruction of temperature data, lacking correlation modeling between temperature and stress and effective spatial matching calibration methods, thus failing to support in-depth analysis of thermo-mechanical coupling. Furthermore, some solutions, although attempting to integrate optical systems and DIC data, neglect the key compensation link in thermocouple measurements, failing to integrate cold junction compensation and high-temperature nonlinearity compensation mechanisms. Spatial matching also relies on fixed hardware angles, limiting the accuracy and scenario adaptability of temperature-stress collaborative measurements.

[0003] The existing patent number CN202511063731.5 (A dynamic compensation processing method and system for thermocouple signals) relies on a heat dissipation device to adjust thermal resistance in terms of hardware, uses a non-contact temperature measuring device to synchronously collect temperature data, and transmits signals through compensation wires. In terms of algorithm, it builds a compensation database based on the "response delay-temperature change" correlation model, separates interference signals through frequency domain decomposition, reconstructs the compensation coefficients in combination with thermal resistance parameters, and performs closed-loop iterative correction. The above technical solution relies on the physical control of the heat dissipation device, which limits its adaptability and accuracy, and its interference processing is singular and cannot cope with multiple types of complex interference.

[0004] The existing patent number CN202510657741.5 (A Multi-point Distributed Thermocouple Array and Real-time Temperature Field Monitoring Method) uses a multi-material gradient thermocouple array to achieve temperature measurement from -200℃ to 2000℃, and uses a GaN-HEMT switching matrix to achieve a full array scan of ≤1ms, transmitting data via fiber optic or wireless links; the algorithm uses a cold-end constant temperature reference zone to determine the cold-end temperature, uses a fifth-order polynomial lookup table method to linearize the thermoelectric potential, and uses an FPGA to implement inverse distance weighted interpolation to reconstruct the temperature field; the above technical solution only focuses on the temperature field, without temperature-stress coupling analysis and multi-interference compensation, and the spatial matching calibration is insufficient.

[0005] The existing patent number CN202510601732.4 (an experimental measurement method for simultaneous measurement of dynamic caustics and DIC) uses hardware consisting of a dynamic caustics high-speed camera, a DIC bias lens, and optical components to construct a system that collects multiple optical and environmental parameters. The algorithm calculates and analyzes coefficients through three weighted coupling operations and compares them with standard values ​​to correct errors. The above technical solution focuses on the fusion of dynamic caustics and DIC data, but lacks a key thermocouple compensation mechanism. The accuracy of high-temperature measurement is not guaranteed, a temperature-stress correlation model is not established, and spatial matching depends on a fixed lens angle.

[0006] To address the aforementioned shortcomings, this application proposes a method for fusing array thermocouple multimodal compensation with DIC stress field spatiotemporal coupling. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method for the fusion of array thermocouple multimodal compensation and DIC stress field spatiotemporal coupling. This method revolves around the collaborative design of a thermal signal intelligent processing and compensation unit and a DIC vision and temperature data joint analysis module. Through hardware optimization and algorithm innovation, it solves the core accuracy problem of temperature and stress detection under high-temperature environments.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A method for fusing multimodal compensation of array thermocouples with spatiotemporal coupling of DIC stress field includes the following steps:

[0010] Step 1: In the intelligent processing and compensation unit for thermal signals, complete the hardware and algorithm co-design of the electronic cold junction compensation module;

[0011] Step 2: Construct a multi-algorithm fusion compensation module, integrating nonlinear correction, drift compensation and interference suppression functions, and accurately responding to various interference signals through a dynamic weighting strategy;

[0012] Step 3: Employ subpixel-level matching algorithm and homography matrix calibration technology to achieve high-precision spatial alignment of temperature and stress measurement units; finally, establish the nonlinear correlation between temperature and stress based on the improved Gaussian process regression (GPR) model, and introduce spatial regularization constraints to improve modeling accuracy.

[0013] Preferably, step 1 includes a hardware layer and an algorithm layer, wherein the hardware layer is a 10×10 thermocouple array (denoted as T). i,j, representing the thermocouple in the i-th row and j-th column of the array. Each thermocouple in the array is equipped with an independent miniature platinum resistance thermometer with an accuracy of ±0.1℃, and is directly attached to the cold junction of the thermocouple to collect the cold junction temperature T0 in real time. At the same time, a high-temperature heat-insulating data transmission cable is used to transmit the thermoelectric potential signal and the cold junction temperature signal. The outer layer of the cable is wrapped with a ceramic fiber heat insulation layer with a temperature resistance of ≥1000℃, and the inner layer is made of twisted pair to suppress electromagnetic interference.

[0014] At the algorithm level, a piecewise thermoelectric potential-temperature function fitting is adopted. For K-type thermocouples with an applicable range of -200℃ to 1300℃, the temperature range is divided into five sub-intervals: [-200, 0), [0, 400), [400, 800), [800, 1000), and [1000, 1300]. Each sub-interval is fitted with a high-order polynomial thermoelectric characteristic function f using the least squares method. k (t)(k=1,2,…,5)(e.g., within a certain interval, E=a0+a1T+a2T) 2 +...+a n T n The function coefficients are calibrated using a thermocouple calibration table, and the fitting error is <0.05mV.

[0015] Preferably, in step 1, the cold junction compensation calculation includes the following method:

[0016] Step 1: Measure the thermoelectric potential E corresponding to the temperature of the uncompensated measuring end. m,i,j =f(t) m,i,j )-f(t 0,i,j ), t m,i,j The temperature at the uncompensated measuring end; t 0,i,j E represents the cold junction temperature of the thermocouple. m,i,j This represents the uncompensated thermoelectric potential of the thermocouple.

[0017] The second step is to calculate the cold junction thermoelectric potential E based on the interval where the cold junction temperature T0 falls, using the corresponding thermoelectric characteristic function for that interval. 0,i,j =f k (t 0,i,j According to t 0,i,j Choose f in the interval k ;f k Let be the thermoelectric characteristic function corresponding to the k-th temperature range;

[0018] Step 3, via E total =E m,i,j +E 0,i,j The total thermoelectric potential after compensation is obtained;

[0019] Step 4: Substitute the interval function corresponding to the total thermoelectric potential and solve it using the Newton-Raphson iteration. Iteration accuracy |t (k+1) -t(k) If the temperature is less than 0.001℃, calculate the temperature at the measuring end. t (k) Let t be the temperature value in the k-th iteration. (k+1) Let f(t) be the temperature value in the (k+1)th iteration. (k) E represents the thermoelectric characteristic function value corresponding to the k-th temperature range. c，i，j To compensate for the total thermoelectric potential, f′(t) (k) Let be the first derivative of the thermoelectric characteristic function at temperature t^(k). It is the inverse function of the thermoelectric characteristic function, used to inversely calculate the temperature from the thermoelectric potential.

[0020] Preferably, in step 2, the basic compensation algorithm set includes three core algorithms:

[0021] Algorithm 1 is a nonlinear correction algorithm based on a BP neural network for t. c,i,j Temperature correction is applied, with the input being the thermoelectric potential-related parameter t. c,i,j The output correction value Δt1 is related to the ambient humidity h. Its training set uses calibration data in a high-temperature furnace, with a sample size of 1000 groups.

[0022] Algorithm 2, for drift compensation, employs a sliding window mean filter with a window size N = 50 to suppress long-term drift. Calculate the correction amount, where Δt2 is the historical average drift, and Δt2 is the temperature correction calculated by the drift compensation algorithm. Let be the compensated temperature of the thermocouple located in the i-th row and j-th column at time (t-nΔt);

[0023] Algorithm 3: For interference suppression, a soft threshold wavelet thresholding method based on db4 wavelet is used to remove the spike error caused by electromagnetic interference and obtain the correction amount Δt3. The fusion strategy adopts dynamic weight adaptive adjustment.

[0024] The dynamic weight fusion strategy includes the following methods:

[0025] Step 1: Define the instantaneous error ε of each algorithm through real-time error evaluation. k,i,j =|t c,i,j +Δt k -t ref |, where t ref ε is the reference temperature for the standard thermocouple with an accuracy of ±0.05℃ built into the high-temperature furnace. k,i,j The instantaneous error of the k-th algorithm thermocouple is Δt. k This is the correction amount for the k-th algorithm;

[0026] Step 2: Calculate the weights using a two-factor model that combines the inverse error and time decay, satisfying the following:

[0027]

[0028] Where Δt lag The algorithm response lag time is expressed in milliseconds (ms), and β = 0.02 is the attenuation coefficient. Algorithms with fast response and small error are preferred. k,i,j For the dynamic weighting coefficient of the k-th algorithm thermocouple position;

[0029] Step 3: Final Pass The temperature T after fusion is obtained f,i,j , where t f,i,j This is the final temperature of the thermocouple after being fused by multiple algorithms.

[0030] Preferably, step 3 includes the following method:

[0031] In the DIC vision and temperature data joint analysis module, spatial matching calibration is first performed. During the DIC stress element division stage, the specimen surface is photographed using a DIC camera, and a sub-pixel matching algorithm with an accuracy of 0.01 pixels is used to divide the specimen surface into mesh elements, which are then compared with the thermocouple array T. i,j One-to-one correspondence, the stress of the corresponding element is denoted as σ. i,j In the spatial coordinate calibration stage, a checkerboard calibration plate with an accuracy of 0.01mm was used to establish the DIC pixel coordinates (u). i,j ,v i,j ) and thermocouple physical coordinates (x i,j ,y i,j The transformation relation of ) satisfies:

[0032]

[0033] Where H is the homography matrix, which is solved by the Levenberg-Marquardt algorithm with a calibration error of <0.05mm.

[0034] Preferably, step 3 further includes the following method:

[0035] Next, temperature-stress nonlinear correlation modeling is performed. The algorithm used is a modified Gaussian process regression (GPR) combined with spatial regularization, with the input variable being the fused temperature t. f,i,j and the temperature gradient calculated by differentially analyzing adjacent thermocouple data. The output variable is the DIC measured stress σ. i,j (including normal stress σ) x σ y and shear stress τ xy Kernel function optimization uses a composite kernel function:

[0036] k(X p ,Xq )=k SE +k Matern

[0037] in The squared exponential kernel is used to capture smooth trends. The Maternal core is used to enhance adaptability to local mutations, σ m Let l be the signal variance and l be the length scale. In the spatial regularization operation, an adjacent element stress constraint term is added to the GPR objective function to penalize spatial discontinuity errors. The objective function satisfies:

[0038]

[0039] Where θ is the kernel function hyperparameter, λ is the regularization coefficient, and neighbors are the stresses of the four adjacent elements (top, bottom, left, and right) of element (i,j). In the stress prediction stage, for a new temperature sample... Predicted stress pass Calculate, where k * The kernel vectors of the new sample and the training sample. Solve using maximum likelihood estimation.

[0040] By adopting the above technical solution: First, at the hardware level, each thermocouple is equipped with an independent cold junction temperature sensing element, and a dedicated transmission cable with an outer heat insulation layer and an inner anti-interference layer is used. Then, combined with a segmented thermoelectric potential-temperature function fitting algorithm, cold junction temperature drift is suppressed from both hardware and algorithm aspects.

[0041] Based on this, a multi-algorithm fusion compensation module is constructed, integrating nonlinear correction, drift compensation and interference suppression functions, and accurately responding to various interference signals through a dynamic weighting strategy.

[0042] Furthermore, a sub-pixel-level matching algorithm and homography matrix calibration technique are employed to achieve high-precision spatial alignment of the temperature and stress measurement units. Finally, a nonlinear correlation between temperature and stress is established based on an improved Gaussian process regression (GPR) model, and spatial regularization constraints are introduced to improve modeling accuracy.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. This invention solves the problem of thermoelectric potential measurement error caused by the susceptibility of cold junction temperature to environmental interference in array thermocouples. It achieves accurate cold junction compensation by configuring an independent miniature platinum resistance thermometer for each thermocouple to collect cold junction temperature in real time and combining it with a piecewise thermoelectric potential-temperature function fitting algorithm.

[0045] 2. This invention addresses the problems of thermocouple nonlinearity, signal drift, and electromagnetic interference exacerbating temperature measurement deviations at high temperatures, as well as the insufficient adaptability of a single compensation algorithm. It constructs a basic algorithm set that includes nonlinearity correction, drift compensation, and interference suppression, and then integrates the various algorithms through a dynamic weight adaptive adjustment strategy to improve the accuracy of temperature measurement.

[0046] 3. This invention eliminates the spatial matching error between stress units and thermocouple temperature units in DIC visual inspection. The surface of the specimen is divided into grid units that correspond one-to-one with the thermocouple array by using a DIC camera, and spatial matching calibration is achieved by establishing coordinate transformation relationship using a calibration plate.

[0047] 4. This invention improves the problem of low accuracy caused by the strong nonlinear relationship between temperature and stress in traditional linear analysis. It adopts an improved Gaussian process regression combined with a composite kernel function and adds spatial regularization constraints to construct a nonlinear correlation model between temperature and stress, thereby improving the accuracy of the correlation analysis.

[0048] 5. This invention ensures the stability of signal transmission and processing in high-temperature environments through hardware and algorithm collaboration. It uses high-temperature heat-insulated data transmission cables to suppress electromagnetic interference, and integrates cold-end compensation, multi-algorithm fusion, and spatiotemporal coupling analysis to form an integrated process that is suitable for real-time monitoring needs in high-temperature scenarios. Attached Figure Description

[0049] Figure 1 This is a flowchart of the electronic cold junction compensation module in this invention;

[0050] Figure 2 This is a flowchart of the multi-algorithm fusion compensation module in this invention;

[0051] Figure 3 This is a flowchart of the DIC vision and temperature data joint analysis module in this invention;

[0052] Figure 4 This is a schematic diagram of the structural installation of an embodiment of the present invention;

[0053] Figure 5 This is the overall flowchart of the present invention. Detailed Implementation

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings, so that those skilled in the art can better understand the advantages and features of the present invention, thereby making a clearer definition of the scope of protection of the present invention. The embodiments described in this invention are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0055] A method for fusing multimodal compensation of array thermocouples with spatiotemporal coupling of DIC stress field includes the following steps:

[0056] Step 1: In the intelligent processing and compensation unit for thermal signals, first complete the hardware and algorithm co-design of the electronic cold junction compensation module;

[0057] Step 2: Construct a multi-algorithm fusion compensation module, integrating nonlinear correction, drift compensation and interference suppression functions, and accurately responding to various interference signals through a dynamic weighting strategy;

[0058] Step 3: Employ subpixel-level matching algorithm and homography matrix calibration technology to achieve high-precision spatial alignment of temperature and stress measurement units; finally, establish the nonlinear correlation between temperature and stress based on the improved Gaussian process regression (GPR) model, and introduce spatial regularization constraints to improve modeling accuracy.

[0059] Specifically, step 1 includes both hardware and algorithmic aspects. The hardware aspect consists of a 10×10 thermocouple array (denoted as T). i,j , representing the thermocouple in the i-th row and j-th column of the array. Each thermocouple in the array is equipped with an independent miniature platinum resistance thermometer with an accuracy of ±0.1℃, and is directly attached to the cold junction of the thermocouple to collect the cold junction temperature T0 in real time. At the same time, a high-temperature heat-insulating data transmission cable is used to transmit the thermoelectric potential signal and the cold junction temperature signal. The outer layer of the cable is wrapped with a ceramic fiber heat insulation layer with a temperature resistance of ≥1000℃, and the inner layer is made of twisted pair to suppress electromagnetic interference.

[0060] At the algorithm level, a piecewise thermoelectric potential-temperature function fitting is adopted. For K-type thermocouples with an applicable range of -200℃ to 1300℃, the temperature range is divided into five sub-intervals: [-200, 0), [0, 400), [400, 800), [800, 1000), and [1000, 1300]. Each sub-interval is fitted with a high-order polynomial thermoelectric characteristic function f using the least squares method. k (t)(k=1,2,…,5)(e.g., within a certain interval, E=a0+a1T+a2T) 2 +…+a n T n The function coefficients are calibrated using a thermocouple calibration table, and the fitting error is <0.05mV.

[0061] Specifically, such as Figure 1 As shown, in step 1, the cold junction compensation calculation includes the following methods:

[0062] Step 1: Measure the thermoelectric potential E corresponding to the temperature of the uncompensated measuring end. m,i,j =f(t) m,i,j )-f(t 0,i,j ), t m,i,j The temperature at the uncompensated measuring end; t0,i,j E represents the cold junction temperature of the thermocouple. m,i,j This represents the uncompensated thermoelectric potential of the thermocouple.

[0063] The second step is to calculate the cold junction thermoelectric potential E based on the interval where the cold junction temperature T0 falls, using the corresponding thermoelectric characteristic function for that interval. 0,i,j =f k (t 0,i,j According to t 0,i,j Choose f in the interval k ;f k Let be the thermoelectric characteristic function corresponding to the k-th temperature range;

[0064] Step 3, via E total =E m,i,j +E 0,i,j The total thermoelectric potential after compensation is obtained;

[0065] Step 4: Substitute the interval function corresponding to the total thermoelectric potential and solve it using the Newton-Raphson iteration. Iteration accuracy |t (k+1) -t (k) If the temperature is less than 0.001℃, calculate the temperature at the measuring end. t (k) Let t be the temperature value in the k-th iteration. (k+1) Let f(t) be the temperature value in the (k+1)th iteration. (k) E represents the thermoelectric characteristic function value corresponding to the k-th temperature range. c，i，j To compensate for the total thermoelectric potential, f′(t) (k) Let be the first derivative of the thermoelectric characteristic function at temperature t^(k). It is the inverse function of the thermoelectric characteristic function, used to inversely calculate the temperature from the thermoelectric potential.

[0066] Specifically, such as Figure 2 As shown, in step 2, the basic compensation algorithm set includes three core algorithms:

[0067] Algorithm 1 is a nonlinear correction algorithm based on a BP neural network for t. c,i,j Temperature correction is applied, with the input being the thermoelectric potential-related parameter t. c,i,j The output correction value Δt1 is related to the ambient humidity h. Its training set uses calibration data in a high-temperature furnace, with a sample size of 1000 groups.

[0068] Algorithm 2, for drift compensation, employs a sliding window mean filter with a window size N = 50 to suppress long-term drift. Calculate the correction amount, where Δt2 is the historical average drift; Δt2 is the temperature correction calculated by the drift compensation algorithm. Let be the compensated temperature of the thermocouple located in the i-th row and j-th column at time (t-nΔt);

[0069] Algorithm 3: For interference suppression, a soft threshold wavelet thresholding method based on db4 wavelet is used to remove the spike error caused by electromagnetic interference and obtain the correction amount Δt3. The fusion strategy adopts dynamic weight adaptive adjustment.

[0070] The dynamic weight fusion strategy includes the following methods:

[0071] Step 1: Define the instantaneous error ε of each algorithm through real-time error evaluation. k,i,j =|t c,i,j +Δt k -t ref |, where t ref The reference temperature for a standard thermocouple with an accuracy of ±0.05℃ built into the high-temperature furnace; ε k,i,j The instantaneous error of the k-th algorithm thermocouple is Δt. k This is the correction amount for the k-th algorithm;

[0072] Step 2: Calculate the weights using a two-factor model that combines the inverse error and time decay, satisfying the following:

[0073]

[0074] Where Δt lag The algorithm response lag time is expressed in milliseconds (ms), and β = 0.02 is the attenuation coefficient. Algorithms with fast response and small error are preferred. k,i,j For the dynamic weighting coefficient of the k-th algorithm thermocouple position;

[0075] Step 3: Final Pass The temperature T after fusion is obtained f,i,j , where t f,i,j This is the final temperature of the thermocouple after being fused by multiple algorithms.

[0076] Specifically, such as Figure 3 As shown, step 3 includes the following methods:

[0077] In the DIC vision and temperature data joint analysis module, spatial matching calibration is first performed. During the DIC stress element division stage, the specimen surface is photographed using a DIC camera, and a sub-pixel matching algorithm with an accuracy of 0.01 pixels is used to divide the specimen surface into mesh elements, which are then compared with the thermocouple array T. i,j One-to-one correspondence, the stress of the corresponding element is denoted as σ. i,j In the spatial coordinate calibration stage, a checkerboard calibration plate with an accuracy of 0.01mm was used to establish the DIC pixel coordinates (u). i,j ,v i,j ) and thermocouple physical coordinates (xi,j ,y i,j The transformation relation of ) satisfies:

[0078]

[0079] Where H is the homography matrix, which is solved by the Levenberg-Marquardt algorithm with a calibration error of <0.05mm.

[0080] Specifically, step 3 also includes the following methods:

[0081] Next, temperature-stress nonlinear correlation modeling is performed. The algorithm used is a modified Gaussian process regression (GPR) combined with spatial regularization, with the input variable being the fused temperature t. f,i,j and the temperature gradient calculated by differentially analyzing adjacent thermocouple data. The output variable is the DIC measured stress σ. i,j (including normal stress σ) x σ y and shear stress τ xy Kernel function optimization uses a composite kernel function:

[0082] k(X p ,X q )=k SE +k Matern

[0083] in The squared exponential kernel is used to capture smooth trends. The Maternal core is used to enhance adaptability to local mutations, σ m Let l be the signal variance and l be the length scale. In the spatial regularization operation, an adjacent element stress constraint term is added to the GPR objective function to penalize spatial discontinuity errors. The objective function satisfies:

[0084]

[0085] Where θ is the kernel function hyperparameter, λ is the regularization coefficient, and neighbors are the stresses of the four adjacent elements (top, bottom, left, and right) of element (i,j). In the stress prediction stage, for a new temperature sample... Predicted stress pass Calculate, where k * The kernel vectors of the new sample and the training sample. Solve using maximum likelihood estimation.

[0086] Example

[0087] This embodiment focuses on the thermo-mechanical coupling testing of high-temperature alloy molds, and determines the actual product technical solution as follows:

[0088] like Figure 4 As shown, the hardware system is first built and deployed. In this step, the test piece 6 is installed in the high-temperature environment heating furnace 5 using a high-temperature precision clamping and fixing mechanism 4. Each of the 10×10 array thermocouples (each thermocouple has an independent number, forming a high-temperature multi-channel array thermocouple sensing device 3) installed on the test piece is equipped with an independent miniature platinum resistance thermometer with an accuracy of ±0.1℃, ensuring that the platinum resistance thermometer is directly and tightly attached to the cold junction of the thermocouple to collect the cold junction temperature in real time. At the same time, a high-temperature heat-insulating data transmission cable 2 with an outer layer of ceramic fiber heat insulation layer with a temperature resistance of ≥1000℃ and an inner layer of twisted pair is selected to connect the high-temperature multi-channel array thermocouple sensing device 3 and the thermal signal intelligent processing and compensation unit 1 (integrating electronic cold junction compensation and multi-algorithm fusion compensation functions) to achieve stable transmission of thermoelectric potential signal and cold junction temperature signal, and to suppress electromagnetic interference through the twisted pair structure. In addition, a DIC visual inspection camera 7 (used to photograph the surface of the test piece and conduct digital image correlation analysis) is set up at a suitable location outside the heating furnace to clearly photograph the surface of the test piece 6. A checkerboard calibration plate with an accuracy of 0.01mm is prepared, and the thermal signal intelligent processing and compensation unit 1 and the DIC visual inspection camera 7 are connected to the data acquisition and analysis computer 8 to complete the overall hardware link construction, laying the foundation for subsequent data acquisition and processing.

[0089] Next, the intelligent thermal signal processing and compensation unit 1 is activated, first executing the electronic cold junction compensation process. For the applicable range of the K-type thermocouple from -200℃ to 1300℃, it is divided into five sub-intervals: [-200, 0), [0, 400), [400, 800), [800, 1000), and [1000, 1300]. Each sub-interval is fitted with a high-order polynomial thermoelectric characteristic function using the least squares method (fitting error < 0.05mV), and the function coefficients need to be calibrated in advance according to the thermocouple calibration table. Then, the thermoelectric potential corresponding to the uncompensated measuring junction temperature of the thermocouple in the high-temperature multi-channel array thermocouple sensing device 3 is measured. Based on the cold junction temperature collected by the platinum resistance thermometer, its sub-interval is determined, and the corresponding fitting function is selected to calculate the cold junction thermoelectric potential, thus obtaining the compensated total thermoelectric potential (total thermoelectric potential = measured thermoelectric potential + cold junction thermoelectric potential). Finally, the measuring junction temperature is calculated back using an iterative formula (the iteration accuracy meets the detection requirements), completing the electronic cold junction compensation.

[0090] After the electronic cold junction compensation is completed, the multi-algorithm fusion compensation module of the thermal signal intelligent processing and compensation unit 1 is activated. First, the basic compensation algorithm set is called: a BP neural network based on 1000 sets of calibration data from a high-temperature furnace as the training set, using the temperature and ambient humidity h after electronic cold junction compensation as input, to calculate the nonlinear correction amount; a sliding window mean filter with a preset window size is used, combined with historical average drift, to obtain the drift compensation correction amount; the db4 wavelet soft threshold denoising method is used to remove spike errors caused by electromagnetic interference, obtaining the interference suppression correction amount. Then, using the reference temperature of the standard thermocouple with an accuracy of ±0.05℃ built into the high-temperature ambient heating furnace 5 as a benchmark, the instantaneous error of each algorithm is calculated; based on the "error reciprocal - time decay" two-factor model (where is the algorithm response lag time (ms) and is the decay coefficient), the weights of each algorithm are calculated (the sum of the weights is 1), and finally, the fused temperature is obtained according to the fused temperature formula.

[0091] Finally, joint analysis of the DIC visual inspection camera 7 and temperature data was conducted. The first step was spatial matching calibration: the surface of the test piece 6 was photographed by the DIC visual inspection camera 7, and the surface of the test piece was divided into 10×10 grid units using a sub-pixel matching algorithm with an accuracy of 0.01 pixels, ensuring a one-to-one correspondence with the thermocouple array of the high-temperature multi-channel array thermocouple sensing device 3; using a checkerboard calibration plate, the homography matrix was solved using the Levenberg-Marquardt algorithm to establish the conversion relationship between DIC pixel coordinates and thermocouple physical coordinates (calibration error <0.05mm), thus completing the spatial matching calibration. The second step involves constructing a temperature-stress nonlinear correlation model: An improved Gaussian process regression (GPR) algorithm with spatial regularization is selected. The input variables are the fused temperature and the temperature gradient calculated through the difference of adjacent thermocouple data. The output variables are the stress data containing normal and shear stress measured by the DIC visual inspection camera 7. A composite kernel function consisting of a squared exponential kernel (capturing smooth trends) and a Matern kernel (ν = 3 / 2, enhancing adaptability to local abrupt changes) is used. Adjacent element stress constraint terms (where is the kernel function hyperparameter, is the regularization coefficient, and represents the four adjacent elements in the top, bottom, left, and right directions) are added to the GPR objective function to penalize spatial discontinuity errors. For new temperature samples, stress prediction is performed based on the formula for predicting stress (the noise variance is solved through maximum likelihood estimation). Ultimately, the spatiotemporal coupling and fusion analysis of temperature and stress data is achieved, obtaining high-precision temperature-stress correlation data for the test specimen.

[0092] The present invention proposes a hardware and algorithm-based electronic cold junction compensation technology. Its core lies in independently configuring a high-precision cold junction temperature acquisition element for each thermocouple in the array to obtain the cold junction temperature in real time. At the algorithm level, it innovatively employs a piecewise thermoelectric potential-temperature function fitting method, dividing the wide temperature range into multiple sub-intervals and performing high-precision polynomial fitting to achieve accurate mathematical compensation for cold junction temperature drift. The advantage of this technology is that it effectively suppresses cold junction thermoelectric potential measurement errors caused by environmental factors from the signal source, significantly improving the reliability of the basic temperature data.

[0093] To address the bottleneck of insufficient adaptability of single compensation algorithms under complex high-temperature conditions, this invention constructs a multi-algorithm fusion compensation mechanism. The key to this technology lies in integrating a set of fundamental algorithms, including nonlinear correction, drift compensation, and interference suppression, and designing a dynamic weight adaptive adjustment strategy based on real-time performance evaluation to intelligently fuse the outputs of each algorithm. Its advantages lie in its ability to comprehensively address multiple error sources such as thermocouple nonlinearity, signal drift, and electromagnetic interference. Through the complementary strengths of the algorithms, it significantly improves the overall accuracy of temperature measurement and the system's robustness under transient changes.

[0094] This invention proposes a high-precision spatial matching calibration technique. The key to this technique lies in using a pixel-level image matching algorithm to establish a one-to-one correspondence between the surface grid of the specimen measured by DIC and the thermocouple array. By introducing a calibration plate and employing an optimization algorithm to solve the coordinate transformation relationship, an unbiased transformation from the pixel coordinate system to the physical coordinate system is achieved. Its advantage is that it fundamentally eliminates the spatial registration error caused by the separation of the measurement system, ensuring the consistency of temperature and stress data in spatial location, and laying a solid foundation for subsequent spatiotemporal coupling analysis.

[0095] To delve deeper into the intrinsic relationship between temperature and stress, this invention employs a nonlinear correlation modeling technique based on Gaussian process regression (GPR). The key innovation of this technique lies in constructing a composite kernel function with different characteristics to better capture the global trend and local abrupt changes in the temperature-stress relationship. Simultaneously, spatial regularization constraints are introduced into the model's objective function, forcing stress prediction results in adjacent regions to remain smooth and continuous. The advantage of this approach is that it overcomes the limitations of traditional linear models in describing complex thermo-mechanical coupling phenomena. Through a nonlinear, non-parametric modeling method, combined with spatial prior knowledge, it significantly improves the accuracy and physical plausibility of stress field prediction.

[0096] In summary, this invention focuses on the collaborative design of a thermal signal intelligent processing and compensation unit and a DIC vision and temperature data joint analysis module. Through hardware optimization and algorithm innovation, it solves the core accuracy problem of temperature and stress detection under high-temperature environments.

[0097] The descriptions and practices disclosed in this invention are readily apparent and understandable to those skilled in the art, and various modifications and refinements can be made without departing from the principles of this invention. Therefore, any modifications or improvements made without departing from the spirit of this invention should also be considered within the scope of protection of this invention.

Claims

1. A method for array thermocouple multi-modal compensation and DIC stress field space-time coupling fusion, characterized in that, Comprising the following steps: Step 1: In the intelligent processing and compensation unit of thermal signals, the hardware and algorithm collaborative design of the electronic cold-end compensation module is completed; Step 2: A multi-algorithm fusion compensation module is constructed, integrating nonlinear correction, drift compensation and interference suppression functions, and accurately responding to various interference signals through a dynamic weighting strategy; Step 3: Sub-pixel level matching algorithm and homography matrix calibration technology are used to realize high-precision spatial alignment of the temperature and stress measurement unit; finally, an improved Gaussian process regression (GPR) model is established to establish the nonlinear relationship between temperature and stress, and spatial regularization constraints are introduced to improve the modeling accuracy.

2. The method of claim 1, wherein, In step 1, both hardware and algorithm are included, where the hardware is a 10x10 thermocouple array, and each thermocouple in the array is configured with an independent micro platinum resistor; the algorithm uses a segmented thermoelectric potential-temperature function fitting.

3. The method of claim 1, wherein, In step 1, the cold-end compensation calculation includes the following methods: First step, measure the thermoelectric force E corresponding to the uncompensated measuring end temperature m,i,j = f(t m,i,j ) - f(t 0,i,j ), t m,i,j is the uncompensated measuring end temperature; t 0,i,j is the cold end temperature of the thermocouple; E m,i,j is the uncompensated thermoelectric force of the f thermocouple; Second step, according to the interval where the cold end temperature T0 is located, the corresponding interval of thermoelectric characteristic function is called to calculate the cold end thermoelectric potential E 0,i,j = f k (t 0,i,j ), according to the interval where t 0,i,j is located, f k is selected; f k is the thermoelectric characteristic function corresponding to the kth temperature interval; Third step, by E total = E m,i,j + E 0,i,j resulting compensated total thermoelectric power; Fourth step, substitute the interval function corresponding to the total thermoelectric potential, and solve by Newton-Raphson iteration Iteration precision |t (k+1) -t (k) |<0.001℃, back-calculate the measured end temperature t (k) Temperature value of the kth iteration, t (k+1) Temperature value of the k+1th iteration, f(t (k) ) is the thermoelectric characteristic function value corresponding to the kth temperature interval, E c,i,j Total thermoelectric potential after compensation, f'(t (k) ) is the first-order derivative of the thermoelectric characteristic function at temperature t^(k), Inverse function of the thermoelectric characteristic function, used to back-calculate the temperature according to the thermoelectric potential.

4. The method of claim 1, wherein, In step 2, the basic compensation algorithm set contains three types of core algorithms: Algorithm 1, for nonlinear correction, based on BP neural network to t c,i,j temperature correction, input is thermoelectric potential parameter t c,i,j and ambient humidity h, output correction amount Δt1, its training set uses high-temperature furnace calibration data, sample amount reaches 1000 groups; Algorithm 2, for drift compensation, uses a sliding window mean filter with window size N = 50 to suppress long term drift, and the temperature correction is calculated by where Δt2is the temperature correction calculated by the drift compensation algorithm, is the historical average drift, Δt2is the temperature correction calculated by the drift compensation algorithm, is the compensated temperature of the thermocouple at time (t - nΔt) located at the i-th row and j-th column. Algorithm 3: For interference suppression, a soft threshold wavelet threshold denoising method based on db4 wavelet is used to remove the spike errors caused by electromagnetic interference to obtain the correction Δt3, and a dynamic weight adaptive adjustment is used for fusion strategy; Where the dynamic weight fusion strategy includes the following methods: Step 1: Define the instantaneous error ε of each algorithm by real-time error evaluation k,i,j = |t c,i,j + Δt k - t ref |, where t ref is the reference temperature of the standard thermocouple with built-in precision of ±0.05℃ in the high-temperature furnace; ε k,i,j is the instantaneous error of the kth algorithm thermocouple; Δt k is the correction amount of the kth algorithm Step 2: Calculate the weight through the error reciprocal-time decay double factor model, which satisfies: where Δt lag is the algorithm response lag time, unit: ms, β = 0.02 is the attenuation coefficient k,i,j is the dynamic weight coefficient of the thermocouple position of the kth algorithm; Step3: Final pass Obtain fusion temperature T f,i,j Where t f,i,j is the final temperature of the thermocouple after the multi-algorithm fusion.

5. The method of claim 1, wherein, In step 3, the following methods are included: In the DIC vision and temperature data joint analysis module, firstly, spatial matching calibration is carried out, DIC stress unit division stage, through DIC camera shooting test piece surface, using the sub-pixel level matching algorithm with accuracy of 0.01 pixels, the test piece surface is divided into grid units, and the thermal couple array T i,j One-to-one correspondence, the stress of the corresponding unit is denoted as σ i,j ; the spatial coordinate calibration link uses the chessboard calibration plate with accuracy of 0.01 mm, and the conversion relationship between the DIC pixel coordinates (u i,j ,v i,j ) and the thermal couple physical coordinates (x i,j ,y i,j ) is established, and the following is met: Where H is the homography matrix, which is solved by the Levenberg-Marquardt algorithm, and the calibration error is <0.05mm.

6. The method of claim 1, wherein, In step 3, the following methods are also included: Then the temperature-stress nonlinear correlation modeling is carried out, the algorithm selects the improved Gaussian process regression GPR combined with spatial regularization, the input variable is the fusion temperature t f,i,j And the temperature gradient calculated by the difference of adjacent thermocouple data The output variable is the DIC measured stress σ i,j The kernel function optimization adopts a composite kernel function: k(X p ,X q ) = k SE +k Matern where is a squared exponential kernel, capturing smooth trends, is a Matern kernel, enhancing adaptability to local mutations, σ m is the signal variance, l is the length scale; in the spatial regularization operation, a stress constraint term of adjacent units is added in the GPR objective function, which punishes the spatial discontinuity error, and the objective function satisfies: where θ is the kernel hyper-parameter, λ is the regularization coefficient, neighbors are the stresses of the four neighboring cells above, below, left and right of the (i, j) cell, in the stress prediction phase, for a new temperature sample predicted stress by where k * is the kernel vector of the new sample and the training samples, solved by maximum likelihood estimation.

Images