A thermoelectric material performance detection system based on thermoelectric textiles

Abstract

The application relates to the technical field of a thermoelectric material performance detection system based on a thermoelectric textile, and particularly discloses a thermoelectric material performance detection system based on a thermoelectric textile. The system comprises a temperature control loading platform, an electric heating signal acquisition unit, a computer vision strain sensing unit, an adaptive deformation compensation module and a central processing unit; the temperature control loading platform applies a controllable temperature gradient and contact pressure, the electric heating signal acquisition unit synchronously acquires voltage and current signals, the computer vision unit analyzes full-field strain through image sequences, the adaptive deformation compensation module dynamically corrects the electric heating signal in combination with a physical information neural network, and accurate Seebeck coefficients and electric conductivities are output. Through multi-modal sensing and physical constraint model fusion, the application realizes high-fidelity characterization of the performance of the easily-deformable thermoelectric textile under dynamic load, and improves test accuracy and repeatability.

Description

Technical Field

[0001] This invention belongs to the field of thermoelectric material performance testing and intelligent textile testing technology, specifically relating to a thermoelectric material performance testing system based on thermoelectric textiles. Background Technology

[0002] With the rapid development of wearable electronic devices and human body heat recovery technology, thermoelectric textiles, as a novel flexible material capable of converting environmental heat energy into electrical energy, have shown great application potential in the fields of smart wearables and mobile energy management. By combining high-performance semiconductor components with a flexible fiber substrate, thermoelectric textiles possess thermoelectric conversion capabilities while maintaining the fabric's lightweight and breathable properties, making them a key branch of current flexible electronics research. In materials science research and industrial applications, accurately characterizing the electrical and thermal properties of thermoelectric textiles is a core prerequisite for evaluating their energy conversion efficiency and reliability.

[0003] Thermoelectric material performance testing systems, as specialized tools for evaluating material quality, primarily calculate the Seebeck coefficient and conductivity by acquiring open-circuit voltage and current signals of samples at specific temperature differences. These systems typically integrate sophisticated temperature control units and weak signal processing modules to simulate and record the material's response curves under real-world operating conditions. To meet the growing demand for flexible testing, existing performance testing systems are gradually exploring how to achieve high-precision measurements under dynamic environments, aiming to provide accurate data support for optimizing the performance of flexible thermoelectric materials.

[0004] Traditional testing systems are mostly based on static testing logic for rigid samples, making them incompatible with the flexible and porous characteristics of thermoelectric textiles. In actual testing processes, the physical pressure applied by the testing device can easily cause irreversible physical collapse of the textile's microstructure, leading to drastic fluctuations in contact resistance and resulting in severe distortion of the measured Seebeck coefficient and conductivity. Because existing technologies lack real-time monitoring methods for sample morphology changes, it is impossible to distinguish whether performance errors originate from the material itself or from mechanical deformation during the testing process. The system lacks analytical logic for the complex nonlinear relationship between deformation state and electrothermal performance, making it difficult to achieve adaptive compensation of measurement results under dynamic strain conditions. Therefore, a thermoelectric material performance testing system based on thermoelectric textiles is desired. Summary of the Invention

[0005] The purpose of this invention is to provide a thermoelectric material performance testing system based on thermoelectric textiles, which can solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A thermoelectric material performance testing system based on thermoelectric textiles includes a temperature-controlled loading platform, an electrothermal signal acquisition unit, a computer vision strain sensing unit, an adaptive deformation compensation module, and a central processing unit.

[0008] The temperature-controlled loading platform is configured to apply a controllable temperature gradient to the thermoelectric textile sample and provide adjustable contact pressure during testing to simulate the mechanical load environment that may occur in actual use.

[0009] The electrothermal signal acquisition unit is configured to simultaneously acquire the open-circuit voltage signal and short-circuit current signal generated by the thermoelectric textile under the action of temperature difference, and transmit the acquired raw electrical signal to the central processing unit.

[0010] The computer vision strain sensing unit is configured to acquire high-resolution image sequences of the surface of thermoelectric textiles in real time during the test, and analyze the full-field geometric deformation information of the sample under pressure based on the digital image association algorithm, including the local strain distribution and the degree of microstructure collapse.

[0011] The adaptive deformation compensation module is configured to receive deformation data from the computer vision strain sensing unit and raw electrothermal signals from the electrothermal signal acquisition unit. Combined with the embedded physical information neural network model, it dynamically corrects the changes in contact resistance and electrothermal performance distortion caused by deformation, and outputs the compensated Seebeck coefficient and conductivity.

[0012] The central processing unit is configured to coordinate the working sequence of each component, perform data fusion, model inference and result output, and provide the final performance evaluation report to the user interface.

[0013] Preferably, the temperature-controlled loading platform includes an upper heating plate and a lower cooling plate, with a pressure adjustment mechanism between them. This pressure adjustment mechanism can dynamically adjust the vertical clamping force on the thermoelectric textile according to preset parameters to ensure stable thermal contact without damaging the microstructure of the sample.

[0014] Furthermore, the electrothermal signal acquisition unit includes a high-precision voltmeter and a nanoampere-level ammeter, and its sampling frequency is synchronized with the image frame rate of the computer vision strain sensing unit to achieve time alignment between the electrothermal response and the deformation state.

[0015] Furthermore, the computer vision strain sensing unit includes a high-resolution industrial camera and a ring light source array. The industrial camera is positioned facing the test area of ​​the thermoelectric textile and can continuously capture changes in the speckle pattern on the sample surface under non-contact conditions, and calculate the two-dimensional or three-dimensional strain field through a digital image association algorithm.

[0016] Preferably, the physical information neural network model embedded in the adaptive deformation compensation module is trained based on a large amount of experimental data. Its input variables include local strain tensor, contact pressure value and original electrothermal signal, and the output variables are the modified Seebeck coefficient and conductivity. The physical information neural network model introduces thermoelectric transport equation as physical constraint during the training process to ensure that the output results conform to the basic laws of thermodynamics.

[0017] Furthermore, the physical information neural network model can identify sudden changes in contact area caused by the collapse of the porous structure of textiles, and dynamically adjust the correction weight of conductivity accordingly to distinguish between the degradation of intrinsic material properties and measurement deviations caused by poor test contact.

[0018] Furthermore, the central processing unit is also equipped with a data storage module and a visualization interface, which can structure and store the raw data, deformation field images, performance parameters before and after compensation, and confidence index of each test, and support users to conduct retrospective analysis and horizontal comparison.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] 1. The thermoelectric material performance testing system based on thermoelectric textiles provided by this invention, through the integration of computer vision strain perception and a physical information neural network-driven adaptive deformation compensation mechanism, achieves for the first time high-fidelity performance characterization of flexible, porous, and easily deformable thermoelectric textiles under dynamic loading conditions. The system can monitor the microstructural changes of the sample in real time during the testing process and accurately correct the distortion of electrothermal signals caused by poor contact or structural collapse based on physical laws, distinguishing between the intrinsic properties of the material and testing interference factors.

[0021] 2. This system breaks through the limitations of the traditional rigid sample testing paradigm, improves the accuracy and repeatability of performance evaluation of thermoelectric textiles, and provides reliable technical support for the research and development, process optimization and industrial quality control of flexible thermoelectric materials. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall technical solution architecture of a thermoelectric material performance testing system based on thermoelectric textiles proposed in this invention;

[0023] Figure 2 This is a schematic diagram of the core principle framework of adaptive deformation compensation based on physical information neural network in this invention;

[0024] Figure 3 This is a logical flowchart of the synchronous acquisition of electrothermal signals and the full-field strain perception by computer vision in this invention.

[0025] Figure 4This is a schematic diagram of the multi-level interaction relationship and data flow between the temperature control loading, sensor acquisition and model compensation modules in this invention;

[0026] Figure 5 This is a flowchart illustrating the logical flow of the present invention, which uses a physical information neural network to identify structural collapse in textiles and dynamically adjusts the conductivity correction weights. Detailed Implementation

[0027] Example 1: Please refer to the appendix Figure 1 To be continued Figure 5 To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0028] A thermoelectric material performance testing system based on thermoelectric textiles includes a temperature-controlled loading platform, an electrothermal signal acquisition unit, a computer vision strain sensing unit, an adaptive deformation compensation module, and a central processing unit.

[0029] The temperature-controlled loading platform is configured to apply a controllable temperature gradient to the thermoelectric textile sample and provide adjustable contact pressure during testing to simulate the mechanical load environment that may occur in actual use. The temperature-controlled loading platform includes a precision temperature control component and a pressure mechanical component. The precision temperature control component is configured to use a semiconductor cooling chip or resistance wire heating array as a heat source, and transfer heat energy to both ends of the thermoelectric textile sample through oxygen-free copper material with high thermal conductivity to build a stable and uniformly distributed temperature field. The pressure mechanical component is configured to apply a vertical pressure load to the surface of the thermoelectric textile sample through a precision ball screw mechanism driven by a servo motor, and the magnitude of the pressure load can be steplessly adjusted between 0.1 Newtons and 500 Newtons according to a preset test protocol.

[0030] The temperature-controlled loading platform integrates a high-sensitivity pressure sensor, which, together with the servo motor's control driver, forms a closed-loop control circuit. This circuit dynamically compensates for pressure fluctuations caused by thermal expansion or textile structural collapse during testing, ensuring constant contact pressure. Both the upper heating plate and lower cooling plate of the temperature-controlled loading platform are chemically nickel-plated to reduce heat radiation loss and improve heat exchange efficiency with the textile surface.

[0031] The electrothermal signal acquisition unit is configured to simultaneously acquire the open-circuit voltage and short-circuit current signals generated by the thermoelectric textile under temperature difference, and transmit the acquired raw electrical signals to the central processing unit. The electrothermal signal acquisition unit integrates a high-input-impedance nanovoltmeter and a nanoampere-level low-noise ammeter. Its analog front-end circuit is connected to the electrode contacts of the thermoelectric textile sample via a multiplexer. The electrode contacts employ a gold-plated copper spring pin array to ensure reliable electrical contact even when the textile undergoes deformation.

[0032] The electrothermal signal acquisition unit also includes a precision synchronization trigger module, which is configured to send a pulse synchronization signal to the computer vision strain sensing unit. Whenever the electrothermal signal acquisition unit completes an analog-to-digital conversion of voltage or current data, the pulse synchronization signal immediately triggers the computer vision strain sensing unit to capture an image, ensuring precise alignment of the electrothermal performance parameters with the geometric deformation state of the sample on a nanosecond-level timescale. Furthermore, the electrothermal signal acquisition unit is equipped with a three-stage active filter to filter out 50 Hz power frequency interference from the environment and thermoelectric noise generated by the temperature control mechanism.

[0033] The computer vision strain sensing unit is configured to acquire high-resolution image sequences of the thermoelectric textile surface in real time during testing, and analyze the full-field geometric deformation information of the sample under pressure based on a digital image association algorithm, including local strain distribution and the degree of microstructural collapse. The computer vision strain sensing unit includes an industrial-grade high-resolution area array camera, a telephoto macro lens, and a ring LED light source array. The industrial-grade high-resolution area array camera has a pixel scale of no less than 20 million pixels and is mounted directly above the temperature-controlled loading platform, with its optical axis perpendicular to the stress plane of the thermoelectric textile sample.

[0034] The computer vision strain sensing unit internally runs a full-field strain analysis algorithm module. This module is configured to first perform grayscale processing and Gaussian filtering denoising on the acquired raw image, then define a series of non-overlapping sub-regions in the image, and use an iterative algorithm based on the least squares criterion to find the optimal matching position of these sub-regions before and after deformation. By calculating the displacement gradient of the centroid coordinates of the sub-regions, the full-field strain analysis algorithm module can generate a two-dimensional strain tensor field covering the entire sample surface. The strain tensor field contains the tensile strain, compressive strain, and shear strain data of the thermoelectric textile in the horizontal and vertical directions. In addition, the algorithm module can also identify abrupt changes in the grayscale gradient in the image to determine whether physical structural collapse has occurred in the fiber bundles inside the textile.

[0035] The adaptive deformation compensation module is configured to receive deformation data from the computer vision strain sensing unit and raw electrothermal signals from the electrothermal signal acquisition unit. Combined with the embedded physical information neural network model, it dynamically corrects the changes in contact resistance and electrothermal performance distortion caused by deformation, and outputs the compensated Seebeck coefficient and conductivity. The physical information neural network model embedded in the adaptive deformation compensation module is a deep learning architecture. Its input vector includes the local strain tensor components at the current moment, the currently applied contact pressure value, the measured temperature difference distribution data, and the original electrical signal sampling value.

[0036] The physical information neural network model is configured to introduce fundamental physical constraints on thermoelectric transport into the loss function. These constraints require that the output correction parameters strictly adhere to the linear coupling relationship between the law of energy conservation and the Seebeck effect; that is, the corrected output must satisfy the vector correlation characteristic between electric field strength and temperature gradient. The model is configured to identify spurious conductivity gains caused by reduced porosity in textiles through deep learning of historical experimental data, and to calculate a correction operator to counteract these spurious gains using a nonlinear mapping function. This correction operator is a real-number vector that, through weighted multiplication with the original measurement data, eliminates measurement system errors introduced by mechanical deformation.

[0037] The central processing unit (CPU) is configured to coordinate the working timing of various components, perform data fusion, model inference, and result output, and provide a final performance evaluation report to the user interface. The CPU is implemented based on a high-performance industrial controller, and its core is configured to run a multi-threaded real-time operating system. The CPU is responsible for managing the PID control parameters of the temperature-controlled loading platform, ensuring that the absolute value of temperature field fluctuations is less than 0.05 degrees Celsius.

[0038] The central processing unit (CPU) internally operates a data fusion center configured to align the discrete electrical signal stream provided by the electrothermal signal acquisition unit with the continuous strain field image stream provided by the computer vision strain sensing unit in a spatial coordinate system. By topologically mapping the high-strain regions in the strain field to the electrode contact regions, the data fusion center can quantitatively assess the contact resistance variation trend at each contact point. The CPU is also connected to a high-capacity solid-state storage medium to store the original RAW format image data, timestamp sequences, and high-precision performance indicators processed by the adaptive deformation compensation module generated for each test. The user interface is configured to render dynamic strain thermograms and performance parameter evolution curves of the sample in real time.

[0039] Furthermore, the temperature-controlled loading platform includes an upper heating plate and a lower cooling plate, with a pressure regulating mechanism between them. This mechanism dynamically adjusts the vertical clamping force on the thermoelectric textile according to preset parameters, ensuring stable thermal contact without damaging the sample's microstructure. A flexible heat flow meter is attached to the bottom surface of the upper heating plate to monitor the heat flow entering the thermoelectric textile sample and assist in calculating the material's thermal conductivity. The pressure regulating mechanism employs a closed-loop force control algorithm. When it detects changes in internal tension caused by thermal shrinkage of the textile, the mechanism automatically adjusts the screw displacement to keep the total normal pressure acting on the sample within a preset error range.

[0040] Furthermore, the electrothermal signal acquisition unit includes a high-precision voltmeter and a nanoampere-level ammeter, whose sampling frequency is synchronized with the image frame rate of the computer vision strain sensing unit to achieve time alignment between the electrothermal response and the deformation state. The sampling frequency is set to no less than 60 times per second to capture the transient electrical response process of the textile during compression. To eliminate the influence of lead resistance on conductivity measurement, the electrothermal signal acquisition unit is configured to use a four-wire measurement method, i.e., injecting a known current through one pair of independent leads and acquiring the potential difference through another pair of independent leads.

[0041] Furthermore, the computer vision strain sensing unit includes a high-resolution industrial camera and a ring light source array. The industrial camera faces the test area of ​​the thermoelectric textile and can continuously capture changes in the speckle pattern on the sample surface under non-contact conditions, and calculate the two-dimensional or three-dimensional strain field through a digital image association algorithm. The ring light source array is configured to use narrow-spectrum monochromatic light illumination, and with the filter in front of the camera lens, it can shield against background interference from natural light in the laboratory and thermal radiation from the temperature control platform. Before the test begins, the surface of the thermoelectric textile sample is pre-sprayed with a uniformly distributed submicron-level random speckle pattern, which serves as feature points for subpixel displacement tracking by the computer vision algorithm.

[0042] Furthermore, the physical information neural network model embedded in the adaptive deformation compensation module is trained based on a large amount of experimental data. Its input variables include the local strain tensor, contact pressure value, and original electrothermal signal, while the output variables are the corrected Seebeck coefficient and conductivity. During the training process, the thermoelectric transport equation is introduced as a physical constraint to ensure that the output results conform to the fundamental laws of thermodynamics. Specifically, the thermoelectric transport equation is transformed into a loss function term in the neural network. This loss function term calculates the entropy production rate of the output parameters under known temperature difference conditions, forcibly constraining the model to prevent it from outputting parameter combinations that violate the second law of thermodynamics.

[0043] Furthermore, the physical information neural network model can identify abrupt changes in contact area caused by the collapse of the porous structure of textiles, and dynamically adjust the correction weight of conductivity accordingly, distinguishing between intrinsic material performance degradation and measurement deviations caused by poor test contact. The identification process is configured such that when the local strain value output by the computer vision strain sensing unit exceeds a preset structural integrity threshold, the model automatically triggers a nonlinear compensation factor. This nonlinear compensation factor is configured to reduce the contribution of the current measured conductivity value to the final performance evaluation, replacing it with an interpolated predicted value based on physical laws, thus eliminating abrupt interference caused by microfiber bundle breakage or overlap.

[0044] Furthermore, the central processing unit is also equipped with a data storage module and a visualization interface, which can structurally store the raw data, deformation field images, performance parameters before and after compensation, and confidence indices of each test, and support user retrospective analysis and horizontal comparison. The visualization interface is configured to support three-dimensional view switching, allowing users to intuitively observe how the actual Seebeck coefficient deviates from the ideal physical model as the degree of compression of the textile increases during the pressure loading process.

[0045] In this embodiment, the system first activates the temperature-controlled loading platform during operation to achieve a preset temperature difference equilibrium between the upper heating plate and the lower cooling plate. The central processing unit applies a stepped, progressively increasing pressure load to the thermoelectric textile sample by controlling the pressure regulating mechanism. Under each pressure load level, the electrothermal signal acquisition unit and the computer vision strain sensing unit acquire the sample's electrical parameters and surface image with sub-millisecond synchronization accuracy, respectively.

[0046] The acquired data is pushed to the adaptive deformation compensation module in real time. This module uses an embedded physical information neural network to nonlinearly decouple the data, removing interference components introduced by the reduction in textile thickness, porosity, and contact resistance. The central processing unit outputs thermoelectric performance indicators corrected by physical constraints, ensuring that the intrinsic physical properties of the material are reflected under any deformation state.

[0047] Example 2: As another alternative to Example 1, this example provides a distributed thermoelectric material performance testing system based on an edge computing architecture. In this architecture, the system's computational load is rationally distributed among multiple field edge computing nodes and the central control server to improve the system's real-time performance and stability during large-scale parallel testing of samples.

[0048] A thermoelectric material performance testing system based on thermoelectric textiles includes a multi-station temperature-controlled loading array, a distributed electrothermal signal acquisition cluster, a multi-angle computer vision perception module, an edge adaptive compensation node, and a cloud processing platform.

[0049] The multi-station temperature-controlled loading array is configured to consist of several independent temperature control units, each with its own independent temperature control algorithm logic. This parallel design allows the system to simultaneously evaluate the performance of multiple thermoelectric textile samples of different materials or weaving techniques. Each temperature control unit employs a miniaturized thermopile design, enabling extremely fast heating and cooling rates. Its control commands are issued via a central bus and support independent pressure parameter configuration to meet diverse experimental needs.

[0050] The distributed electrothermal signal acquisition cluster comprises multiple signal processing terminals deployed near the test stations. Each terminal integrates a multi-channel analog-to-digital converter and is configured to digitize the thermoelectric signals using high bit depth sampling technology. By digitizing near the signal source, the distributed cluster reduces attenuation caused by long-distance analog transmission and external electromagnetic noise. Each terminal transmits the digitized voltage and current signals to the corresponding edge adaptive compensation node via a high-speed Ethernet interface.

[0051] The multi-angle computer vision sensing module is configured with at least two industrial cameras positioned to the side and above each testing station. This multi-view configuration can capture the three-dimensional deformation information of textiles during compression, especially for three-dimensional fabrics with complex surface topologies. It can reconstruct the volume changes and thickness compression field of the sample using binocular vision principles. The multi-angle sensing module is connected to the electrothermal acquisition cluster via a hardware synchronization trigger line, ensuring that all sensor data have a unified time origin.

[0052] The edge adaptive compensation node's hardware core is an embedded computing module equipped with a dedicated tensor processing unit. This edge adaptive compensation node is configured to run a lightweight physical information neural network model. Compared to Example 1, this physical information neural network has undergone pruning and quantization optimization, focusing on rapid real-time compensation of data within a specific pressure range. This edge adaptive compensation node can directly process large amounts of video stream data and electrical signal streams on-site and calculate the dynamic thermoelectric performance parameters of the sample in real time.

[0053] The edge adaptive compensation node also integrates anomaly diagnosis logic, configured to monitor the health status of each sensor in real time. If a decrease in image contrast or an abnormal increase in electrode contact impedance is detected at a certain workstation, the logic immediately marks the data for that period as low confidence and sends a warning signal to the operator. By performing primary data cleaning and physical model mapping at the edge, the bandwidth pressure on the backend network transmission is greatly reduced.

[0054] The cloud processing platform is configured as the core brain of the entire system, responsible for receiving the processed, streamlined datasets from various edge nodes. The cloud platform runs a full-scale deep neural network that utilizes distributed computing resources to perform cross-sample correlation analysis on the test results from each workstation.

[0055] The cloud processing platform also maintains a vast database of textile structural mechanics characteristics, encompassing typical deformation patterns of different fiber densities and weaving structures under various pressure conditions. The cloud platform is configured to utilize this prior knowledge to perform online parameter migration and optimization adjustments on the physical information neural network model at the edge. Through continuous feedback from big data, the system can continuously improve the accuracy of deformation compensation for novel thermoelectric textiles.

[0056] Furthermore, each station in the temperature-controlled loading array is equipped with a laser ranging component, which is configured as an auxiliary means of visual perception to measure the physical distance between the upper heating plate and the lower cooling plate in real time. The central processing unit can more accurately calculate the apparent density change of the thermoelectric textile during the compression process by fusing the laser ranging data with the surface strain analyzed by computer vision.

[0057] Furthermore, the distributed electrothermal signal acquisition cluster employs differential sampling technology, introducing a reference branch in each test loop to cancel common-mode noise caused by the high-current switching action of the temperature control platform in real time. The hardware circuitry of the electrothermal signal acquisition cluster is encapsulated in an aluminum alloy housing with magnetic shielding, and its internal circuit board layout undergoes rigorous impedance matching design to ensure the integrity of nanoampere-level weak signals.

[0058] Furthermore, the multi-angle computer vision perception module is configured to support infrared thermal imaging. By spatially registering the infrared thermal imager with a visible light camera, the system can acquire real-time images of the dynamic temperature field distribution on the sample surface. The edge adaptive compensation node is configured to use this spatially continuous temperature gradient information as an important boundary condition for the physical information neural network, more realistically simulating the electrothermal transport process inside the material and eliminating measurement deviations caused by local hot spots or thermal short circuits.

[0059] Furthermore, the adaptive deformation compensation module employs a continuum mechanics description method based on the Lagrange coordinate system when processing three-dimensional deformation data. This continuum mechanics description method is configured to describe the microscopic collapse of textiles as a nonlinear evolution process of the equivalent elastic modulus. In this process, the physical information neural network not only learns the mapping relationship of electrical signals but must also satisfy the constraints of the mechanical equilibrium equations, meaning that the stress distribution inside the sample must be consistent with the pressure load applied externally. This multi-physics coupling constraint mechanism ensures a high degree of physical and logical rigor in the compensated performance parameters.

[0060] Furthermore, the cloud processing platform provides an open API interface to support integration with external Materials Research and Management Systems (LIMS). This interface is configured to automatically upload the final characterization report for each sample group. The report includes the compensated Seebeck coefficient as a function of strain, the stability index of electrical conductivity under different pressures, and the power factor evaluation results of the material. Through this structured data output, researchers can quickly identify key process factors affecting the performance of flexible thermoelectric materials.

[0061] Example 3: Building upon Examples 1 and 2, this example further proposes an intelligent thermoelectric material performance testing system with self-healing monitoring capabilities. This system is specifically designed for flexible thermoelectric textiles with self-healing capabilities, aiming to accurately assess the evolution of the material's electrothermal properties during damage and repair processes.

[0062] A thermoelectric material performance testing system based on thermoelectric textiles includes an intelligent temperature control loading unit, a high dynamic electrothermal sampling module, a super-resolution visual analysis unit, an evolutionary deformation compensation module, and a logic decision center.

[0063] The intelligent temperature-controlled loading unit is configured to apply programmable micro-vibration. In addition to conventional static pressure loading, the piezoelectric ceramic actuator within the unit can apply minute dynamic mechanical disturbances with frequencies ranging from 0.1 Hz to 1000 Hz to thermoelectric textiles. These disturbances are configured to simulate the actual force conditions experienced by wearable devices during human movement.

[0064] The intelligent temperature-controlled loading unit also integrates a high-precision thermal resistance measurement and compensation logic. This logic is configured to detect the contact thermal resistance between the sample and the electrode in real time using a transient thermal pulse method. By feeding this thermal resistance value back to the logic decision center in real time, the system can automatically adjust the output power of the upper heating plate to maintain a constant temperature difference between the two ends of the sample, thus eliminating system interference caused by changes in contact thermal resistance with pressure on the Seebeck coefficient measurement.

[0065] The high-dynamic electrothermal sampling module is configured with an ultra-high sampling rate digital oscilloscope-grade front end. This configuration enables the system to capture transient voltage pulse signals generated by the breakage or re-contact of the microstructures within textiles. These pulse signals typically have a time width on the order of microseconds, making them difficult to record with traditional acquisition equipment.

[0066] The high-dynamic electrothermal sampling module also includes a real-time signal decomposition engine, which is configured to use a wavelet transform algorithm to decompose the original electrical signal into intrinsic performance components, mechanical noise components, and structural abrupt change components. This multi-scale signal analysis scheme provides finer-grained feature inputs for backend deformation compensation.

[0067] The super-resolution visual analysis unit is configured to employ an image enhancement module with a deep learning super-resolution algorithm. This image enhancement module can reconstruct original images from low-light or high-noise environments into clear, high-contrast images, enabling digital image association algorithms to capture even subtle fiber slippage phenomena.

[0068] The super-resolution visual analysis unit is also configured to identify minute cracks or voids on the surface of thermoelectric textiles. By integrating an edge detection operator, the super-resolution visual analysis unit can calculate the percentage of macroscopic damage to the sample in real time. This damage feature is transmitted in real time to the evolutionary deformation compensation module, serving as a crucial trigger switch for model switching.

[0069] The evolutionary deformation compensation module is based on a recurrent physical information neural network with long short-term memory. This model is configured to consider not only the current deformation state but also the impact of the deformation history on material properties. For flexible textiles, the structural hysteresis effect caused by mechanical fatigue is significant. The evolutionary compensation module can accurately distinguish between performance fluctuations caused by elastic deformation and performance degradation caused by plastic deformation by memorizing the sample's stress history through its internal state units.

[0070] The evolutionary deformation compensation module is also configured as an online learning loop. When the system detects a novel weave structure and its compensation residual exceeds a preset threshold, the module automatically initiates a parameter fine-tuning procedure. This procedure compares the deviation between the theoretical predictions of a known physical model and experimental observations, and uses the backpropagation algorithm to quickly update the weight parameters of the neural network, thus achieving the self-evolution of the compensation logic.

[0071] The logical decision-making center is configured as the coordination and command center for the entire system. Its core task is to dynamically adjust the working mode of each module according to the current testing phase.

[0072] The logical decision center integrates a reliability evaluation engine, which is configured to comprehensively consider structural integrity data from the vision unit and electrical signal stability data from the acquisition module. By calculating a normalized performance confidence index, the center can clearly demonstrate the true reliability of the measurement results to the end user. If the confidence index falls below a certain preset level, the center will automatically control the loading unit to reduce the pressure increase rate and enter a high-precision scanning mode.

[0073] Furthermore, the pressure regulation mechanism of the intelligent temperature-controlled loading unit is configured with a force-displacement dual closed-loop control mode. When testing certain ultralight thermoelectric textiles with extremely high porosity, the system can switch to displacement control mode to prevent sample pulverization and collapse by precisely limiting the downward displacement of the upper pressure head.

[0074] Furthermore, the hardware circuit of the high-dynamic electrothermal sampling module adopts a fully isolated differential input architecture, with each sampling channel having an independent reference level. This design can completely isolate conducted interference caused by the high-power drive circuit of the temperature control system, ensuring that the noise floor remains at an extremely low level when measuring nanoampere current.

[0075] Furthermore, the super-resolution visual analysis unit is configured to support holographic technology. By recording the interference pattern of light waves, the system can acquire the phase information of the sample, enabling real-time monitoring of textile surface roughness at the nanometer level. This high-precision surface topography data can help physical information neural networks more accurately establish contact resistance models.

[0076] Furthermore, the evolutionary deformation compensation module employs a variational algorithm to execute physical constraints. This algorithm is configured to transform the thermoelectric performance measurement problem into an optimization problem constrained by physical equations. By minimizing the weighted sum of measurement residuals and physical consistency residuals, it extracts the closest material intrinsic parameters to reality under dynamic testing environments with severe noise interference. Compared to purely data-driven models, this method possesses extremely strong extrapolation capabilities and physical interpretability.

[0077] Furthermore, the logic decision center is also equipped with a remote monitoring gateway, which supports synchronizing the test status to mobile terminals via an encrypted wireless channel. This remote monitoring gateway is configured to support voice alarm functionality; in the event of an emergency such as temperature runaway or pressure overload during testing, the system can immediately cut off the heating power supply and release mechanical pressure through preset logic, ensuring the absolute safety of the sample and equipment.

[0078] Example 4: This example describes an online thermoelectric material performance testing system for large-scale industrial production. The system is configured to be integrated into a continuous winding production line for thermoelectric textiles, enabling real-time scanning and quality control of the properties of the rolled materials.

[0079] A thermoelectric material performance testing system based on thermoelectric textiles includes a continuous belt loading mechanism, a synchronous non-contact electrothermal detection unit, a high-speed production line visual quality inspection module, a real-time flow adaptive compensator, and a production execution coordination system.

[0080] The continuous conveyor loading mechanism is configured to achieve smooth transport of thermoelectric textile samples via a series of tension control rollers. This continuous conveyor loading mechanism is configured to apply constant radial pressure while maintaining material movement via a pair of precisely aligned pressure rollers.

[0081] The continuous conveyor loading mechanism also includes an infrared preheating zone, configured to non-contactly heat the material before it enters the pressure roller, establishing a preset temperature gradient on its surface. By adjusting the preheating power and the ambient cooling airflow, the system can maintain a relatively stable steady-state thermal field on the continuously moving conveyor belt.

[0082] The synchronous non-contact electrothermal detection unit is configured to detect the local voltage and current distribution of a moving textile using electromagnetic induction or electrostatic coupling technology. To address the contact stability issue during motion, the synchronous non-contact electrothermal detection unit is also equipped with a set of slidable flexible graphite brush electrodes. These electrodes are suspended by constant pressure springs, enabling them to maintain a continuous electrical path with the material surface with minimal frictional resistance.

[0083] The synchronous non-contact electrothermal detection unit also integrates a multi-frequency impedance spectroscopy analysis module. This module is configured to detect the complex impedance characteristics of the material by injecting a weak AC sweep signal. By analyzing the response of the complex impedance plane at different frequencies, the system can calculate the bonding quality of the semiconductor particles in the textile and the dielectric properties of the substrate in real time.

[0084] The high-speed production line visual quality inspection module is configured to use a line scan camera for scanning and imaging. As the material moves, this module can stitch together and generate a continuous high-definition surface map of the entire roll of textile. The module integrates a motion blur compensation algorithm specifically designed for high-speed movement, ensuring that the interweaving pattern of the fibers remains clearly visible even at belt speeds reaching 30 meters per minute.

[0085] The high-speed production line visual quality inspection module is also configured to detect defects in the production process, such as fiber breakage, coating peeling, or uneven weaving. The location coordinates of these defects are recorded in real time and transmitted to the backend as spatial indexes.

[0086] The real-time streaming adaptive compensator is configured to employ hardware acceleration technology based on a Field-Programmable Gate Array (FPGA). Due to the massive data volume generated in pipeline production, traditional general-purpose processors struggle to meet real-time requirements. The compensator is configured to implement the inference logic of a physical information neural network at the hardware circuit level, enabling real-time processing of each row of pixel data and each set of electrical signal sample values.

[0087] The real-time flow adaptive compensator is configured to employ a spatially correlated sliding window algorithm. This algorithm comprehensively analyzes deformation data before and after the current sampling point, filtering out spurious strain signals caused by material turbulence on the roller conveyor. The parameter stream output by the compensator is dynamically calibrated, eliminating the systematic bias in the evaluation of electrothermal performance due to production line tension.

[0088] The production execution coordination system is configured as part of the overall factory digital architecture. It is responsible for receiving performance data from the real-time streaming compensator and automatically associating it with metadata such as production batch number and raw material batch.

[0089] The production execution coordination system integrates a trend prediction module, which is configured to monitor the wear and tear of production equipment based on long-term test data. If the conductivity correction weight of the entire roll of material shows a monotonically increasing trend, the system will automatically determine that the surface flatness of the pressure roller has decreased and will issue a prompt to maintenance personnel.

[0090] Furthermore, the surface of the pressure roller of the continuous conveyor loading mechanism is covered with a Teflon film with a high thermal conductivity. This Teflon film is configured to prevent thermoelectric materials from adhering to the roller under high temperature and pressure, while ensuring excellent heat transfer effect.

[0091] Furthermore, the synchronous non-contact electrothermal detection unit employs an active thermal management system to prevent temperature drift in its internal high-precision circuitry due to proximity to the high-temperature production line. This active thermal management system includes a circulating air-cooling channel and multiple thermocouple monitoring points, ensuring that the ambient temperature at the signal acquisition front end remains constant within a range of 25 degrees Celsius ± 1 degree Celsius.

[0092] Furthermore, the high-speed pipeline visual quality inspection module is configured to support polarized light illumination. By changing the polarization direction, the system can highlight the thickness inhomogeneity of semiconductor thin films in textiles, as these regions exhibit different polarization reflection characteristics due to multi-directional interference. This feature provides important auxiliary evidence for physical information neural networks to identify local conductivity anomalies.

[0093] Furthermore, the neural network structure within the real-time streaming adaptive compensator is configured to support online incremental learning. During production breaks, the system automatically retrieves some data samples with low confidence levels, retrains them using high-performance operators in the cloud, and then distributes the updated weight parameters online. This mechanism ensures that the detection system can quickly adapt to new types of thermoelectric textiles on the production line without requiring downtime for recalibration.

[0094] Furthermore, the production execution coordination system provides a 3D visualized digital twin interface configured to display the mass distribution of the entire roll of material in real time. Users can zoom in and out to view details of performance degradation caused by weaving stress concentration at specific locations. This deep transparency of quality improves the yield of flexible thermoelectric materials in industrial applications.

[0095] The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made by those skilled in the art without departing from the core principles of the present invention should be included within the scope of protection of the present invention. The modules, units, components, and their embedded algorithm logic within the system can be flexibly combined and adjusted according to the computational performance requirements, hardware cost constraints, and testing accuracy targets of the actual application scenario.

[0096] The compensation mechanism based on physical information neural networks involved in this invention is also applicable to the performance evaluation of other flexible sensors and flexible functional materials with complex deformation characteristics. All physical constraints, parameter correction logic, and timing synchronization schemes based on textual descriptions are designed to accurately reproduce the essential physical characteristics of flexible textiles.

[0097] All hardware components and their configurations described in this invention are essentially technical means to implement the stated functional logic and should not be construed as dependence on specific hardware models. In practical engineering, users can optimize and upgrade the interface protocol, bus architecture, and data structure according to changes in industry standards. The scope of protection of this invention is defined by the following claims.

Claims

1. A thermoelectric material performance testing system based on thermoelectric textiles, characterized in that, include: A temperature-controlled loading platform is configured to apply a controlled temperature gradient to thermoelectric textile samples and provide adjustable contact pressure during testing to simulate a mechanical load environment. An electrothermal signal acquisition unit is configured to simultaneously acquire the open-circuit voltage signal and short-circuit current signal generated by the thermoelectric textile sample under the action of temperature difference; A computer vision strain sensing unit is configured to acquire the surface image sequence of the thermoelectric textile sample in real time during the test, and analyze the full-field geometric deformation information of the thermoelectric textile sample under pressure based on a digital image association algorithm. An adaptive deformation compensation module is connected to the computer vision strain sensing unit and the electrothermal signal acquisition unit. It is configured to combine the embedded physical information neural network model to dynamically correct the open-circuit voltage signal and short-circuit current signal based on the full-field geometric deformation information, and output the compensated thermoelectric performance parameters. The central processing unit is connected to the temperature-controlled loading platform, the electrothermal signal acquisition unit, the computer vision strain perception unit, and the adaptive deformation compensation module, respectively. It is configured to coordinate the working sequence of the components, perform data fusion and model inference, and output a performance evaluation report.

2. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 1, characterized in that, The temperature-controlled loading platform includes a precision temperature control component and a pressure mechanical component; The precision temperature control component is configured to use a semiconductor cooling chip or resistance wire heating array as a heat source, and transfer heat energy to both ends of the thermoelectric textile sample through oxygen-free copper material with high thermal conductivity to build a stable temperature field. The pressure mechanical assembly includes an upper heating plate, a lower cooling plate, and a precision ball screw mechanism driven by a servo motor, the precision ball screw mechanism being configured to apply a vertical pressure load to the surface of the thermoelectric textile sample. The temperature-controlled loading platform integrates a high-sensitivity pressure sensor, which forms a closed-loop control circuit with the control driver of the servo motor. This circuit is used to dynamically compensate for pressure fluctuations caused by thermal expansion or collapse of the textile structure during the test, ensuring the constantness of the contact pressure. The surfaces of both the upper heating plate and the lower cooling plate are chemically nickel-plated to reduce heat radiation loss.

3. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 1, characterized in that, The electrothermal signal acquisition unit includes a high input impedance nanovoltmeter, a nanoampere-level low-noise ammeter, and an analog front-end circuit. The analog front-end circuit is connected to the electrode contacts of the thermoelectric textile sample via a multiplexer. The electrode contacts are made of gold-plated copper spring pin array, configured to maintain electrical contact when the thermoelectric textile sample is deformed. The electrothermal signal acquisition unit also includes a precision synchronization triggering module, which is configured to send a pulse synchronization signal to the computer vision strain perception unit when the electrical signal analog-to-digital conversion is completed, so as to achieve precise alignment of the electrothermal response data and the geometric deformation state on the time scale. The electrothermal signal acquisition unit is also equipped with a three-stage active filter to filter out power frequency interference in the environment and thermoelectric noise generated by the temperature control mechanism. The electrothermal signal acquisition unit is configured to use a four-wire measurement method, which uses independent leads to perform current injection and potential difference acquisition to eliminate the influence of lead resistance on the measurement results.

4. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 1, characterized in that, The computer vision strain sensing unit includes an industrial-grade high-resolution area array camera, a telephoto macro lens, a ring light source array, and a full-field strain analysis algorithm module. The industrial-grade high-resolution area array camera is mounted directly above the temperature-controlled loading platform, and its optical axis is perpendicular to the force plane of the thermoelectric textile sample. The ring light source array uses narrow-spectrum monochromatic light illumination, and in conjunction with the filter at the front end of the telephoto macro lens, it shields the background interference generated by natural light and thermal radiation from the temperature control platform. The full-field strain analysis algorithm module is configured to perform grayscale processing and Gaussian filtering to denoise the acquired original image, define a series of non-overlapping sub-regions in the image, find the best matching position of the sub-regions before and after deformation through an iterative algorithm based on the least squares criterion, and generate a two-dimensional strain tensor field covering the sample by calculating the displacement gradient of the centroid coordinates of the sub-regions. The full-field geometric deformation information includes the tensile strain, compressive strain, and shear strain data of the thermoelectric textile sample in the horizontal and vertical directions.

5. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 1, characterized in that, The physical information neural network model embedded in the adaptive deformation compensation module adopts a deep learning architecture. Its input vector includes the local strain tensor components at the current moment, the currently applied contact pressure value, the measured temperature difference distribution data, and the original electrical signal sample value. The physical information neural network model is configured to introduce thermoelectric transport physical constraints into the loss function. These constraints require that the output correction parameters follow the linear coupling relationship between the law of energy conservation and the Seebeck effect, ensuring that the corrected output satisfies the vector correlation characteristics between the electric field strength and the temperature gradient. The physical information neural network model is configured to identify spurious conductivity gains caused by reduced porosity of the thermoelectric textile sample by learning from historical experimental data, and to calculate a correction operator to offset the spurious conductivity gains using a nonlinear mapping function. By weighting the correction operator with the original measurement data, measurement errors introduced by mechanical deformation are eliminated.

6. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 5, characterized in that, The physical information neural network model is configured with structural collapse recognition logic to identify abrupt changes in contact area caused by the collapse of the porous structure of the thermoelectric textile sample; the recognition logic is configured as follows: When the local strain value output by the computer vision strain sensing unit exceeds the preset structural integrity threshold, the nonlinear compensation factor is automatically triggered. The nonlinear compensation factor is configured to reduce the contribution of the current measured conductivity value to the final performance evaluation and is replaced by an interpolated predicted value based on physical laws to eliminate the interference of abrupt changes in performance data caused by microfiber bundle breakage or overlap. The physical information neural network model also incorporates an entropy production rate constraint term during training. By calculating the entropy production rate of the output parameters under known temperature difference conditions, the model outputs a parameter combination that conforms to the second law of thermodynamics.

7. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 1, characterized in that, The central processing unit has a data fusion center running inside, which is configured to align the discrete electrical signal stream provided by the electrothermal signal acquisition unit with the continuous strain field image stream provided by the computer vision strain perception unit in a spatial coordinate system. By performing a topological mapping between high-strain regions in the strain field and electrode contact regions, the data fusion center is configured to quantitatively assess the trend of contact resistance variation at each contact point. The central processing unit is also connected to a large-capacity solid-state storage medium for storing the original image data, timestamp sequence, and performance indicators processed by the adaptive deformation compensation module. The central processing unit is also equipped with a visualization interface, which is configured to render dynamic strain thermograms and performance parameter evolution curves of the sample in real time, and supports three-dimensional view switching to show the degree to which the material properties deviate from the ideal physical model during pressure loading.

8. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 2, characterized in that, The temperature-controlled loading platform also includes a flexible heat flow meter and a laser ranging component; The flexible heat flow meter is attached to the bottom surface of the upper heating plate and is used to monitor the heat flow entering the thermoelectric textile sample and assist in calculating the thermal conductivity of the material. The laser ranging component is configured to measure the physical distance between the upper heating plate and the lower cooling plate in real time; The central processing unit is configured to fuse the physical distance data measured by the laser ranging component with the surface strain information parsed by the computer vision strain sensing unit to calculate the apparent density change of the thermoelectric textile sample during the compression process. The pressure adjustment mechanism of the temperature-controlled loading platform adopts a dual closed-loop control mode of force and displacement. It is configured to automatically adjust the displacement to keep the total positive pressure acting on the sample within a preset error range when the internal tension of the thermoelectric textile sample changes due to thermal shrinkage, or to prevent the sample from shattering and collapsing by limiting the downward displacement.

9. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 1, characterized in that, The system adopts a distributed architecture, including a multi-station temperature control loading array, a distributed electrothermal signal acquisition cluster, a multi-angle computer vision perception module, an edge adaptive compensation node, and a cloud processing platform. The multi-station temperature-controlled loading array consists of several independent temperature control units, supporting parallel performance evaluation of multiple thermoelectric textile samples. The distributed electrothermal signal acquisition cluster performs digital processing near the signal source through multiple signal processing terminals and uses differential sampling technology to cancel the common-mode noise generated by the temperature control platform. The multi-angle computer vision perception module includes multiple cameras arranged on the side and above the test station, which reconstruct the volume change field of the sample using the principle of binocular vision, and integrates an infrared thermal imager to obtain dynamic temperature field distribution images of the sample surface. The edge adaptive compensation node is configured to run a pruned and optimized physical information neural network model to perform real-time data cleaning and primary compensation. The cloud processing platform is configured to perform online migration and optimization of model parameters for each edge node using cross-sample association analysis and prior databases.

10. The thermoelectric material performance testing system based on thermoelectric textiles according to claim 1, characterized in that, The temperature-controlled loading platform has a programmable micro-vibration application function, which applies dynamic mechanical disturbance to the thermoelectric textile sample through an internal piezoelectric ceramic actuator to simulate the motion force state. The electrothermal signal acquisition unit includes a real-time signal decomposition engine, which is configured to use a wavelet transform algorithm to decompose the original electrical signal into intrinsic performance components, mechanical noise components and structural mutation components. The adaptive deformation compensation module employs a recurrent physical information neural network with long short-term memory capability, configured to memorize the stress history path of the thermoelectric textile sample through internal state units, and distinguish between performance fluctuations caused by elastic deformation and performance degradation caused by mechanical fatigue. The computer vision strain perception unit is also equipped with a deep learning super-resolution enhancement module, which is used to reconstruct high-contrast images to capture fiber slippage phenomena, and integrates an edge detection operator to calculate the percentage of macroscopic damage to the sample in real time, serving as a model switching trigger signal for the recurrent physical information neural network.

Images