Intelligent thermal characteristic nondestructive testing method for LED light source module

Abstract

The invention relates to an intelligent thermal characteristic nondestructive testing method for an LED light source module in the technical field of new-generation information, and the method comprises the steps: applying short-time thermal excitation to the surface of the LED module through a preset pulse sequence, obtaining initial temperature distribution data, and determining a thermal state reference when the excitation is started; according to the initial temperature distribution data, a transient thermal analysis algorithm is adopted to process a continuously collected temperature signal sequence, and the dynamic change trend of heat flow propagation in the module is determined; if the dynamic change trend exceeds a preset heat flux density threshold value, adjusting parameters of the pulse sequence, and obtaining an optimized excitation parameter set; constructing a thermal network model of the module according to the heat conduction contribution degree, and obtaining preliminary estimation values of thermal resistance and thermal capacity parameters; and time domain features are extracted from the continuously collected temperature signal sequence, and a deviation vector of the preliminary estimation value is determined.

Description

Technical Field

[0001] This invention relates to the field of next-generation information technology, and in particular to a method for intelligent non-destructive testing of the thermal characteristics of LED light source modules. Background Technology

[0002] As a core component in modern lighting and display fields, the thermal characteristics of light-emitting diode (LED) light source modules directly determine the product's performance stability, lifespan, and energy efficiency. With the widespread application of high-power LED modules in automotive lighting, industrial lighting, and other fields, rapid and accurate assessment of their thermal characteristics has become crucial to ensuring product quality.

[0003] However, current thermal characteristic testing methods have significant limitations, failing to meet the high-efficiency and accurate demands of modern production lines, especially in rapid iteration and high-throughput production environments. Traditional thermal characteristic testing primarily relies on steady-state measurements, requiring the LED module to be heated to thermal equilibrium, a process that often takes several hours, severely restricting production efficiency. This is particularly true for high-power LED modules, whose high heat flux density necessitates even longer thermal equilibrium times, and prolonged high-temperature operation during testing can damage the device. This inefficient testing method cannot meet the demands of modern production for rapid feedback and online monitoring, limiting the real-time nature of product quality control. The core technical challenge in thermal characteristic testing lies in obtaining accurate thermal resistance and thermal capacity parameters in a short time while ensuring the testing process is non-destructive to the device. Thermal resistance and thermal capacity parameters are key indicators describing the thermal conductivity of LED modules, directly impacting the optimization of thermal management design.

[0004] However, rapidly applying thermal excitation and accurately capturing the temperature response presents a complex technical challenge. Short-duration thermal excitation may result in insufficient temperature change signals, making it difficult to accurately extract the thermal characteristics of each material layer; conversely, excessive excitation may cause overheating damage to the device, affecting the reliability of the test and the safety of the device. For example, when testing a high-power LED module for automotive headlights, improper control of the pulse excitation time or intensity may cause the chip junction temperature to rise rapidly, triggering thermal runaway, which in turn affects the accuracy of the test data or directly damages the sample.

[0005] Therefore, how to apply appropriate thermal excitation within a short time to quickly and accurately capture the temperature response curve and extract reliable thermal resistance and thermal capacity parameters while avoiding device damage has become a key issue in the field of rapid thermal characteristic testing. This issue not only requires the test system to have high-precision temperature monitoring capabilities, but also to achieve dynamic balance in the control of thermal excitation to adapt to the testing needs of LED modules with different power and packaging structures. Summary of the Invention

[0006] This invention provides an intelligent non-destructive testing method for the thermal properties of LED light source modules, mainly including:

[0007] A short-term thermal excitation is applied to the surface of the LED module using a preset pulse sequence to obtain initial temperature distribution data and determine the thermal state baseline at the start of the excitation. Based on the initial temperature distribution data, a transient thermal analysis algorithm is used to process the continuously acquired temperature signal sequence to determine the dynamic change trend of heat flow propagation within the module. If the dynamic change trend exceeds a preset heat flow density threshold, the parameters of the pulse sequence are adjusted to obtain an optimized excitation parameter set. An adjusted temperature response curve is obtained from the optimized excitation parameter set, and the frequency components in the temperature response curve are decomposed to determine the heat conduction contribution of each layer of material. A thermal network model of the module is constructed based on the heat conduction contribution to obtain preliminary estimates of thermal resistance and thermal capacity parameters. Time-domain features are extracted from the continuously acquired temperature signal sequence to determine the deviation vector of the preliminary estimates. If the absolute value of an element in the deviation vector is greater than a preset accuracy threshold, the node connection weights in the thermal network model are iteratively updated to obtain a refined set of thermal resistance and thermal capacity parameters. The refined set of thermal resistance and thermal capacity parameters is compared with the initial temperature distribution data to determine the non-destructive compliance of the overall thermal characteristics and obtain test evaluation report data. Furthermore, the step of applying short-term thermal excitation to the surface of the LED module using a preset pulse sequence to obtain initial temperature distribution data and determine the thermal state benchmark at the start of excitation includes: applying the short-term thermal excitation to the surface of the LED module using the pulse sequence; collecting the temperature value of the LED module surface using an infrared thermal imager to obtain the initial temperature distribution data; calculating the standard deviation of the initial temperature distribution data as a uniformity index; if the uniformity index is lower than a preset threshold, reapplying the thermal excitation by adjusting the parameters of the pulse sequence to obtain adjusted data; and extracting timestamp information from the adjusted data to determine the thermal state benchmark at the start of excitation. Furthermore, the step of processing the continuously acquired temperature signal sequence using a transient thermal analysis algorithm based on the initial temperature distribution data to determine the dynamic change trend of heat flow propagation within the module includes: processing the continuously acquired temperature signal sequence using the finite difference method based on the initial temperature distribution data, calculating the temperature change within a time step, and obtaining the temperature sequence processing result; calculating the heat flow propagation trend within the module using the Fourier heat conduction equation based on the temperature sequence processing result, and obtaining the heat conduction simulation output; if the heat conduction simulation output exceeds a preset threshold, adjusting the heat flow propagation trend through signal noise filtering to obtain heat flow path tracking data; integrating the data preprocessing steps from the heat flow path tracking data, determining the integrity of the heat distribution simulation, and obtaining the dynamic change trend.Furthermore, if the dynamic change trend exceeds a preset heat flux density threshold, the parameters of the pulse sequence are adjusted to obtain an optimized excitation parameter set. This includes: acquiring real-time heat flux density data through a sensor to obtain the dynamic change trend; if the dynamic change trend exceeds the preset heat flux density threshold, adjusting the amplitude and duration of the pulse sequence to obtain adjusted sequence parameters; extracting key features from the adjusted sequence parameters to obtain the optimized excitation parameter set; and updating the heat flux monitoring mechanism based on the optimized excitation parameter set to obtain the final parameter set update result. Furthermore, the step of obtaining the adjusted temperature response curve from the optimized excitation parameter set, decomposing the frequency components in the temperature response curve, and determining the thermal conductivity contribution of each material layer includes: obtaining the adjusted temperature response curve from the optimized excitation parameter set; decomposing the frequency components in the adjusted temperature response curve using a Fourier transform algorithm to obtain a set of frequency components; obtaining a contribution index for the set of frequency components by comparing the amplitude of each frequency with a preset material thermal conductivity frequency correlation threshold; determining the response curve analysis result by matching the contribution index with a preset mapping table for material layer separation; and obtaining the thermal conductivity distribution characteristics of each material layer by superimposing the response curve analysis result with thermal conductivity frequency correlation distribution data. Furthermore, the step of constructing a thermal network model of the module based on the heat conduction contribution to obtain preliminary estimates of thermal resistance and heat capacity parameters includes: obtaining heat flow path data from the module structure based on the heat conduction contribution to determine a preliminary value of thermal resistance; performing mesh generation on the preliminary value of thermal resistance using a finite element simulation algorithm to obtain a thermal network model framework; calculating the temperature gradient inside the module using the thermal network model framework to determine the range of heat capacity parameters; if the range of heat capacity parameters exceeds a preset threshold, adjusting the heat flow path data by correcting the contribution weight to obtain an optimized thermal resistance estimate; and extracting heat distribution features from the optimized thermal resistance estimate to obtain a preliminary estimate of the heat capacity parameters. Furthermore, the step of extracting time-domain features from the continuously acquired temperature signal sequence to determine the deviation vector of the preliminary estimate includes: obtaining time-domain features from the continuously acquired temperature signal sequence to determine the deviation basis of the preliminary estimate; extracting the temperature change trend from the time-domain features for the preliminary estimate to determine the initial deviation; processing the continuously acquired temperature signal sequence using a deviation correction process based on the initial deviation to obtain a deviation vector component; if the deviation vector component exceeds a preset threshold, adjusting the parameter optimization to fuse the estimation deviation analysis to determine the correction direction of the deviation vector; and integrating the temperature signal sequence and the time-domain features from the correction direction to determine the deviation vector.Furthermore, if the absolute value of an element in the deviation vector is greater than a preset precision threshold, the node connection weights in the thermal network model are iteratively updated to obtain a refined set of thermal resistance and thermal capacity parameters. This includes: extracting the absolute value of an element from the deviation vector; determining whether to trigger iteration by comparing it with the preset precision threshold; obtaining the set of node connection weights to be updated; adjusting the weight values ​​using a gradient descent algorithm for the set of node connection weights to obtain preliminary refined thermal resistance parameters; determining the iteration direction by integrating the thermal capacity calculation module based on the preliminary refined thermal resistance parameters; if the iteration direction of the thermal capacity parameters is related to the deviation of the thermal resistance parameters, obtaining integrated parameter pairs through multiple iterations; verifying thermal balance from the integrated parameter pairs to obtain the refined set of thermal resistance and thermal capacity parameters.

[0008] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0009] This invention discloses an intelligent non-destructive testing method for the thermal properties of LED light source modules. Addressing the difficulty of traditional thermal analysis methods in accurately capturing the dynamic changes in heat flow and the contribution of heat conduction within the module, this invention proposes a complete solution. The invention obtains the initial temperature distribution by applying a short-duration thermal excitation pulse sequence and uses a transient thermal analysis algorithm to process the continuous temperature signal, accurately determining the heat flow propagation trend. When the heat flow density exceeds a threshold, the pulse parameters are adaptively adjusted to optimize the excitation effect and extract the temperature response curve. The frequency components are decomposed using Fourier transform to determine the contribution of each layer of material to heat conduction, and a finite element thermal network model is further constructed to preliminarily estimate thermal resistance and thermal capacity parameters. This invention combines time-domain feature extraction and deviation vector iteration to refine the model parameters, and finally generates a high-precision thermal property evaluation report by comparing it with the initial data. This invention significantly improves the non-destructive nature and accuracy of thermal property testing and is suitable for thermal management optimization of complex LED modules. Attached Figure Description

[0010] Figure 1 This is a flowchart of a non-destructive testing method for the intelligent thermal characteristics of an LED light source module according to the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0012] like Figure 1 This embodiment of a non-destructive testing method for the intelligent thermal characteristics of an LED light source module may specifically include:

[0013] S101. Apply short-time thermal excitation through a preset pulse sequence to obtain initial temperature distribution data from the surface of the LED module and obtain the thermal state reference at the start of excitation.

[0014] The initial temperature distribution is obtained by applying a short-duration thermal excitation to the surface of the LED module using a preset pulse sequence. Based on this initial temperature distribution, an infrared thermal imager is used to collect temperature values ​​on the LED module surface, yielding temperature distribution data. The standard deviation of the temperature distribution data is calculated as a uniformity index. If the uniformity index is lower than a preset threshold, the thermal excitation application is adjusted by changing the pulse sequence parameters, thus obtaining the data acquisition process. Timestamp information is extracted from the data acquisition process to determine the start time of the excitation, obtaining the initial thermal baseline.

[0015] In one implementation, the process of applying short-term thermal excitation through a preset pulse sequence first requires preparing the test environment for the LED module.

[0016] For example, the LED module is placed in a constant temperature chamber, ensuring the ambient temperature remains stable between 20 and 25 degrees Celsius. This preparation helps reduce the impact of external interference on thermal excitation. The preset pulse sequence refers to a series of electrical pulse signals with specific intervals, which are input to the LED module's drive circuit through the power supply module.

[0017] It should be noted that the pulse sequence design is based on the thermal response characteristics of the LED module. For example, the sequence includes multiple pulses with widths ranging from 1 to 5 milliseconds and intervals ranging from 10 to 50 milliseconds to achieve short-duration thermal excitation without causing overheating damage. Through this sequence, thermal excitation generates instantaneous heat within the LED chip, thereby creating a measurable temperature gradient on the module surface. Furthermore, an infrared thermal imager is used as the acquisition device to obtain initial temperature distribution data from the LED module surface. The infrared thermal imager can capture surface thermal radiation and convert it into a temperature distribution map.

[0018] Specifically, acquisition begins immediately after the first pulse input, at the start of the applied pulse sequence, recording the temperature field data of the module surface. This data includes the temperature values ​​corresponding to multiple pixels, forming a two-dimensional temperature distribution matrix.

[0019] For example, in a typical LED display module, the acquisition resolution is 320 by 240 pixels to ensure coverage of the entire module surface. The acquired data reflects the thermal state at the start of excitation, including hotspot areas and uniformity distribution; this information is used for subsequent analysis.

[0020] Preferably, obtaining the thermal state baseline at the start of excitation involves processing the acquired data. The thermal state baseline refers to a reference temperature distribution model used to compare subsequent thermal changes. Specifically, the initial temperature distribution data is processed using an averaging filtering algorithm to remove noise points, and then the average temperature and standard deviation are calculated as baseline parameters.

[0021] In one possible implementation, for a backlight LED module, the benchmark calculation takes into account the module's geometry, such as the temperature gradient of a rectangular region decreasing from the center to the edge. This benchmark establishes the thermal equilibrium state of the LED module under no-load excitation, providing a basis for detecting thermal faults.

[0022] In one embodiment, different types of LED modules are considered to demonstrate the versatility of the method.

[0023] For example, for small LED lighting modules, the pulse sequence can be adjusted to lower power to accommodate their smaller heat capacity.

[0024] Specifically, when applying thermal excitation, the pulse amplitude is controlled between 5 and 10 volts, and the sequence duration does not exceed 1 second. After acquiring temperature data from the surface, the reference formation process includes a calibration step, using a known temperature reference point to correct the thermal imager readings. This implementation ensures that in lighting applications, the thermal state reference accurately reflects the initial thermal distribution of the module, facilitating the monitoring of thermal decay over long-term operation. Further, in another embodiment, for large-size LED display modules, the pulse sequence is designed as a multi-stage process, with a first stage being a rapid heating pulse and a second stage being a sustaining pulse, to simulate actual workload. When acquiring initial temperature distribution data, multi-angle thermal imager scanning is used to ensure coverage of curved or irregular surfaces. The process of obtaining the thermal state reference involves region segmentation, dividing the module surface into multiple sub-regions, with temperature statistics calculated independently for each sub-region. This method enhances the accuracy of the reference, enabling effective assessment of thermal non-uniformity in large modules and resulting in stable thermal management in the display field.

[0025] Understandably, the implementation of this technical solution can also be extended to integrated LED modules.

[0026] For example, before applying a short-term thermal excitation, the module is pre-powered and stabilized to simulate real-world usage scenarios. Temperature distribution data is acquired using a high-speed infrared sensor to capture the dynamic changes at the moment of excitation initiation. The reference is formed through vector representation, converting the temperature distribution into a set of feature vectors for subsequent comparison. This approach demonstrates the flexibility of the solution in integrated circuit LEDs, ensuring that the thermal state reference supports various detection requirements.

[0027] Specifically, in practice, the application of the pulse sequence is achieved through a dedicated controller connected to the LED driver board. The principle of the thermal excitation process is based on the Joule heating effect, where the electrical pulse is converted into heat energy that rapidly diffuses to the surface. After data acquisition, benchmark calculations avoid the influence of environmental variables by subtracting the background temperature using a differential method. This detailed process ensures that the technical features of the claims are fully realized and provides reliable support in the field of LED module thermal detection.

[0028] S102. Based on the acquired initial temperature distribution data, the transient thermal analysis algorithm is used to process the continuously acquired temperature signal sequence to determine the dynamic change trend of heat flow propagation inside the module.

[0029] The continuous signal acquisition is processed using the finite difference method based on the initial temperature distribution. This method calculates temperature changes within a time step based on temperature grid points, yielding a temperature sequence processing result. Based on this result, the internal trend of the module is calculated using the Fourier heat conduction equation to determine the heat transfer simulation output, considering the dynamics of heat flow propagation. If the simulation output exceeds a preset threshold, the dynamic trend prediction is adjusted through signal noise filtering, using low-pass filtering to remove high-frequency interference, resulting in heat flow path tracking data. The heat flow propagation dynamics are obtained from the heat flow path tracking data. This data preprocessing step is then integrated for module fault warning, including normalization and assessment of the integrity of the heat distribution simulation. Based on the integrity of the heat distribution simulation and the internal trend of the module, the dynamic change trend is obtained.

[0030] In one implementation, initial temperature distribution data of the module is first acquired. This data can be collected by multiple temperature sensors arranged inside the module, for example, in a battery module, where the sensors are placed at key locations in the battery cells, such as the positive electrode, negative electrode, and separator. The initial temperature distribution data reflects the thermal state of the module at the initial moment, providing a benchmark for subsequent analysis.

[0031] Specifically, by simultaneously acquiring temperature values ​​from multiple points, a temperature field matrix is ​​formed to characterize the non-uniformity of heat distribution. This data acquisition method ensures the accuracy of the analysis because it captures local thermal differences within the module; for example, hot spots may appear in certain areas during high-load operation. Furthermore, a transient thermal analysis algorithm is employed to process the continuously acquired temperature signal sequence. Based on the heat conduction equation and considering the influence of time variables, the transient thermal analysis algorithm simulates the heat propagation process within the module.

[0032] For example, the algorithm can use the finite difference method to discretize the module space into grid points and iteratively calculate the temperature change at each time step. The specific process includes: first, initializing the temperature values ​​on the grid using initial distribution data; then, calculating the heat flow vector for each grid point according to the heat conduction equation, where the heat flow vector is determined by the temperature gradient and the material's thermal conductivity; and finally, updating the temperature distribution for the next time step. This processing method can capture the dynamic evolution of heat flow, such as the path and speed of heat transfer from high-temperature areas to low-temperature areas when heat dissipation is uneven in the module. Preferably, when processing continuously acquired temperature signal sequences, the signals need to be pre-filtered to remove noise.

[0033] For example, a low-pass filter is used to smooth the signal sequence, ensuring the reliability of the algorithm's input data. Continuous acquisition means that the temperature signal is acquired at a fixed sampling rate, such as once per second, forming a time series. Based on this, the transient thermal analysis algorithm simulates the thermal diffusion process by solving partial differential equations.

[0034] Specifically, the algorithm considers parameters such as the heat capacity and density of the module materials to calculate the speed and direction of heat flow propagation, thereby revealing potential thermal runaway risks. In one possible implementation, determining the dynamic trend of heat flow propagation within the module is achieved by analyzing the temperature field evolution sequence output by the algorithm. The dynamic trend includes the dominant direction of heat flow, propagation rate, and hotspot migration path.

[0035] For example, in an electronic module, if the heat flow trend shows diffusion towards the edge of the circuit board, it indicates that the heat dissipation design is effective. The algorithm outputs a trend vector sequence, where each vector represents the average heat flow intensity and direction over a specific time period. By comparing consecutive sequences, it identifies whether the trend is accelerating or slowing down. This determination method supports real-time monitoring, helping to optimize the module's thermal management strategy. It should be noted that the transient thermal analysis algorithm can be implemented in conjunction with numerical simulation software for auxiliary calculations.

[0036] Specifically, the temperature signal sequence is input into a simulation environment, and the algorithm iteratively updates the heat flow state until it reaches stability or a specified time point. For example, in battery module applications, if the initial temperature distribution shows high temperatures in the central region, the algorithm can predict the trend of heat flow propagation to the periphery, thereby assessing the cooling system requirements. This method enhances the practicality of the technical solution because it provides quantifiable dynamic indicators. Furthermore, to adapt to different module types, algorithm parameters, such as time step or grid resolution, can be adjusted.

[0037] In one embodiment, a finer mesh is used for small electronic modules to capture microscopic heat flow changes, while a coarser mesh is used in large battery modules to improve computational efficiency. This flexibility ensures the versatility of the solution while maintaining consistency in applications within the thermal management field. For example, in practical operation, continuously acquired temperature signal sequences may contain sudden peaks. Transient thermal analysis algorithms handle these anomalies by introducing boundary conditions, such as setting the heat dissipation coefficient of the module's outer wall. When calculating the heat flow propagation trend, the algorithm generates a visual chart showing changes in the heat flow vector field, helping users understand the dynamic process. It is understood that this technical solution achieves accurate tracking of the module's internal thermal state through detailed algorithmic processing. In implementation, the trend determination step can also integrate an alarm mechanism, triggering an alert when the heat flow propagation rate exceeds a threshold, thereby enhancing module safety. In another embodiment, the acquisition of initial temperature distribution data can be supplemented by infrared imaging technology to complement sensor data. The specific process includes scanning the module surface, generating a thermal image, and mapping it to the internal structure. This combination improves the comprehensiveness of the data, providing richer input for transient analysis. Overall, these implementation methods cover the entire process from data acquisition to trend determination, ensuring the effective application of the technical solutions in module thermal management.

[0038] S103. If the dynamic change trend exceeds the preset heat flux density threshold, adjust the amplitude and duration of the pulse sequence to obtain the optimized excitation parameter set.

[0039] Real-time heat flux density data is collected by sensors to obtain dynamic trends. If these trends exceed a preset heat flux density threshold, the pulse sequence amplitude and duration are adjusted to determine the adjusted sequence parameters. Key features are extracted from the adjusted sequence parameters to obtain an optimized excitation parameter set. The heat flux monitoring mechanism is updated based on the optimized excitation parameter set to obtain the final parameter set update.

[0040] In one implementation, the system first monitors the dynamic change trend of the material, for example, by acquiring temperature and heat flow data in real time through sensors, and calculates trend indicators.

[0041] Specifically, the dynamic change trend can be determined through differential analysis of continuous time series data, such as comparing the current heat flux density value with the value at the previous moment to obtain the rate of change. If this rate of change exceeds a preset threshold, an adjustment mechanism is triggered. This monitoring process ensures timely response of the excitation parameters, maintaining the thermal equilibrium of the material in the field of laser processing and avoiding overheating or under-excitation. In this way, the system achieves precise control of heat flux density, providing a data foundation for subsequent parameter optimization. Furthermore, the preset heat flux density threshold is determined based on the thermal conductivity of the material and environmental conditions.

[0042] For example, in a laser pulse excitation scenario, the threshold can be set to 80% of the material's melting point to prevent phase transition from occurring.

[0043] It should be noted that the calculation of the heat flux density threshold involves a simplified form of the heat conduction equation, such as considering heat flux density q as equal to thermal conductivity k multiplied by the temperature gradient, where the threshold serves as a safety upper limit. If the dynamic trend, such as the rate of temperature rise, exceeds 5 degrees Celsius per second, it is determined that the threshold has been exceeded. This threshold preset process emphasizes the preventative nature of operations, reducing defect generation and ensuring stable processing quality in practical laser welding applications. The detailed implementation of this step contributes to the system's adaptability to different material types, such as adjusting the threshold parameters for metal alloys or ceramic materials, thus demonstrating the versatility of the technical solution.

[0044] Preferably, when the dynamic change trend exceeds the threshold, the system adjusts the amplitude and duration of the pulse sequence.

[0045] Specifically, the amplitude adjustment can be achieved through a proportional-integral-derivative control algorithm. For example, if the trend exceeds the upward limit, the amplitude is reduced by 10% to reduce heat input; the duration is shortened or extended according to the trend slope. For example, when the trend changes sharply, the duration is halved to 50% of the original value.

[0046] In one possible implementation, this adjustment process first calculates the deviation value, i.e., the difference between the actual trend and the threshold, and then applies a feedback loop to progressively optimize the parameters. This detailed adjustment logic is particularly useful in laser cutting operations, enabling real-time responses to changes in material thickness and preventing uneven cutting or over-ablation. Through multiple iterations, the system obtains a set of amplitude and duration combinations, forming a preliminary set of excitation parameters, thereby improving overall processing efficiency.

[0047] For example, in laser surface treatment applications, assuming an initial pulse amplitude of 100 watts and a duration of 1 millisecond, if the detected heat flux density trend exceeds a threshold, such as reaching 500 watts per square centimeter, the system automatically reduces the amplitude to 80 watts and extends the duration to 1.5 milliseconds to balance the heat distribution. This adjustment not only optimizes the parameters but also ensures the integrity of the material's microstructure.

[0048] Understandably, this process can be implemented through software modules, with data flow between modules forming a closed-loop control from monitoring to adjustment, further enhancing the system's robustness. In another embodiment, for high-precision laser etching scenarios, the assessment of dynamic trends can be incorporated into machine learning models, such as using time series prediction to predict in advance whether a threshold will be exceeded.

[0049] Specifically, the model takes historical heat flow data as input and outputs a predicted trend. If the predicted value exceeds a threshold, the pulse parameters are pre-adjusted, such as fine-tuning the amplitude from the standard value by 5%, with the duration changing accordingly. This method expands the predictive capabilities of the business, reducing scrap rates and achieving finer control in semiconductor material processing.

[0050] It should be noted that the model training process is based on a large amount of experimental data to ensure the accuracy of the adjustments, thus providing reliable support for optimizing the stimulus parameter set. Furthermore, the optimized stimulus parameter set is obtained through iterative optimization algorithms, such as using gradient descent to gradually minimize the deviation between the trend and the threshold.

[0051] Specifically, the algorithm initializes the parameter set and then updates the amplitude and duration in each iteration until the trend stabilizes within a threshold. When applied in the field of laser medical devices, this process optimizes pulses to match tissue thermal response and avoids damage. Through multiple iterations, the final parameter set includes a set of amplitude and duration values, forming a complete excitation scheme, demonstrating the flexibility of the technology within the same domain.

[0052] In one embodiment, the system may also integrate a feedback sensor for further refinement.

[0053] For example, if the trend still exceeds the limit after adjustment, the process is repeated until optimization is complete. This iterative mechanism ensures the reliability of the parameter set, enabling it to adapt to different hole diameter requirements and improve production consistency in industrial laser drilling operations.

[0054] Preferably, in actual deployment, the optimized set of incentive parameters can be stored as a template for quick invocation in similar scenarios.

[0055] For example, multiple sets of parameters can be pre-stored for specific material types, facilitating switching. This storage and retrieval process simplifies operations and enhances the practicality of the solution.

[0056] Understandably, through the above steps, the system achieves effective management of heat flux density, providing stable excitation control in the field of laser processing.

[0057] For example, in laser cladding applications, the adjusted parameter set can improve coating uniformity, reduce thermal stress concentration, and thus extend the material's service life.

[0058] S104. Obtain the adjusted temperature response curve from the optimized excitation parameter set, decompose the frequency components in the curve using the Fourier transform algorithm, and determine the contribution of each layer of material to heat conduction.

[0059] An initial temperature response curve is obtained from a preset set of excitation parameters, and an optimized temperature response curve is obtained by adjusting the parameter set. Fourier transform is used to decompose the frequency components in the optimized temperature response curve. The Fourier transform input is the optimized temperature response curve, and the output is a set of frequency components. By representing the curve as time-series data, a Fourier transform formula is applied to the sequence to convert the time-domain signal into a frequency-domain signal, obtaining the set of frequency components. The contribution of each layer of material to thermal conduction is determined based on the set of frequency components. By comparing the amplitude of each frequency in the set with a preset material thermal conduction frequency correlation threshold, a contribution index is obtained. This contribution index reflects the proportion of each layer of material's contribution to overall thermal conduction. Based on the correlation between the contribution index and the material layer separation, the result of the response curve analysis is determined by matching the contribution index to a preset mapping table of material layer separation. The result of the response curve analysis is fused with the thermal conduction frequency correlation. By superimposing the result of the response curve analysis with the distribution data of the thermal conduction frequency correlation, the thermal conduction distribution characteristics of each layer of material are obtained.

[0060] In one implementation, obtaining the adjusted temperature response curve from the optimized set of excitation parameters first requires understanding the optimization process of the excitation parameter set. These parameters include heat source intensity, excitation frequency, and duration, which are adjusted through iterative algorithms to minimize noise effects and ensure that the curve reflects the true thermal response.

[0061] For example, in the field of multilayer composite material testing, the optimized parameter set can be applied to the thermal imaging testing of aerospace composite panels, and the obtained temperature response curve represents the change of material surface temperature over time. Furthermore, the frequency components in the curve are decomposed using a Fourier transform algorithm. The Fourier transform is a mathematical tool that converts a time-domain signal into a frequency-domain signal, and here it is used to extract the frequency spectrum of the temperature curve.

[0062] Specifically, the temperature response curve is used as the input signal, and its spectrum is calculated using a Fast Fourier Transform algorithm to decompose it into different frequency components. These components correspond to the propagation characteristics of heat waves in the material; low-frequency components are usually related to deeper material layers, while high-frequency components reflect the response of shallower layers. This decomposition isolates the amplitude and phase information of each frequency, providing a basis for subsequent judgment.

[0063] It should be noted that the assessment of the thermal conductivity contribution of each material layer is based on frequency component analysis. Thermal conductivity contribution refers to the relative influence of each material layer on the overall heat transfer. In the decomposed spectrum, different frequencies correspond to different thermal diffusion depths because the penetration depth of a heat wave is inversely proportional to its frequency.

[0064] For example, a larger amplitude of low-frequency components indicates that the thermal conductivity of the deeper material dominates the response. The contribution can be calculated by comparing the energy distribution of each frequency component, for example, by using the proportion of the sum of squared amplitudes to quantify the contribution of each layer.

[0065] In one possible implementation, a multilayer material thermal conductivity model is considered. The material is assumed to consist of three layers: a surface coating, a substrate, and a bottom support. After obtaining the temperature response curve, a Fourier transform is performed to obtain the frequency spectrum. Then, based on the thermal diffusion equation, the frequency is mapped to the depth.

[0066] For example, the relationship between frequency f and depth d is approximately d proportional to 1 / sqrt(f). By analyzing the intensity of components within a specific frequency range, the contribution of the corresponding layer to thermal conductivity can be determined. For instance, if the mid-frequency components decay slowly, the second layer will contribute more to thermal conductivity. d proportional to 1 / sqrt(f) is a physical law describing the fundamental relationship between signal penetration depth and frequency. It implies that high frequencies indicate shallow depths, while low frequencies indicate deep depths. In thermal analysis, it is used as a "yardstick" to deduce the thermal properties (such as thermal conductivity) of material layers at different depths by analyzing the intensity of signals at different frequencies.

[0067] Preferably, in actual testing scenarios, a reference curve can be introduced for calibration.

[0068] For example, optimized excitation is applied to a multilayer sample with known thermal conductivity to obtain a reference temperature response curve and decompose its frequency components. By comparing the spectrum with that of the sample under test, the differences in contribution of each layer are quantified. This method enhances the accuracy of the judgment and is applicable to testing different batches of materials in the same field.

[0069] For example, in the production quality control of aerospace composite materials, the above process can be used to identify problems of uneven interlayer heat conduction.

[0070] Specifically, if the Fourier transform shows abnormal high-frequency components, the surface layer contribution is low, which may indicate a defect. This assessment allows for timely adjustments to manufacturing parameters, optimizing thermal performance.

[0071] It is understood that the implementation of this technical solution is not limited to a single excitation method. In another embodiment, using a pulse excitation parameter set to obtain the transient temperature response curve, and then determining the contribution after Fourier transform decomposition, is equally effective. This flexibility supports its application in the detection of thermal barrier coatings, ensuring the thermal insulation effect of multilayer structures. Furthermore, the results show that, through frequency component decomposition, the determination of the thermal conduction contribution of each layer can provide quantitative analysis, supporting material design improvements.

[0072] S105. Based on the determined contribution of heat conduction, a thermal network model of the module is constructed using a finite element simulation algorithm to obtain preliminary estimates of thermal resistance and thermal capacity parameters.

[0073] Based on the contribution of heat conduction, heat flow path data is obtained from the module structure to determine the preliminary value of thermal resistance. A finite element method (FEM) simulation algorithm is used to mesh the preliminary thermal resistance value. The FEM algorithm calculates the heat conduction equation by discretizing the module structural elements to obtain a thermal network model framework. Using this thermal network model framework, the internal temperature gradient of the module is calculated. The temperature gradient is determined by solving for the difference value based on the heat flow path data to determine the range of the heat capacity parameter. If the heat capacity parameter range exceeds a preset threshold, the heat flow path data is adjusted by correcting the contribution weights to obtain an optimized thermal resistance estimate. From the optimized thermal resistance estimate, heat distribution features are extracted. These heat distribution features are obtained by integrating the temperature gradient data to obtain the final preliminary value of the heat capacity parameter.

[0074] In one implementation, the contribution of heat conduction is first determined, which is evaluated based on the material thermal conductivity and geometry of each component in the module.

[0075] Specifically, by analyzing the heat flow path of the module, the contribution of each component to the overall heat conduction is quantified.

[0076] For example, in the field of electronic modules, the heat transfer efficiency of chips and heat sinks is considered.

[0077] It should be noted that this judgment helps identify the dominant heat conduction factors, thus providing basic data for subsequent simulations. Based on the above judgment, a thermal network model of the module is constructed using the finite element method (FEM). The FEM is a numerical method that discretizes the physical structure of the module into a finite number of element meshes, solving the heat conduction equations within each element.

[0078] For example, in the thermal management scenario of an electronic module, a three-dimensional geometric model is first established, and then boundary conditions such as heat sources and ambient temperature are applied. Furthermore, through iterative solving, the distribution and flow of heat within the module are simulated. The advantage of this algorithm is its ability to handle complex geometries, ensuring that the model accurately reflects actual thermal behavior.

[0079] In one possible implementation, the thermal network model is constructed by transforming the finite element simulation results into an equivalent thermal resistance network. The thermal network model is similar to a circuit network, where nodes represent temperature points and branches represent thermal resistance.

[0080] Specifically, the thermal resistance of each path is calculated based on the heat flux density calculated by the finite element method.

[0081] For example, the module can be divided into multiple sub-regions, where the thermal resistance of each region is determined by the material's thermal conductivity and path length. In this way, the model simplifies complex heat conduction problems and facilitates parameter analysis.

[0082] Preferably, preliminary estimates of thermal resistance and heat capacity parameters are obtained by solving a thermal network model. The thermal resistance parameter represents the resistance to heat flow through the path, while the heat capacity parameter reflects the module's ability to store heat.

[0083] In one embodiment, a transient thermal load is applied to the model, and the temperature response curve is observed to estimate these parameters.

[0084] For example, in a scenario within the same electronic module field, the thermal response under different workloads is simulated, and the thermal resistance is calculated to be approximately a few degrees per watt, with the heat capacity derived based on the time constant. This estimate can be used to optimize module design and improve thermal stability. Furthermore, in another embodiment, when considering the multi-layered structure of the module, the contribution of heat conduction can be further subdivided into inter-layer and intra-layer contributions. By refining the mesh density using the finite element method, a more accurate thermal network model is constructed, and a correction factor is introduced when estimating the final parameters to accommodate actual manufacturing deviations. This method demonstrates the versatility of the technical solution in the thermal management of electronic modules, enabling a reliable preliminary assessment of thermal parameters.

[0085] Understandably, the effect of this process is to provide accurate thermal parameter estimates, supporting iterative thermal design of the module without introducing subjective optimization terms.

[0086] S106. Based on the preliminary estimate, extract time-domain features from the continuously acquired temperature signal sequence to determine the deviation vector of the parameter estimation.

[0087] By continuously acquiring temperature signal sequences, time-domain features are obtained to determine the initial deviation basis corresponding to the preliminary estimate. For the preliminary estimate, the temperature change trend is extracted from the acquired time-domain features to determine the initial deviation of the parameter estimation. Based on the initial deviation, a deviation correction process is used to process the continuously acquired data. This process obtains a deviation vector component by comparing the difference between the initial deviation and the continuously acquired data. If the deviation vector component exceeds a preset threshold, parameter optimization is performed to fuse the estimation deviation analysis. This parameter optimization determines the correction direction of the deviation vector by fusing the deviation distribution of the estimation deviation analysis based on the amplitude change of the deviation vector component. The temperature signal sequence and time-domain features are integrated from the correction direction to determine the deviation vector of the parameter estimation.

[0088] In one implementation, extracting time-domain features from continuously acquired temperature signal sequences first requires understanding the acquisition process of these signals. These sequences typically originate from sensors on industrial equipment such as boilers or motors, continuously recording temperature changes to reflect the equipment's operating status. Preliminary estimates can be temperature forecasts calculated based on historical data or models, such as initial parameters obtained through averaging. Extracting time-domain features from these preliminary estimates helps quantify the statistical properties of the signal, thereby identifying biases.

[0089] Specifically, time-domain feature extraction includes calculating metrics such as the mean, variance, and peak value of the signal.

[0090] For example, for a temperature signal sequence within an acquisition period, its arithmetic mean is first calculated as the mean feature, representing the central trend of the signal; then the variance is calculated to measure the dispersion of temperature fluctuations; furthermore, peak values, i.e., the maximum temperature points in the sequence, can be extracted to capture sudden changes. These features form a vector for subsequent deviation analysis. This extraction process ensures that key information is extracted from the original signal, supporting the optimization of parameter estimation. Further, the deviation vector of the parameter estimation is determined based on a comparison between the extracted time-domain features and the preliminary estimated values.

[0091] In one possible implementation, the bias vector is obtained by subtracting the expected features corresponding to the preliminary estimates element by element from the feature vector.

[0092] For example, if the mean of the initial estimate is 50 degrees Celsius, while the extracted actual mean is 52 degrees Celsius, the deviation is 2 degrees Celsius; similarly, variance and peak deviation are handled. This deviation vector quantifies the estimation error and can be used to correct model parameters, improving the accuracy of temperature predictions.

[0093] Preferably, in industrial temperature monitoring scenarios, multiple acquisition intervals can be used to demonstrate the versatility of the solution.

[0094] For example, in boiler temperature control, features are extracted using sequences acquired once per second, and deviation vectors are calculated to adjust heating parameters; similarly, in motor overheat monitoring, time-domain features are extracted using sequences acquired every minute to determine deviations and prevent faults. These embodiments remain within the field of temperature signal processing, avoiding cross-domain expansion.

[0095] It should be noted that the process of determining the deviation vector emphasizes iterative optimization.

[0096] In one embodiment, the initial estimate is first initialized, and then the error range is gradually reduced by extracting features and calculating the deviation multiple times.

[0097] For example, when the initial bias vector is large, the parameters of the estimation model can be adjusted, such as by increasing the weighting factor, until the bias vector approaches zero. This method achieves dynamic correction of parameter estimation in continuous signal processing. In another implementation, feature extraction under noise interference is considered.

[0098] For example, filtering preprocessing is applied to the temperature signal sequence to remove noise, and then time-domain features are extracted to ensure the accuracy of the deviation vector.

[0099] For example, in noisy environments, the filtered mean and variance are calculated and compared with the initial estimate to generate a deviation. This approach enhances the robustness of the solution in practical industrial applications.

[0100] Understandably, the logical sequence of these steps, from data acquisition to feature extraction and then to bias determination, forms a coherent process. Furthermore, in terms of effectiveness, this method enables precise adjustment of parameter estimates, such as reducing energy loss caused by temperature deviations during equipment maintenance.

[0101] For example, in a specific scenario, for a continuously acquired temperature signal sequence, the extracted feature vector includes a mean of 45 degrees Celsius, a variance of 3, and a peak value of 55 degrees Celsius, while the initial estimate is a mean of 40 degrees Celsius, a variance of 2, and a peak value of 50 degrees Celsius. The deviation vector is (5,1,5), which is used to guide subsequent parameter correction.

[0102] In one embodiment, the process is extended to temperature sequences fused from multiple sensors to extract comprehensive time-domain features and determine the overall deviation vector. This fusion is achieved through a weighted average of the signal sequences, supporting more comprehensive parameter estimation. Finally, after the deviation vector is determined, it can be input into a feedback loop to further optimize the initial estimate, ensuring system stability.

[0103] S107. If the absolute value of the elements in the deviation vector is greater than the preset accuracy threshold, then iteratively update the node connection weights in the thermal network model to obtain the refined set of thermal resistance and thermal capacity parameters.

[0104] The absolute values ​​of elements are extracted from the deviation vector. By comparing them with a preset precision threshold, it is determined whether an iteration is triggered, resulting in the set of node connection weights that need to be updated. For this set of node connection weights, a gradient descent algorithm is used to adjust the weight values, obtaining preliminary refined thermal resistance parameters. Based on these preliminary refined thermal resistance parameters, a heat capacity calculation module is integrated. This module determines the iteration direction by multiplying a preset initial heat capacity value with the thermal resistance parameter. If the iteration direction of the heat capacity parameter is related to the thermal resistance parameter deviation, multiple iterations are used to update and obtain integrated parameter pairs. Thermal balance is verified from these integrated parameter pairs by comparing the heat flow difference between parameter pairs with a preset balance threshold, resulting in a refined set of thermal resistance and heat capacity parameters.

[0105] In one implementation, the thermal network model is constructed as a network of nodes to simulate the heat transfer process in electronic devices.

[0106] Specifically, the model treats electronic components as nodes, with connections between nodes representing thermal resistance, and each node itself associated with a thermal capacity parameter. This structure allows the model to describe the flow of heat from a heat source to a heat dissipation path. For example, in simulating the thermal distribution of chips on a circuit board, nodes correspond to different components, and connection weights are initially set based on material properties. Furthermore, the deviation vector is a vector generated by comparing the model-predicted temperature with the actual measured temperature, where each element represents the temperature deviation at a specific node.

[0107] For example, in a thermal testing scenario for electronic devices, actual temperature data is first collected, and then a thermal network model is run to obtain predicted values. The deviation vector is the set of differences between the two. This vector helps quantify the accuracy of the model; if the absolute value of an element is greater than a preset accuracy threshold, such as 0.5 degrees Celsius, it indicates that the model parameters need further adjustment.

[0108] In one possible implementation, the iterative update process optimizes the node connection weights in the hot network model.

[0109] Specifically, the update uses the gradient descent method to adjust the weight values ​​based on the gradient information of the deviation vector.

[0110] For example, in a thermal network with multiple nodes, the initial weights correspond to the reciprocals of the thermal resistance. When the deviation is large, the deviation is gradually reduced by calculating the partial derivatives. This iteration is repeated until the absolute values ​​of all deviation elements are less than a threshold, thus obtaining a refined set of thermal resistance and thermal capacity parameters. In electronic device design, this refined parameter set can be used for more accurate thermal simulation, avoiding the risk of overheating.

[0111] Preferably, a regularization term can be introduced during the iteration process to prevent overfitting.

[0112] Specifically, regularization ensures that the updated weights remain physically reasonable by adding a weight penalty function; for example, the thermal resistance value is not lower than the material's minimum thermal conductivity. This approach enhances the model's robustness and enables it to handle the impact of noisy data in practical applications such as thermal management of mobile phone chips.

[0113] It's important to note that thermal resistance represents the resistance to heat transfer, while thermal capacity represents the node's ability to store heat. During the refinement process, these parameters are indirectly optimized through weight updates; for example, the initial thermal resistance is estimated based on Fourier's law, and after iteration, it better matches measured data. This parameter set allows the thermal network model to be used across different scenarios within the same domain, such as design optimization from desktop computers to portable devices.

[0114] For example, in one specific embodiment, for thermal simulation of a server chassis, a thermal network model is first established, including CPU nodes and fan nodes. Temperature data is collected to generate a deviation vector. If the absolute value exceeds a threshold, the connection weights are iteratively updated. After several iterations, the thermal resistance parameter is refined from the initial 10K / W to 8K / W, and the thermal capacity parameter is adjusted accordingly. This refinement improves the model's prediction accuracy, enabling more effective thermal design and reducing equipment failure rates. Furthermore, in another embodiment, applied to thermal management of embedded systems, the model nodes are extended to sensors and the enclosure. After determining the deviation vector, iterative updates focus on dynamic weight adjustments, such as adaptive modifications based on changes in ambient temperature. This approach demonstrates the flexibility of the technical solution, ensuring the accuracy of the parameter set under varying temperature conditions.

[0115] Understandably, the refined parameter set obtained through the aforementioned iterative mechanism supports multiple rounds of verification. In the field of electronic thermal simulation, this process not only refines the parameters but also provides a reproducible optimization path, such as rapidly iterating the model in prototype testing to improve design efficiency.

[0116] In one embodiment, the accuracy threshold is chosen based on application requirements, such as 0.1 degrees Celsius for high-precision scenarios. After iterative updates, the parameter set is used to generate heat map, assisting engineers in optimizing the layout. This objective and refined approach ensures the practical value of the heat network model in real-world business applications.

[0117] S108. By comparing the refined thermal resistance and thermal capacity parameter set with the initial temperature distribution data, the non-destructive compliance of the overall thermal characteristics is determined, and the final test evaluation report data is obtained.

[0118] An initial temperature map is obtained by fusing the thermal resistance refinement set and the thermal capacity parameter set, and a comparison value is obtained by determining the distribution data source. Based on the comparison value, a non-destructive compliance calculation is performed, wherein the non-destructive compliance calculation is obtained by the ratio of the comparison value to a preset standard deviation to determine the overall characteristic evaluation. From the overall characteristic evaluation, test evaluation items are obtained to obtain a report data chain. For the report data chain, the parameter set is fused with the temperature distribution domain. If the final judgment kernel exceeds a preset threshold, the non-destructive compliance of the thermal characteristics is determined.

[0119] In one implementation, the refined set of thermal resistance and thermal capacity parameters is compared with the initial temperature distribution data to determine the non-destructive compliance of the overall thermal characteristics, thereby obtaining the final test evaluation report data. This process begins with the collection of initial temperature distribution data. For example, in the thermal testing scenario of electronic equipment, an infrared thermal imager is used to record the temperature field distribution of the device under standard conditions. These data form a multidimensional array representing the temperature values ​​at different locations.

[0120] It should be noted that the initial temperature distribution data must be acquired under conditions free from external interference to provide a reliable baseline. Furthermore, the generation of the refined thermal resistance and thermal capacity parameter set is based on a heat transfer model.

[0121] Specifically, thermal resistance parameter represents the ability of a material or device to block heat flow, and is usually calculated by the ratio of heat flux density to temperature gradient, while heat capacity parameter reflects the capacity to store heat, and is determined by the product of specific heat capacity and mass.

[0122] In one possible implementation, the initially estimated parameters are refined, for example, by using iterative optimization methods to adjust the parameter values ​​to match the experimental observation data, thereby obtaining a refined parameter set. This refinement process can be understood as optimizing by minimizing an error function, such as comparing the difference between the simulated temperature distribution and the actual measured values ​​to ensure the accuracy of the parameter set.

[0123] For example, in thermal characteristic testing, a refined set of thermal resistance and thermal capacity parameters is applied to a thermal network model to simulate temperature distribution. The specific process involves constructing an equivalent thermal circuit, where thermal resistance corresponds to resistive elements and thermal capacity corresponds to capacitive elements. By inputting initial conditions, such as ambient temperature and heat source power, the model calculates predicted temperature distribution data. This is then compared with actual initial temperature distribution data, for example, by calculating the temperature difference at key points or an overall deviation index, such as root mean square error. This comparison helps determine the degree of non-destructive compliance, i.e., assessing whether the thermal characteristics remain intact and undamaged.

[0124] Preferably, a threshold comparison mechanism is used to determine the degree of non-destructive compliance.

[0125] Specifically, a compliance score is defined, for example, by calculating the percentage of deviation using a formula. If the score is lower than a preset threshold, such as 5%, it is considered compliant; otherwise, it indicates that there may be a defect.

[0126] In one embodiment, for thermal testing of semiconductor chips, the comparison results are quantified. For example, the refined thermal resistance parameter is 2.5 K / W, and the heat capacity is 1.2 J / K. A deviation of 3% compared to the initial temperature distribution indicates no damage to the thermal properties. This judgment process ensures the objectivity of the evaluation and directly supports report generation. In another embodiment, the process is extended to the thermal property testing scenario of composite materials, also limited to the field of non-destructive testing. The initial temperature distribution data can be obtained through laser scanning to form a high-resolution temperature map. The refined parameter set involves a multi-layer material model, such as calculating the thermal resistance and heat capacity of each layer. During comparison, a vector matching method is used to compare the simulated vector generated by the parameter set with the initial data vector, and the cosine similarity is calculated as a conformity index. If the similarity is higher than 0.95, it is determined to be non-destructive. Furthermore, based on the above comparison results, the final test evaluation report data is generated.

[0127] Specifically, the report includes a list of parameter values, a compliance score, and visualizations such as a temperature distribution comparison chart. This report data can be output via automated scripts, ensuring standardized formatting.

[0128] In one embodiment, the report also includes sensitivity analysis, such as adjusting the thermal resistance parameter and observing changes in compliance, thereby verifying the robustness of the judgment. This technical solution enables efficient non-destructive thermal property assessment, providing a reliable basis in the testing field.

[0129] It should be noted that the entire process emphasizes the iterative nature of parameter refinement. For example, initial parameters are initially estimated through finite element analysis and then refined until convergence. In this way, the non-destructive compliance assessment is more accurate, and the results are objectively presented in the report, such as a deviation distribution chart, to help users identify potential thermal problems.

[0130] For example, in the thermal testing of battery modules, initial temperature distribution data is recorded to show the temperature gradient on the battery surface. A refined set of thermal resistance parameters considers internal heat conduction paths; if the comparison shows a high degree of conformity, a report confirms the integrity of the thermal characteristics, avoiding the need for destructive testing. This implementation demonstrates the versatility of the solution within the same field.

[0131] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of the embodiments of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. A non-destructive intelligent thermal characteristic testing method for LED light source modules, characterized in that, include: A short-term thermal excitation is applied to the surface of the LED module by a preset pulse sequence to obtain initial temperature distribution data and determine the thermal state reference at the start of the excitation. Based on the initial temperature distribution data, a transient thermal analysis algorithm is used to process the continuously acquired temperature signal sequence to determine the dynamic trend of heat flow propagation within the module. If the dynamic trend exceeds a preset heat flux density threshold, the parameters of the pulse sequence are adjusted to obtain an optimized excitation parameter set. An adjusted temperature response curve is obtained from the optimized excitation parameter set, and the frequency components in the temperature response curve are decomposed to determine the heat conduction contribution of each layer of material. A thermal network model of the module is constructed based on the heat conduction contribution to obtain preliminary estimates of thermal resistance and heat capacity parameters. Time-domain features are extracted from the continuously acquired temperature signal sequence to determine the deviation vector of the preliminary estimates. If the absolute value of an element in the deviation vector is greater than a preset accuracy threshold, the node connection weights in the thermal network model are iteratively updated to obtain a refined set of thermal resistance and heat capacity parameters. The refined set of thermal resistance and heat capacity parameters is compared with the initial temperature distribution data to determine the non-destructive compliance of the overall thermal characteristics, and test evaluation report data is obtained.

2. The intelligent non-destructive testing method for the thermal characteristics of an LED light source module as described in claim 1, characterized in that, The step of applying short-term thermal excitation to the surface of the LED module using a preset pulse sequence, acquiring initial temperature distribution data, and determining the thermal state benchmark at the start of excitation includes: applying the short-term thermal excitation to the surface of the LED module using the pulse sequence, acquiring the temperature value of the LED module surface using an infrared thermal imager, and acquiring the initial temperature distribution data; calculating the standard deviation of the initial temperature distribution data as a uniformity index; if the uniformity index is lower than a preset threshold, reapplying the thermal excitation by adjusting the parameters of the pulse sequence, and acquiring the adjusted data; extracting timestamp information from the adjusted data to determine the thermal state benchmark at the start of excitation.

3. The intelligent non-destructive testing method for the thermal characteristics of an LED light source module as described in claim 1, characterized in that, The step of processing continuously acquired temperature signal sequences using a transient thermal analysis algorithm based on the initial temperature distribution data to determine the dynamic trend of heat flow propagation within the module includes: processing the continuously acquired temperature signal sequences using the finite difference method based on the initial temperature distribution data, calculating the temperature change within a time step, and obtaining the temperature sequence processing result; calculating the heat flow propagation trend within the module using the Fourier heat conduction equation based on the temperature sequence processing result, and obtaining the heat conduction simulation output; if the heat conduction simulation output exceeds a preset threshold, adjusting the heat flow propagation trend through signal noise filtering to obtain heat flow path tracking data; integrating data preprocessing steps from the heat flow path tracking data, determining the integrity of the heat distribution simulation, and obtaining the dynamic trend.

4. The intelligent non-destructive testing method for the thermal characteristics of an LED light source module as described in claim 1, characterized in that, If the dynamic change trend exceeds a preset heat flux density threshold, the parameters of the pulse sequence are adjusted to obtain an optimized excitation parameter set, including: acquiring real-time heat flux density data through a sensor to obtain the dynamic change trend; if the dynamic change trend exceeds the preset heat flux density threshold, adjusting the amplitude and duration of the pulse sequence to obtain adjusted sequence parameters; extracting key features from the adjusted sequence parameters to obtain the optimized excitation parameter set; and updating the heat flux monitoring mechanism based on the optimized excitation parameter set to obtain the final parameter set update result.

5. The intelligent non-destructive testing method for the thermal characteristics of an LED light source module as described in claim 1, characterized in that, The process of obtaining the adjusted temperature response curve from the optimized excitation parameter set, decomposing the frequency components in the temperature response curve, and determining the thermal conductivity contribution of each material layer includes: obtaining the adjusted temperature response curve from the optimized excitation parameter set; decomposing the frequency components in the adjusted temperature response curve using a Fourier transform algorithm to obtain a set of frequency components; obtaining a contribution index for the set of frequency components by comparing the amplitude of each frequency with a preset material thermal conductivity frequency correlation threshold; determining the response curve analysis result by matching the contribution index with a preset mapping table for material layer separation; and obtaining the thermal conductivity distribution characteristics of each material layer by superimposing the response curve analysis result with thermal conductivity frequency correlation distribution data.

6. A non-destructive intelligent thermal characteristic testing method for an LED light source module as described in any one of claims 1-5, characterized in that, The step of constructing a thermal network model of the module based on the heat conduction contribution to obtain preliminary estimates of thermal resistance and heat capacity parameters includes: obtaining heat flow path data from the module structure based on the heat conduction contribution to determine a preliminary value of thermal resistance; performing mesh generation on the preliminary value of thermal resistance using a finite element simulation algorithm to obtain a thermal network model framework; calculating the temperature gradient inside the module using the thermal network model framework to determine the range of heat capacity parameters; if the range of heat capacity parameters exceeds a preset threshold, adjusting the heat flow path data by correcting the contribution weight to obtain an optimized thermal resistance estimate; and extracting heat distribution features from the optimized thermal resistance estimate to obtain a preliminary estimate of the heat capacity parameters.

7. A non-destructive intelligent thermal characteristic testing method for an LED light source module as described in any one of claims 1-5, characterized in that, The step of extracting time-domain features from the continuously acquired temperature signal sequence to determine the deviation vector of the preliminary estimate includes: obtaining time-domain features from the continuously acquired temperature signal sequence to determine the deviation basis of the preliminary estimate; extracting the temperature change trend from the time-domain features for the preliminary estimate to determine the initial deviation; processing the continuously acquired temperature signal sequence using a deviation correction process based on the initial deviation to obtain a deviation vector component; if the deviation vector component exceeds a preset threshold, adjusting the parameter optimization to fuse the estimation deviation analysis and determining the correction direction of the deviation vector; and integrating the temperature signal sequence and the time-domain features from the correction direction to determine the deviation vector.

8. A non-destructive intelligent thermal characteristic testing method for an LED light source module as described in any one of claims 1-5, characterized in that, If the absolute value of an element in the deviation vector is greater than a preset precision threshold, the node connection weights in the thermal network model are iteratively updated to obtain a refined set of thermal resistance and thermal capacity parameters. This includes: extracting the absolute value of an element from the deviation vector, comparing it with the preset precision threshold to determine whether iteration is triggered, and obtaining the set of node connection weights to be updated; adjusting the weight values ​​using a gradient descent algorithm for the set of node connection weights to obtain preliminary refined thermal resistance parameters; determining the iteration direction by integrating the thermal capacity calculation module based on the preliminary refined thermal resistance parameters; if the iteration direction of the thermal capacity parameters is related to the deviation of the thermal resistance parameters, obtaining integrated parameter pairs through multiple iterations; verifying thermal balance from the integrated parameter pairs to obtain the refined set of thermal resistance and thermal capacity parameters.

Images