LED chip life prediction method and related equipment

Abstract

The invention relates to an LED chip life prediction method and related equipment, and the method comprises the following steps: carrying out the electroluminescent spectrum measurement of an LED chip, and obtaining an initial spectrum characteristic curve of the LED chip; performing Gaussian peak decomposition on the initial spectral characteristic curve to obtain multi-peak spectral parameters of the LED chip; performing spectrum evolution simulation on the LED chip through a spectrum reconstruction technology based on the multi-peak spectrum parameters to obtain a spectrum change trend chart of the LED chip; performing spectral feature extraction based on the spectral change trend chart to obtain a spectral attenuation feature vector of the LED chip; the technical problems that in the prior art, the service life of the LED chip is evaluated generally by combining an accelerated aging experiment with an empirical model, however, the method is long in period, high in cost and difficult to reflect the performance degradation process under the actual working condition in the prior art are solved in the prior art, the service life of the LED chip is predicted and analyzed on the basis of the spectral attenuation feature vector, and the service life prediction result of the LED chip is obtained.

Description

Technical Field

[0001] This invention relates to the field of LED chip technology, and in particular to a method and related equipment for predicting the lifetime of LED chips. Background Technology

[0002] With the rapid development of semiconductor lighting technology, LED chips have been widely used in general lighting, display backlighting, automotive lighting, and special lighting due to their advantages such as high energy efficiency, long lifespan, and environmental friendliness. However, the actual lifespan of LED chips is affected by various factors such as material aging, packaging degradation, and thermal stress accumulation, causing their light output performance to degrade over time, thus affecting the stability and reliability of lighting systems. Therefore, accurately predicting the lifespan of LED chips is crucial for product design optimization, quality control, and user experience. Traditionally, accelerated aging tests combined with empirical models are used for lifespan assessment, but this method is often time-consuming, costly, and fails to reflect the technical issues of performance degradation under actual operating conditions. Summary of the Invention

[0003] The main objective of this invention is to provide a method for predicting the lifespan of LED chips, which solves the technical problem that the traditional method of using accelerated aging experiments combined with empirical models for lifespan assessment is often time-consuming, costly, and difficult to reflect the performance degradation process under actual working conditions.

[0004] To achieve the above objectives, the present invention provides a method for predicting the lifetime of an LED chip, comprising the following steps: Electroluminescence spectroscopy was performed on the LED chip to obtain the initial spectral characteristic curve of the LED chip; Gaussian peak decomposition was performed on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip; Based on the multi-peak spectral parameters, the spectral evolution of the LED chip is simulated using spectral reconstruction technology to obtain a spectral change trend diagram of the LED chip; Based on the spectral change trend map, spectral features are extracted to obtain the spectral attenuation feature vector of the LED chip; The lifetime prediction analysis of the LED chip is performed based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip.

[0005] Furthermore, the step of performing electroluminescence spectroscopy measurement on the LED chip to obtain the initial spectral characteristic curve of the LED chip includes: A constant injection current is applied to the LED chip, and the emitted light wavelength of the LED chip after the injection current is scanned by a spectrum analyzer to obtain the original spectral data; The original spectral data is subjected to noise suppression processing using wavelet transform technology to generate the initial spectral characteristic curve of the LED chip.

[0006] Furthermore, the step of performing Gaussian peak decomposition on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip includes: The initial spectral characteristic curve is divided into continuous wavelength intervals to obtain multiple discrete wavelength data segments, and local extreme points of the multiple discrete wavelength data segments are detected to obtain spectral peak candidate points. The peak width of the candidate spectral peaks is estimated to obtain initial peak width parameters, and the initial spectral characteristic curve is fitted with a Gaussian function in a local region based on the initial peak width parameters to obtain multiple Gaussian fitting parameters. Residual analysis is performed on the multiple Gaussian fitting parameters to obtain a fitting residual sequence, and the multiple Gaussian fitting parameters are iteratively optimized based on the fitting residual sequence to obtain the optimized Gaussian peak parameters. The optimized Gaussian peak parameters are subjected to peak overlap discrimination to obtain the effective peak parameter screening results. Based on the effective peak parameter screening results, the initial spectral characteristic curve is reconstructed into a multi-peak spectrum to obtain the multi-peak spectral parameters of the LED chip.

[0007] Furthermore, the step of simulating the spectral evolution of the LED chip based on the multi-peak spectral parameters using spectral reconstruction technology to obtain a spectral change trend diagram of the LED chip includes: The multi-peak spectral parameters are subjected to time-series interpolation to obtain a spectral evolution reference point sequence. The spectral evolution reference point sequence is then reconstructed using spectral reconstruction technology to obtain a time-varying spectral distribution function. The spectral degradation behavior characteristics of the LED chip are analyzed based on the time-varying spectral distribution function. The spectral decay path of the LED chip is analyzed by the spectral degradation behavior characteristics to obtain a spectral evolution trajectory diagram. The spectral evolution trajectory diagram is then mapped to a time scale to obtain a lifetime acceleration factor sequence. Based on the lifetime acceleration factor sequence, the spectral evolution trajectory diagram is corrected on the time axis to obtain a family of real-time spectral change curves, and the family of real-time spectral change curves is extrapolated to obtain the spectral prediction boundary. A reliability analysis was performed on the spectral prediction boundary to obtain a spectral variation trend diagram of the LED chip.

[0008] Furthermore, the step of extracting spectral features based on the spectral change trend map to obtain the spectral attenuation feature vector of the LED chip includes: The spectral variation trend map is decomposed into a time series to obtain spectral trend components and periodic components. Based on the trend components and periodic components, spectral feature points are identified to obtain key attenuation nodes. The spectral degradation process of the LED chip is nonlinearly fitted based on the key attenuation node to obtain the spectral attenuation kinetic equation, and the spectral attenuation rate field is obtained by numerically solving the spectral attenuation kinetic equation. A spatial-temporal joint analysis of the spectral decay rate field is performed to obtain the spectral decay feature vector of the LED chip.

[0009] Furthermore, the step of performing lifetime prediction analysis on the LED chip based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip includes: The spectral attenuation feature vector is subjected to dimensional expansion processing to obtain an expanded feature space, and feature cross-correlation analysis is performed based on the expanded feature space to obtain a feature correlation matrix; Based on the feature correlation matrix, feature weights are assigned to the extended feature space to obtain a weighted feature vector, and the lifetime-sensitive features of the LED chip are analyzed based on the weighted feature vector. A time-series evolution simulation is performed on the lifetime-sensitive features to obtain a lifetime evolution trajectory map, and the failure modes of the LED chip are identified based on the lifetime evolution trajectory map. The remaining lifetime of the LED chip is predicted based on the failure mode, and the lifetime prediction result is obtained.

[0010] Furthermore, the identification of the failure mode of the LED chip based on the lifetime evolution trajectory map includes: Feature points are extracted from the lifetime evolution trajectory map to obtain a set of key trajectory feature points, and abrupt change points are detected from the set of key trajectory feature points to obtain spectral attenuation abrupt change points. Based on the spectral decay abrupt change point, the lifetime evolution trajectory diagram is divided into stages to obtain multiple spectral decay stages, and the decay rate of each stage is calculated to obtain the decay rate value of each stage. The attenuation rate values ​​at each stage are matched and compared with a preset failure feature library to obtain the feature matching degree, and the failure mode of the LED chip is determined based on the feature matching degree.

[0011] The present invention also provides an LED chip lifetime prediction device, comprising: The measurement module is used to perform electroluminescence spectrum measurement on the LED chip to obtain the initial spectral characteristic curve of the LED chip; The decomposition module is used to perform Gaussian peak decomposition on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip. The simulation module is used to simulate the spectral evolution of the LED chip based on the multi-peak spectral parameters using spectral reconstruction technology, and obtain a spectral change trend diagram of the LED chip. The extraction module is used to extract spectral features based on the spectral change trend map to obtain the spectral attenuation feature vector of the LED chip; The prediction module is used to perform lifetime prediction analysis on the LED chip based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip.

[0012] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0013] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0014] This invention provides a method for predicting the lifetime of an LED chip, comprising the following steps: performing electroluminescence spectroscopy measurement on the LED chip to obtain an initial spectral characteristic curve; performing Gaussian peak decomposition on the initial spectral characteristic curve to obtain multi-peak spectral parameters of the LED chip; based on the multi-peak spectral parameters, performing spectral evolution simulation on the LED chip using spectral reconstruction technology to obtain a spectral change trend map of the LED chip; extracting spectral features based on the spectral change trend map to obtain a spectral attenuation feature vector of the LED chip; and performing lifetime prediction analysis on the LED chip based on the spectral attenuation feature vector to obtain a lifetime prediction result of the LED chip. Traditionally, accelerated aging experiments combined with empirical models are commonly used for lifetime assessment, but this method is often time-consuming, costly, and fails to reflect the performance degradation process under actual operating conditions. This invention not only effectively simulates the spectral drift and attenuation phenomena that may occur in LED chips during actual use but also significantly shortens the time required by traditional experimental methods, thus improving efficiency. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the steps of an LED chip lifetime prediction method in one embodiment of the present invention; Figure 2 This is a structural block diagram of an LED chip lifetime prediction system according to an embodiment of the present invention; Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0016] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] like Figure 1 As shown, Figure 1 This invention provides a method for predicting the lifetime of an LED chip, comprising the following steps: Step S1: Perform electroluminescence spectroscopy measurement on the LED chip to obtain the initial spectral characteristic curve of the LED chip.

[0019] Specifically, electroluminescence spectroscopy measurements are performed on the LED chip to obtain its initial spectral characteristic curve. This process is fundamental to the entire lifetime prediction method, aiming to acquire the original spectral information of the LED chip under normal operating conditions, unaffected by aging. In practice, the LED chip is first placed in a constant-temperature, constant-current testing environment. A standard operating current is applied through a precision power supply to ensure stable light emission. Then, a high-resolution spectrometer is used to collect the emitted light signal from a specific angle. After calibration and background subtraction, electroluminescence spectral data reflecting the chip's light-emitting characteristics are obtained—the initial spectral characteristic curve. This curve not only includes macroscopic parameters such as the position, intensity, and full width at half maximum (FWHM) of the main emission peak but also implicitly contains the distribution characteristics of the carrier recombination region in the multi-quantum-well structure, providing the initial basis for subsequent Gaussian peak decomposition. For example, in chip screening scenarios before high-power white LED packaging, by performing the aforementioned electroluminescence spectroscopy measurements on multiple chips from the same production batch, the initial spectral characteristic curve of each chip can be obtained. This allows for the identification of individuals with poor spectral consistency, early elimination of chips with potential reliability risks, and ultimately, improvement in the lifetime uniformity and market competitiveness of the final product. Therefore, this step not only enables non-destructive characterization of the optical properties of LED chips, but also lays the data foundation for subsequent spectral evolution simulation based on multi-peak parameters.

[0020] Step S2: Perform Gaussian peak decomposition on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip.

[0021] Specifically, Gaussian peak decomposition is performed on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip. This process is a key step in further analyzing its internal structure after obtaining the initial spectral characteristic curve. In practice, based on the complex peak shape presented by the initial spectral characteristic curve, a Gaussian function or a Gaussian-Lorentz mixture function is used to mathematically fit the overlapping peaks in the curve. The fitting residual is optimized using the least squares method to gradually separate multiple independent emission peak components. Each peak corresponds to a different carrier recombination mechanism or the emission contribution of the quantum well region, thereby extracting parameters such as the center wavelength, full width at half maximum (FWHM), integral intensity, and peak area of ​​each peak. These parameters together constitute the multi-peak spectral parameters. This process can reveal the non-uniformity of the multi-quantum well structure in the LED chip, the distribution of defect states, and the influence of interface states, allowing subtle features that were originally masked in the overall spectrum to be quantitatively expressed. For example, in the reliability assessment of high-power white LED chips, Gaussian peak decomposition of the initial spectral characteristic curve can identify short-wavelength shoulder peaks or long-wavelength tail peaks that may exist in the starting spectrum in addition to the main peak. These peaks are often related to local stress concentration or impurity segregation, and their corresponding multi-peak spectral parameters can be used as sensitive indicators of early degradation for subsequent spectral evolution simulation, thereby improving the physical accuracy of lifetime prediction and early warning capabilities. Therefore, this step not only deepens the intrinsic analysis of the initial spectral characteristic curve but also provides refined input parameters for evolution trend prediction based on spectral reconstruction technology.

[0022] Step S3: Based on the multi-peak spectral parameters, the spectral evolution of the LED chip is simulated using spectral reconstruction technology to obtain a spectral change trend diagram of the LED chip.

[0023] Specifically, based on the multi-peak spectral parameters, the spectral evolution of the LED chip is simulated using spectral reconstruction technology to obtain a spectral change trend map of the LED chip. This step involves constructing a mathematical model that reflects the dynamic evolution of the LED chip's spectrum during long-term operation, based on the obtained multi-peak spectral parameters. In practice, the multi-peak spectral parameters are first input into the spectral reconstruction model as initial conditions. This model incorporates the degradation kinetics equations of semiconductor materials, carrier leakage mechanisms, defect state growth models, and bandgap effects caused by thermal stress. By setting degradation coefficients at different aging time points, the center wavelength, full width at half maximum (FWHM), and relative intensity of each Gaussian peak are dynamically adjusted. Then, the spectral morphology at different time stages is synthesized frame by frame, ultimately forming a continuously changing spectral sequence, i.e., the spectral change trend map. This process not only preserves the physical interpretability of the original multi-peak structure but also simulates typical aging phenomena such as main peak shift, shoulder peak enhancement, or short-band attenuation. For example, in the lifetime prediction scenario of high-power white LED chips, this method can simulate the phenomenon of redshift of the main peak and defect-related luminescence enhancement caused by the diffusion of In components in the quantum well during 5000 hours of continuous operation. This generates a clear spectral evolution path, providing a visualized and quantifiable basis for subsequent extraction of spectral attenuation feature vectors, thereby supporting high-precision lifetime prediction analysis. Therefore, this step achieves a leap from static parameters to dynamic behavior, enhancing the physical realism and foresight of the prediction model.

[0024] Step S4: Based on the spectral change trend map, spectral features are extracted to obtain the spectral attenuation feature vector of the LED chip.

[0025] Specifically, based on the spectral change trend map, spectral features are extracted to obtain the spectral decay feature vector of the LED chip. This process aims to extract key information characterizing the aging degree of the LED chip from a continuously changing spectral sequence. First, by analyzing the spectral change trend map, the changing patterns of the position, width, and intensity of each Gaussian peak at each time point are identified. These changing patterns directly reflect the physical degradation process of the LED chip during operation. Then, mathematical methods such as principal component analysis (PCA) or independent component analysis (ICA) are used to reduce the dimensionality of these changing patterns to capture the feature parameters that best distinguish different aging stages, thereby constructing a set of spectral decay feature vectors that can describe the spectral evolution trajectory. This step not only relies on the information contained in the original multi-peak spectral parameters but also incorporates phenomena observed during the actual aging process, such as the gradual redshift of the main peak position and the weakening of peak intensity over time, ensuring that the extracted feature vectors have high representativeness and stability. For example, in aging experiments of high-power white LED chips, by tracking their spectral evolution over a 5000-hour operating cycle and extracting feature vectors using the aforementioned method, we can observe a decrease in the relative intensity of the blue light component due to In component diffusion and an increase in the luminescence of the long-wavelength yellow phosphor. These feature vectors provide crucial data support for establishing accurate aging models, facilitating precise prediction and evaluation of LED lifetime. Therefore, this step serves as a vital bridge connecting dynamic spectral changes with aging mechanisms, providing a powerful tool for a deeper understanding of LED chip operating characteristics and optimized design.

[0026] Step S5: Perform lifetime prediction analysis on the LED chip based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip.

[0027] Specifically, the lifespan prediction analysis of the LED chip is performed based on the spectral attenuation feature vector to obtain the LED chip's lifespan prediction result. This process is the ultimate goal of the entire method, aiming to evaluate the working lifespan of the LED chip by analyzing the key feature vector extracted from the spectral change trend map. First, the spectral attenuation feature vector is compared and trained with a known aging dataset using machine learning algorithms or statistical models to establish a prediction model that reflects the relationship between spectral feature changes and actual lifespan. In this process, considering the influence of different factors such as temperature and current density on the LED aging rate, environmental parameters are usually introduced as correction factors to improve prediction accuracy. Next, the spectral attenuation feature vector under specific operating conditions is input into the model, and its evolution over time is calculated to estimate when the LED chip will reach a predetermined failure standard, such as a luminous flux drop to 70% of its initial value. For example, in the application scenario of high-power white LEDs, assuming that we have collected a large amount of aging data on the same model of LED under different operating conditions (such as high temperature and humidity, continuous high current drive), and established a prediction model based on this data. Once we obtain the spectral decay feature vector of a new batch of LED chips, we can input it into this model to predict the expected lifespan of this batch of LEDs under normal operating conditions. If the prediction results show that this batch of LEDs may experience significant performance degradation after 3000 hours, we can adjust the production process or propose improvement suggestions accordingly, thereby effectively extending product lifespan and enhancing market competitiveness. Therefore, this step not only realizes the conversion from spectral characteristics to lifespan prediction but also provides a scientific basis for optimizing LED design.

[0028] In a specific embodiment, the step of performing electroluminescence spectroscopy measurement on the LED chip to obtain the initial spectral characteristic curve of the LED chip includes: A constant injection current is applied to the LED chip, and the emitted light wavelength of the LED chip after the injection current is scanned by a spectrum analyzer to obtain the original spectral data; The original spectral data is subjected to noise suppression processing using wavelet transform technology to generate the initial spectral characteristic curve of the LED chip.

[0029] Specifically, electroluminescence spectroscopy measurements are performed on the LED chip to obtain its initial spectral characteristic curve. This process not only requires acquiring the optical response characteristics of the LED chip in its initial state but also emphasizes the accuracy of the measurement data and optimization of the signal-to-noise ratio to ensure the reliability of subsequent Gaussian peak decomposition and spectral evolution simulation. In practice, a constant injection current is first applied to the LED chip, and the emitted light wavelength is scanned using a spectrometer after the injection current is applied to obtain the raw spectral data. This operation is typically performed in a standard testing environment (such as a constant-temperature darkroom) to eliminate interference from external light and temperature fluctuations on the measurement results. The constant injection current setting must be configured based on the rated operating current of the LED chip to ensure the chip is in a stable light-emitting state and avoid spectral drift or transient effects caused by current fluctuations. After current injection, the LED chip produces electroluminescence, and the emitted light is collected by an integrating sphere or fiber optic probe and introduced into a spectrometer. The spectrometer continuously scans the entire visible or near-infrared emission band at high wavelength resolution (e.g., 0.1 nm) to obtain raw spectral data containing rich details. However, due to factors such as electronic noise in the detection system itself, environmental electromagnetic interference, and optical path scattering, the acquired raw spectral data often contains a certain degree of random noise and baseline drift. These interferences can mask weak shoulder peaks or fine structures in the spectrum, affecting the accuracy of subsequent multi-peak parameter extraction. Therefore, effective noise suppression processing of the raw spectral data is necessary. To this end, wavelet transform technology is used to denoise the raw spectral data. Compared with traditional moving average or Fourier filtering, this method has better time-frequency localization capabilities and can effectively distinguish the true spectral features and high-frequency noise components in the signal. In specific implementation, the raw spectral data is treated as a one-dimensional signal. An appropriate wavelet basis function (such as Daubechies or Symlet series) and decomposition level are selected to perform multi-scale wavelet decomposition on the signal. Then, soft or hard thresholding is applied to the wavelet coefficients of each level to suppress weak coefficients representing noise and retain significant coefficients reflecting true spectral changes. Finally, the signal is reconstructed through inverse wavelet transform, thereby generating a smooth initial spectral feature curve of the LED chip that retains key features. This curve not only accurately reflects the position and intensity of the main emission peak of the LED chip, but also clearly reveals any potential secondary peaks, tails, or asymmetric structures, providing high-quality input for subsequent Gaussian peak decomposition. For example, in the R&D and quality control of high-power white LED chips, after applying a constant current of 350 mA to a batch of chips, their original spectral data were obtained by scanning with a spectrometer. A weak short-wavelength shoulder peak was found near the 450nm blue light main peak, but it was difficult to confirm due to noise. After denoising using wavelet transform, the shoulder peak structure became clearly discernible, indicating the possible presence of local quantum well thickness inhomogeneity or interface defects.The initial spectral characteristic curves generated in this way can be further used for Gaussian peak decomposition to extract multi-peak spectral parameters, thereby supporting subsequent spectral evolution simulation and lifetime prediction analysis. This enables the identification of potential early degradation mechanisms of the chip and accurate lifetime prediction. Therefore, this step, by combining constant current drive, precise spectral scanning, and wavelet denoising techniques, significantly improves the quality of the initial spectral data, laying a solid data foundation for the entire lifetime prediction method.

[0030] In a specific embodiment, the step of performing Gaussian peak decomposition on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip includes: The initial spectral characteristic curve is divided into continuous wavelength intervals to obtain multiple discrete wavelength data segments, and local extreme points of the multiple discrete wavelength data segments are detected to obtain spectral peak candidate points. The peak width of the candidate spectral peaks is estimated to obtain initial peak width parameters, and the initial spectral characteristic curve is fitted with a Gaussian function in a local region based on the initial peak width parameters to obtain multiple Gaussian fitting parameters. Residual analysis is performed on the multiple Gaussian fitting parameters to obtain a fitting residual sequence, and the multiple Gaussian fitting parameters are iteratively optimized based on the fitting residual sequence to obtain the optimized Gaussian peak parameters. The optimized Gaussian peak parameters are subjected to peak overlap discrimination to obtain the effective peak parameter screening results. Based on the effective peak parameter screening results, the initial spectral characteristic curve is reconstructed into a multi-peak spectrum to obtain the multi-peak spectral parameters of the LED chip.

[0031] Specifically, Gaussian peak decomposition is performed on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip. This process is a key step in deeply analyzing the complexity of its internal light-emitting structure based on obtaining high-quality initial spectral data, aiming to separate and quantify multiple overlapping light-emitting components in the overall spectrum. In practice, the initial spectral characteristic curve is first divided into continuous wavelength intervals to obtain multiple discrete wavelength data segments. This division usually covers the entire emission spectrum range with a fixed step size (e.g., 0.5 nm) or an adaptive window method to ensure that no potential spectral feature regions are missed. On this basis, by detecting the local extrema of the multiple discrete wavelength data segments, all possible spectral intensity peak positions are identified. These points are spectral peak candidate points, which may correspond to physical mechanisms such as main quantum well emission, impurity energy level transitions, interface state recombination, or stress-induced sub-peaks. Subsequently, the peak width of the spectral peak candidate points is estimated, usually using the half-width at half-maximum (FWHM) estimation method or based on neighborhood curvature analysis, to preliminarily determine the initial peak width parameter of each candidate peak. This parameter reflects the carrier energy distribution width and material uniformity level of the light-emitting region. Next, based on the initial peak width parameter, Gaussian function fitting is performed in the local region of each candidate peak. That is, using the peak as the center wavelength and the initial peak width as the initial value, a single Gaussian model is constructed and fitted to the local spectral segment, thereby obtaining multiple Gaussian fitting parameters, including center wavelength, peak width, peak intensity, and area. To further improve the fitting accuracy, residual analysis needs to be performed on the multiple Gaussian fitting parameters to calculate the difference between the actual spectrum and the current multi-peak Gaussian superposition model, forming a fitting residual sequence. If there are still significant fluctuations or local peaks in the residuals, it indicates that the current fitting has not completely captured the true spectral structure. Therefore, it is necessary to adjust each Gaussian parameter based on the residual sequence, such as correcting the center position, broadening, or introducing new fitting peaks. Through iterative optimization, the overall residual is continuously reduced until convergence, and finally, the optimized Gaussian peak parameters are obtained. Subsequently, the optimized Gaussian peak parameters need to be evaluated for peak overlap to determine whether there are severe overlaps or physically unexplained redundant peaks between adjacent peaks. For example, if the distance between two peaks is less than 50% of their average half-width, they can be considered highly overlapping and require further merging or refitting to screen out effective peak parameters with independent physical meaning. Finally, based on the effective peak parameter screening results, all retained Gaussian peaks are linearly superimposed to reconstruct a complete multi-peak spectral morphology. This reconstructed spectrum should closely match the original initial spectral characteristic curve, and the parameter set of each component peak is the multi-peak spectral parameter of the LED chip.For example, in the application of high-power white LED chips, the initial spectrum of a certain chip shows a main peak at 450 nm. However, after processing using the above procedure, it was found that it actually consists of three Gaussian peaks: one at 448 nm (main quantum well emission), one at 440 nm (high In composition cluster emission), and one at 458 nm (interface defect-related emission). Peak overlap discrimination confirmed that all three are independent and effective. The resulting multi-peak spectral parameters not only revealed the material growth inhomogeneity but also provided refined input for subsequent spectral evolution simulation, significantly improving the interpretability and accuracy of lifetime prediction. Therefore, this step, through systematic interval division, candidate point detection, iterative optimization, and effectiveness screening, achieved accurate analysis of complex spectral structures.

[0032] In a specific embodiment, the step of simulating the spectral evolution of the LED chip based on the multi-peak spectral parameters using spectral reconstruction technology to obtain a spectral change trend diagram of the LED chip includes: The multi-peak spectral parameters are subjected to time-series interpolation to obtain a spectral evolution reference point sequence. The spectral evolution reference point sequence is then reconstructed using spectral reconstruction technology to obtain a time-varying spectral distribution function. The spectral degradation behavior characteristics of the LED chip are analyzed based on the time-varying spectral distribution function. The spectral decay path of the LED chip is analyzed by the spectral degradation behavior characteristics to obtain a spectral evolution trajectory diagram. The spectral evolution trajectory diagram is then mapped to a time scale to obtain a lifetime acceleration factor sequence. Based on the lifetime acceleration factor sequence, the spectral evolution trajectory diagram is corrected on the time axis to obtain a family of real-time spectral change curves, and the family of real-time spectral change curves is extrapolated to obtain the spectral prediction boundary. A reliability analysis was performed on the spectral prediction boundary to obtain a spectral variation trend diagram of the LED chip.

[0033] Specifically, based on the multi-peak spectral parameters, the spectral evolution simulation of the LED chip is performed using spectral reconstruction technology to obtain a spectral change trend map of the LED chip. This step involves constructing a predictive model that reflects the dynamic evolution of the LED chip's spectrum during long-term operation, based on the obtained accurate multi-peak spectral parameters. In practice, the multi-peak spectral parameters are first subjected to time-series interpolation. This process introduces aging time variables (such as 0 hours, 1000 hours, 2000 hours, etc.) to expand the static multi-peak parameters in the initial state into a discrete data point sequence that evolves over time, i.e., a spectral evolution reference point sequence. These reference points represent the expected changes in the center wavelength, intensity, and full width at half maximum (FWHM) of each Gaussian peak at different aging stages. Their construction relies on the support of historical accelerated aging experimental data or physical degradation mechanism models. Subsequently, spectral reconstruction calculations are performed on the spectral evolution reference point sequence using spectral reconstruction technology. This involves substituting the multi-peak parameters at each time point into a Gaussian superposition model to synthesize the complete spectral morphology at each corresponding moment, thereby generating a continuously changing time-varying spectral distribution function. This function not only describes the redshift or broadening of the main peak but also reflects complex behaviors such as shoulder peak enhancement and short-wavelength attenuation. Based on this, the spectral degradation behavior characteristics of the LED chip are analyzed using the time-varying spectral distribution function. For example, key degradation indicators such as the attenuation rate of the main emission peak intensity, the increasing trend of the proportion of long-wavelength defect emission, and the nonlinear expansion of the full width at half maximum (FWHM) over time are identified. These characteristics reveal the coupling effect of multiple failure mechanisms, including carrier leakage, quantum well degradation, and yellowing of the packaging material within the chip. Furthermore, the spectral degradation path of the LED chip is analyzed using these spectral degradation behavior characteristics. This involves tracing the priority order and spatial distribution evolution pattern of light intensity attenuation in each band, forming a spectral evolution trajectory extending from the initial state to the failure endpoint. This trajectory visually displays the evolution path of the LED chip's color coordinates, color rendering index, and light output characteristics throughout its entire lifespan. To make the trajectory map applicable to real-world applications, it needs to be time-scaled, converting the time axis under accelerated aging conditions in the laboratory to the equivalent time under normal operating conditions. An Arrhenius-like lifetime model is established by introducing stress factors such as temperature and current, calculating the lifetime acceleration factor sequence, and then correcting the time axis of the spectral evolution trajectory map based on this sequence. This generates a family of real-time spectral change curves reflecting actual operating conditions. These curves illustrate the spectral evolution process that the LED chip may undergo under different operating conditions. Finally, the real-time spectral change curves are extrapolated, and statistical regression or machine learning methods are used to predict the spectral morphology over a longer future period, determining the time point at which spectral performance degrades to a critical threshold (e.g., a 30% decrease in luminous flux or a color coordinate shift exceeding the B15 ellipse), thus forming the spectral prediction boundary.Next, a reliability analysis is performed on the spectral prediction boundary to evaluate the confidence interval and sources of uncertainty of the extrapolation results, eliminate outlier paths, and finally obtain a high-confidence, physically interpretable spectral change trend chart of the LED chip. For example, in the application scenario of high-power white LED chips, the multi-peak spectral parameters of a certain model of chip in its initial state show the presence of a main blue peak and a weak yellow-green defect peak. By simulating its evolution process over a 5000-hour operating cycle through the above process, it is predicted that the intensity of the defect peak will significantly increase over time, causing the overall spectrum to shift towards longer wavelengths and the color temperature to decrease, ultimately forming a clear spectral change trend chart, providing a scientific basis for product life rating and packaging process optimization. Therefore, this step realizes a complete closed loop from static parameters to dynamic evolution, and from laboratory data to actual life prediction.

[0034] In a specific embodiment, the step of extracting spectral features based on the spectral change trend map to obtain the spectral attenuation feature vector of the LED chip includes: The spectral variation trend map is decomposed into a time series to obtain spectral trend components and periodic components. Based on the trend components and periodic components, spectral feature points are identified to obtain key attenuation nodes. The spectral degradation process of the LED chip is nonlinearly fitted based on the key attenuation node to obtain the spectral attenuation kinetic equation, and the spectral attenuation rate field is obtained by numerically solving the spectral attenuation kinetic equation. A spatial-temporal joint analysis of the spectral decay rate field is performed to obtain the spectral decay feature vector of the LED chip.

[0035] Specifically, the extraction of spectral features based on the spectral change trend map to obtain the spectral attenuation feature vector of the LED chip is a key technical step in the entire lifetime prediction method, enabling the extraction of quantifiable and modelable degradation indicators from dynamic spectral evolution information. The core of this step lies in deeply exploring the inherent laws governing the evolution of the spectrum over time and transforming them into mathematical features that characterize the aging process of the LED chip. In practice, the spectral change trend map is first decomposed into a time series. This process treats the continuously changing spectral data as a two-dimensional spatiotemporal signal, where one dimension is wavelength and the other is time. By applying time-frequency analysis methods such as STL (Seasonal and Trend Decomposition using Loess) or wavelet decomposition to the light intensity change sequence at each wavelength point, the slowly changing spectral trend component and the potentially existing periodic fluctuation component are separated. The latter may originate from intermittent operating modes, thermal cycling, or periodic interference from the measurement system itself. The spectral trend component reflects the continuous attenuation behavior of the LED chip's light output performance during long-term operation, while the periodic component can be used to identify the fatigue accumulation effect under discontinuous operating conditions. Based on this, spectral feature points are identified using the trend and periodic components, i.e., turning points, inflection points, or abrupt changes with significant physical meaning are located on the wavelength-time plane. Examples include the time point when the main peak intensity first decreases by 5%, the wavelength position where the shoulder peak begins to enhance, or the critical moment when the full width at half maximum (FWHM) expands rapidly. These points collectively constitute key decay nodes, marking important stages in the chip degradation process, such as defect activation, increased carrier leakage, or the onset of carbonization in the packaging material. Subsequently, the spectral degradation process of the LED chip is nonlinearly fitted based on these key decay nodes. Considering that LED aging typically follows nonlinear laws such as exponential decay, power-law degradation, or hyperbolic functions, a suitable mathematical model (such as a double exponential decay model or the Weibull distribution function) is selected to fit the light intensity decay path at each wavelength point, thereby constructing a spectral decay dynamic equation that can describe the degradation behavior across the entire spectral domain. This equation not only includes parameters such as the decay coefficient and time constant but also implicitly contains the coupling relationship between degradation in different wavelength bands. The spectral attenuation kinetic equation is numerically solved, and numerical methods such as finite difference or Runge-Kutta are used to calculate the rate of change of light intensity at different time steps. This generates a spectral attenuation rate field covering the entire wavelength range and the entire time period. The rate field is represented in matrix form, with each element corresponding to the instantaneous attenuation rate at a specific wavelength and time point. This intuitively reveals which bands attenuate the fastest and when. For example, the short-wavelength blue light region may attenuate rapidly in the early stage, while the long-wavelength yellow light region increases in the later stage and then attenuates.Furthermore, a joint spatial-temporal analysis is performed on the spectral decay rate field. This involves treating the wavelength dimension as a spatial axis and constructing a two-dimensional feature space by combining it with the time axis. Principal component analysis (PCA), independent component analysis (ICA), or convolutional neural networks are used to extract the most discriminative patterns. For example, a typical degradation mode of "rapid decay of blue light - plateau period of green light - slow rise of red light" is identified. Finally, these patterns are compressed into a set of low-dimensional numerical vectors, namely the spectral decay feature vector of the LED chip. This vector not only contains the spatial distribution characteristics of the degradation rate but also integrates temporal evolution dynamics information, exhibiting high compressibility and comparability. For example, in the application of high-power white LED chips, after the above process, the spectral decay feature vector of a certain chip showed that the decay rate at the main peak of 450 nm was significantly higher than that of the surrounding bands, and a secondary acceleration occurred after 2000 hours. Combined with the analysis of key decay nodes, it can be determined that this is caused by the combined effect of quantum well In-Ga separation and ohmic contact degradation. This feature vector was then input into the lifetime prediction model, which successfully predicted that the chip would reach the L70 lifetime endpoint at around 6500 hours, with an error of less than 8% compared with the actual aging experiment results. Therefore, this step, through systematic temporal decomposition, dynamic modeling, and joint feature extraction, achieves a high degree of abstraction and accurate characterization of the complex spectral evolution process, providing a solid data foundation and physical basis for subsequent lifetime prediction.

[0036] In a specific embodiment, the step of performing lifetime prediction analysis on the LED chip based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip includes: The spectral attenuation feature vector is subjected to dimensional expansion processing to obtain an expanded feature space, and feature cross-correlation analysis is performed based on the expanded feature space to obtain a feature correlation matrix; Based on the feature correlation matrix, feature weights are assigned to the extended feature space to obtain a weighted feature vector, and the lifetime-sensitive features of the LED chip are analyzed based on the weighted feature vector. A time-series evolution simulation is performed on the lifetime-sensitive features to obtain a lifetime evolution trajectory map, and the failure modes of the LED chip are identified based on the lifetime evolution trajectory map. The remaining lifetime of the LED chip is predicted based on the failure mode, and the lifetime prediction result is obtained.

[0037] Specifically, the lifetime prediction analysis of the LED chip based on the spectral attenuation feature vector, to obtain the lifetime prediction result of the LED chip, is an important step in realizing the transformation from abstract features to specific lifetime estimation. This process first requires dimensional expansion of the spectral attenuation feature vector, that is, increasing the dimension of the original feature space by introducing additional relevant variables or applying mathematical transformation methods (such as polynomial expansion, kernel trick, etc.), thereby obtaining an expanded feature space. For example, in the application scenario of high-power white LEDs, in addition to the directly extracted spectral attenuation features, factors such as ambient temperature, current density, and the aging degree of packaging materials can also be considered and added as new dimensions to the feature space. The purpose of doing so is to enrich the data expression capability, so that information that may have been hidden in the low-dimensional space can be revealed in a higher dimension, providing a richer information foundation for subsequent analysis. Next, feature cross-correlation analysis is performed based on the expanded feature space, the purpose of which is to explore the intrinsic connections and mutual influence relationships between different features. This usually involves calculating the correlation coefficient matrix or covariance matrix between features to identify which features have significant correlations or synergistic change trends. For example, it might be observed that as operating time increases, the spectral decay rate in the short-wavelength region exhibits a strong positive correlation with ambient temperature, while the long-wavelength region shows a relatively independent change pattern. Through such analysis, a feature correlation matrix can be constructed. This matrix not only reveals the strength of the relationships between features but also provides valuable clues for understanding the LED aging mechanism. Assigning feature weights to the extended feature space based on this feature correlation matrix aims to highlight features that play a crucial role in lifetime prediction while reducing the influence of noise or secondary factors. This step can be implemented using various machine learning algorithms, such as gradient boosting decision trees (GBDT), random forests (RF), or support vector machines (SVM), assigning weights based on the contribution of each feature to model performance. For instance, in the above application scenario, if analysis reveals that rapid decay in the short-wavelength region and increased current density at high temperatures are the primary lifetime limiting factors, these features will be given higher weights, forming a weighted feature vector. This vector can more accurately reflect the actual health status of the LED chip and its potential risks throughout its service life. Furthermore, analyzing the lifetime-sensitive characteristics of the LED chip based on the weighted feature vector means focusing on those key characteristics that have a decisive impact on lifetime. By modeling and simulating the time series of the weighted feature vector, the degradation path of the LED chip throughout its entire lifespan can be simulated, generating a lifetime evolution trajectory map. This image dynamically illustrates the gradual deterioration of the LED chip's health over time, including from initial small performance declines to later rapid functional loss.For example, for a specific batch of high-power white LEDs, the lifetime evolution trajectory shows that around 3000 hours of operation, the blue light output begins to weaken significantly, marking the arrival of the early failure stage; and after 5000 hours, the brightness across the entire spectrum shows irreversible degradation, indicating the approach of final failure. Finally, based on the lifetime evolution trajectory, the failure mode of the LED chip is identified, and the remaining lifetime of the LED chip is predicted accordingly, yielding the lifetime prediction result. The identification of failure modes relies on understanding specific patterns and inflection points on the lifetime evolution trajectory, such as whether there is an accelerated decay stage or whether multi-stage degradation behavior occurs. Once the failure mode is determined, corresponding mathematical models (such as the Arrhenius model, power-law model, etc.) can be used to predict the rate and endpoint of future performance degradation. For example, in the aforementioned case, if the LED chip is observed to have experienced three stages—an initial stable period, a rapid mid-term decay period, and a slow but continuous final performance loss—the remaining lifetime can be estimated based on the typical degradation patterns of each stage. Assuming the chip has already run for 4000 hours and is in a rapid degradation phase, with an estimated 1500 hours remaining to reach the L70 standard (i.e., initial luminous flux decreasing to 70%), then its remaining lifetime is predicted to be 1500 hours. This systematic process not only accurately predicts the expected lifetime of a single LED chip but also provides guidance for optimizing design and improving manufacturing processes, thereby enhancing product reliability and durability.

[0038] In a specific embodiment, identifying the failure mode of the LED chip based on the lifetime evolution trajectory map includes: Feature points are extracted from the lifetime evolution trajectory map to obtain a set of key trajectory feature points, and abrupt change points are detected from the set of key trajectory feature points to obtain spectral attenuation abrupt change points. Based on the spectral decay abrupt change point, the lifetime evolution trajectory diagram is divided into stages to obtain multiple spectral decay stages, and the decay rate of each stage is calculated to obtain the decay rate value of each stage. The attenuation rate values ​​at each stage are matched and compared with a preset failure feature library to obtain the feature matching degree, and the failure mode of the LED chip is determined based on the feature matching degree.

[0039] Specifically, identifying the failure modes of the LED chip based on the lifetime evolution trajectory map is a crucial step in the entire lifetime prediction process, mapping the degradation path to the physical failure mechanism. Its purpose is to clarify the aging stages experienced by the LED chip during long-term operation and the dominant failure mechanisms through refined analysis of the lifetime evolution trajectory map. In practice, the lifetime evolution trajectory map is first processed by feature point extraction. This involves discretizing the continuous degradation curve into representative key nodes, including but not limited to initial decay inflection points, points of significant slope change, plateau start and end points, and accelerated degradation initiation points, thus forming a set of key feature points for the trajectory. This extraction process can employ numerical differentiation combined with curvature analysis to identify locations where the second derivative changes significantly, or use wavelet transform to detect singularities in the signal, ensuring that the extracted feature points accurately reflect the structural transformations during the degradation process. Based on this, abrupt change point detection is performed on the set of key feature points to identify locations of discontinuous jumps or sudden rate changes during spectral decay, i.e., spectral decay abrupt change points. These abrupt changes often correspond to physical mutations occurring within the chip, such as metal electrode diffusion penetrating the quantum well, microcracks in the encapsulating silicone leading to moisture intrusion, or irreversible thermal degradation caused by localized hotspots. Abrupt change detection can be achieved using statistical methods such as CUSUM (cumulative sum control chart) or Bayesian change point detection algorithms. Setting reasonable thresholds distinguishes between normal fluctuations and true mutations, ensuring the reliability of the detection results. Subsequently, based on the spectral decay abrupt changes, the lifetime evolution trajectory is divided into stages, decomposing the entire degradation process into multiple time intervals with different degradation characteristics, i.e., multiple spectral decay stages. For example, the first stage may exhibit slow linear decay, mainly caused by the slow growth of carrier nonradiative recombination centers; the second stage shows exponentially accelerated decay, possibly related to ohmic contact degradation or a decrease in p-type layer hole injection efficiency; the third stage may be a plateau period, indicating that the system has entered a relatively stable state. The decay rate is calculated for each of the multiple spectral decay stages. Specifically, a function (such as a linear, exponential, or power-law function) of light intensity versus time is fitted within each stage, and its slope or decay constant is obtained to determine the decay rate value for each stage. These values ​​quantify the severity of aging at different stages, providing a quantitative basis for subsequent pattern recognition. Furthermore, the decay rate values ​​for each stage are matched against a preset failure feature library. This feature library is a database constructed from a large amount of historical experimental data, storing typical multi-stage decay rate combinations and their temporal distribution characteristics corresponding to various known failure modes (such as "quantum well In composition segregation," "n-GaN layer dislocation multiplication," and "silver reflector layer sulfidation"). The matching process can employ metrics such as Euclidean distance, dynamic time warping (DTW), or cosine similarity to calculate the feature matching degree between the decay rate mode of the current chip under test and each mode in the library. A higher matching degree indicates a more similar degradation behavior between the two.Finally, the failure mode of the LED chip is determined based on the feature matching degree, and the candidate mode with the highest matching degree is usually selected as the final judgment result. For example, in the application scenario of high-power white LED chips, after the lifetime evolution trajectory of a certain chip is processed by the above process, two spectral decay abrupt change points are identified: the first occurs at 1200 hours of operation, corresponding to a sharp drop of 8% in the intensity of the blue light main peak; the second occurs at 3800 hours, accompanied by an abnormal enhancement of the yellow light component. Based on this, it is divided into three decay stages, and the calculated decay rate is 0.015% / h in the first stage, rising to 0.042% / h in the second stage, and 0.028% / h in the third stage. Comparing this decay rate sequence with a preset failure feature library, it is found that its matching degree with the typical mode of "quantum well interface defect activation leading to carrier leakage" reaches 92%, which is much higher than other modes. Therefore, it is determined that the main failure mode of the chip is nonradiative recombination enhancement caused by internal defects. This result not only provides mechanistic support for lifetime prediction, but also provides a clear direction for improving epitaxial growth processes and enhancing chip reliability. Therefore, this step, through systematic feature extraction, stage division and pattern matching, achieves accurate identification from degradation trajectory to physical failure mechanism, significantly improving the scientific nature and engineering practicality of lifetime prediction.

[0040] The above describes the LED chip lifetime prediction method in the embodiments of the present invention. The following describes the LED chip lifetime prediction system in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the LED chip lifetime prediction system in this invention includes: Measurement module 21 is used to perform electroluminescence spectrum measurement on the LED chip to obtain the initial spectral characteristic curve of the LED chip; Decomposition module 22 is used to perform Gaussian peak decomposition on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip; The simulation module 23 is used to simulate the spectral evolution of the LED chip based on the multi-peak spectral parameters using spectral reconstruction technology, and obtain a spectral change trend diagram of the LED chip. Extraction module 24 is used to extract spectral features based on the spectral change trend map to obtain the spectral attenuation feature vector of the LED chip; The prediction module 25 is used to perform lifetime prediction analysis on the LED chip based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip.

[0041] In this embodiment, the specific implementation of each unit in the above device embodiment is described in the above method embodiment, and will not be repeated here.

[0042] Reference Figure 3This invention also provides a computer device whose internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0043] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0044] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0045] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0046] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0047] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for predicting the lifetime of an LED chip, characterized in that, Includes the following steps: Electroluminescence spectroscopy was performed on the LED chip to obtain the initial spectral characteristic curve of the LED chip; Gaussian peak decomposition was performed on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip; Based on the multi-peak spectral parameters, the spectral evolution of the LED chip is simulated using spectral reconstruction technology to obtain a spectral change trend diagram of the LED chip; Based on the spectral change trend map, spectral features are extracted to obtain the spectral attenuation feature vector of the LED chip; The lifetime prediction analysis of the LED chip is performed based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip.

2. The method for predicting the lifetime of an LED chip according to claim 1, characterized in that, The step of performing electroluminescence spectroscopy measurement on the LED chip to obtain the initial spectral characteristic curve of the LED chip includes: A constant injection current is applied to the LED chip, and the emitted light wavelength of the LED chip after the injection current is scanned by a spectrum analyzer to obtain the original spectral data; The original spectral data is subjected to noise suppression processing using wavelet transform technology to generate the initial spectral characteristic curve of the LED chip.

3. The method for predicting the lifetime of an LED chip according to claim 1, characterized in that, The step of performing Gaussian peak decomposition on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip includes: The initial spectral characteristic curve is divided into continuous wavelength intervals to obtain multiple discrete wavelength data segments, and local extreme points of the multiple discrete wavelength data segments are detected to obtain spectral peak candidate points. The peak width of the candidate spectral peaks is estimated to obtain initial peak width parameters, and the initial spectral characteristic curve is fitted with a Gaussian function in a local region based on the initial peak width parameters to obtain multiple Gaussian fitting parameters. Residual analysis is performed on the multiple Gaussian fitting parameters to obtain a fitting residual sequence, and the multiple Gaussian fitting parameters are iteratively optimized based on the fitting residual sequence to obtain the optimized Gaussian peak parameters. The optimized Gaussian peak parameters are subjected to peak overlap discrimination to obtain the effective peak parameter screening results. Based on the effective peak parameter screening results, the initial spectral characteristic curve is reconstructed into a multi-peak spectrum to obtain the multi-peak spectral parameters of the LED chip.

4. The method for predicting the lifetime of an LED chip according to claim 1, characterized in that, The step of simulating the spectral evolution of the LED chip using spectral reconstruction technology based on the multi-peak spectral parameters to obtain a spectral change trend diagram of the LED chip includes: The multi-peak spectral parameters are subjected to time-series interpolation to obtain a spectral evolution reference point sequence. The spectral evolution reference point sequence is then reconstructed using spectral reconstruction technology to obtain a time-varying spectral distribution function. The spectral degradation behavior characteristics of the LED chip are analyzed based on the time-varying spectral distribution function. The spectral decay path of the LED chip is analyzed by the spectral degradation behavior characteristics to obtain a spectral evolution trajectory diagram. The spectral evolution trajectory diagram is then mapped to a time scale to obtain a lifetime acceleration factor sequence. Based on the lifetime acceleration factor sequence, the spectral evolution trajectory diagram is corrected on the time axis to obtain a family of real-time spectral change curves, and the family of real-time spectral change curves is extrapolated to obtain the spectral prediction boundary. A reliability analysis was performed on the spectral prediction boundary to obtain a spectral variation trend diagram of the LED chip.

5. The method for predicting the lifetime of an LED chip according to claim 1, characterized in that, The step of extracting spectral features based on the spectral change trend map to obtain the spectral attenuation feature vector of the LED chip includes: The spectral variation trend map is decomposed into a time series to obtain spectral trend components and periodic components, and spectral feature points are identified based on the trend components and periodic components to obtain key attenuation nodes; The spectral degradation process of the LED chip is nonlinearly fitted based on the key attenuation node to obtain the spectral attenuation kinetic equation, and the spectral attenuation rate field is obtained by numerically solving the spectral attenuation kinetic equation. A spatial-temporal joint analysis of the spectral decay rate field is performed to obtain the spectral decay feature vector of the LED chip.

6. The method for predicting the lifetime of an LED chip according to claim 1, characterized in that, The lifetime prediction analysis of the LED chip based on the spectral attenuation feature vector, to obtain the lifetime prediction result of the LED chip, includes: The spectral attenuation feature vector is subjected to dimensional expansion processing to obtain an expanded feature space, and feature cross-correlation analysis is performed based on the expanded feature space to obtain a feature correlation matrix; Based on the feature correlation matrix, feature weights are assigned to the extended feature space to obtain a weighted feature vector, and the lifetime-sensitive features of the LED chip are analyzed based on the weighted feature vector. A time-series evolution simulation is performed on the lifetime-sensitive features to obtain a lifetime evolution trajectory map, and the failure modes of the LED chip are identified based on the lifetime evolution trajectory map. The remaining lifetime of the LED chip is predicted based on the failure mode, and the lifetime prediction result is obtained.

7. The method for predicting the lifetime of an LED chip according to claim 6, characterized in that, The identification of failure modes of the LED chip based on the lifetime evolution trajectory map includes: Feature points are extracted from the lifetime evolution trajectory map to obtain a set of key trajectory feature points, and abrupt change points are detected from the set of key trajectory feature points to obtain spectral attenuation abrupt change points. Based on the spectral decay abrupt change point, the lifetime evolution trajectory diagram is divided into stages to obtain multiple spectral decay stages, and the decay rate of each stage is calculated to obtain the decay rate value of each stage. The attenuation rate values ​​at each stage are matched and compared with a preset failure feature library to obtain the feature matching degree, and the failure mode of the LED chip is determined based on the feature matching degree.

8. A device for predicting the lifetime of an LED chip, characterized in that, include: The measurement module is used to perform electroluminescence spectrum measurement on the LED chip to obtain the initial spectral characteristic curve of the LED chip; The decomposition module is used to perform Gaussian peak decomposition on the initial spectral characteristic curve to obtain the multi-peak spectral parameters of the LED chip. The simulation module is used to simulate the spectral evolution of the LED chip based on the multi-peak spectral parameters using spectral reconstruction technology, and obtain a spectral change trend diagram of the LED chip. The extraction module is used to extract spectral features based on the spectral change trend map to obtain the spectral attenuation feature vector of the LED chip; The prediction module is used to perform lifetime prediction analysis on the LED chip based on the spectral attenuation feature vector to obtain the lifetime prediction result of the LED chip.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images